โŒ

Normal view

There are new articles available, click to refresh the page.
Yesterday โ€” 5 December 2025Main stream

Agents-as-a-service are poised to rewire the software industry and corporate structures

5 December 2025 at 05:00

This was the year of AI agents. Chatbots that simply answered questions are now evolving into autonomous agents that can carry out tasks on a userโ€™s behalf, so enterprises continue to invest in agentic platforms as transformation evolves. Software vendors are investing in it as fast as they can, too.

According to a National Research Group survey of more than 3,000 senior leaders, more than half of executives say their organization is already using AI agents. Of the companies that spend no less than half their AI budget on AI agents, 88% say theyโ€™re already seeing ROI on at least one use case, with top areas being customer service and experience, marketing, cybersecurity, and software development.

On the software provider side, Gartner predicts 40% of enterprise software applications in 2026 will include agentic AI, up from less than 5% today. And agentic AI could drive approximately 30% of enterprise application software revenue by 2035, surpassing $450 billion, up from 2% in 2025. In fact, business users might not have to interact directly with the business applications at all since AI agent ecosystems will carry out user instructions across multiple applications and business functions. At that point, a third of user experiences will shift from native applications to agentic front ends, Gartner predicts.

Itโ€™s already starting. Most enterprise applications will have embedded assistants, a precursor to agentic AI, by the end of this year, adds Gartner.

IDC has similar predictions. By 2028, 45% of IT product and service interactions will use agents as the primary interface, the firm says. Thatโ€™ll change not just how companies work, but how CIOs work as well.

Agents as employees

At financial services provider OneDigital, chief product officer Vinay Gidwaney is already working with AI agents, almost as if they were people.

โ€œWe decided to call them AI coworkers, and we set up an AI staffing team co-owned between my technology team and our chief people officer and her HR team,โ€ he says. โ€œThat team is responsible for hiring AI coworkers and bringing them into the organization.โ€ You heard that right: โ€œhiring.โ€

The first step is to sit down with the business leader and write a job description, which is fed to the AI agent, and then it becomes known as an intern.

โ€œWe have a lot of interns weโ€™re testing at the company,โ€ says Gidwaney. โ€œIf they pass, they get promoted to apprentices and we give them our best practices, guardrails, a personality, and human supervisors responsible for training them, auditing what they do, and writing improvement plans.โ€

The next promotion is to a full-time coworker, and it becomes available to be used by anyone at the company.

โ€œAnyone at our company can go on the corporate intranet, read the skill sets, and get ice breakers if they donโ€™t know how to start,โ€ he says. โ€œYou can pick a coworker off the shelf and start chatting with them.โ€

For example, thereโ€™s Ben, a benefits expert whoโ€™s trained on everything having to do with employee benefits.

โ€œWe have our employee benefits consultants sitting with clients every day,โ€ Gidwaney says. โ€œBen will take all the information and help the consultants strategize how to lower costs, and how to negotiate with carriers. Heโ€™s the consultantsโ€™ thought partner.โ€

There are similar AI coworkers working on retirement planning, and on property and casualty as well. These were built in-house because theyโ€™re core to the companyโ€™s business. But there are also external AI agents who can provide additional functionality in specialized yet less core areas, like legal or marketing content creation. In software development, OneDigital uses third-party AI agents as coding assistants.

When choosing whether to sign up for these agents, Gidwaney says he doesnโ€™t think of it the way he thinks about licensing software, but more to hiring a human consultant or contractor. For example, will the agent be a good cultural fit?

But in some cases, itโ€™s worse than hiring humans since a bad human hire who turns out to be toxic will only interact with a small number of other employees. But an AI agent might interact with thousands of them.

โ€œYou have to apply the same level of scrutiny as how you hire real humans,โ€ he says.

A vendor who looks like a technology company might also, in effect, be a staffing firm. โ€œThey look and feel like humans, and you have to treat them like that,โ€ he adds.

Another way that AI agents are similar to human consultants is when they leave the company, they take their expertise with them, including what they gained along the way. Data can be downloaded, Gidwaney says, but not necessarily the fine-tuning or other improvements the agent received. Realistically, there might not be any practical way to extract that from a third-party agent, and that could lead to AI vendor lock-in.

Edward Tull, VP of technology and operations at JBGoodwin Realtors, says he, too, sees AI agents as something akin to people. โ€œI see it more as a teammate,โ€ he says. โ€œAs we implement more across departments, I can see these teammates talking to each other. It becomes almost like a person.โ€

Today, JBGoodwin uses two main platforms for its AI agents. Zapier lets the company build its own and HubSpot has its own AaaS, and theyโ€™re already pre-built. โ€œThere are lead enrichment agents and workflow agents,โ€ says Tull.

And the company is open to using more. โ€œIn accounting, if someone builds an agent to work with this particular type of accounting software, we might hire that agent,โ€ he says. โ€œOr a marketing coordinator that we could hire thatโ€™s built and ready to go and connected to systems we already use.โ€

With agents, his job is becoming less about technology and more about management, he adds. โ€œItโ€™s less day-to-day building and more governance, and trying to position the company to be competitive in the world of AI,โ€ he says.

Heโ€™s not the only one thinking of AI agents as more akin to human workers than to software.

โ€œWith agents, because the technology is evolving so far, itโ€™s almost like youโ€™re hiring employees,โ€ says Sheldon Monteiro, chief product officer at Publicis Sapient. โ€œYou have to determine whom to hire, how to train them, make sure all the business units are getting value out of them, and figure when to fire them. Itโ€™s a continuous process, and this is very different from the past, where I make a commitment to a platform and stick with it because the solution works for the business.โ€

This changes how the technology solutions are managed, he adds. What companies will need now is a CHRO, but for agentic employees.

Managing outcomes, not persons

Vituity is one of the largest national, privately-held medical groups, with 600 hospitals, 13,800 employees, and nearly 14 million patients. The company is building its own AI agents, but is also using off-the-shelf ones, as AaaS. And AI agents arenโ€™t people, says CIO Amith Nair. โ€œThe agent has no feelings,โ€ he says. โ€œAGI isnโ€™t here yet.โ€

Instead, it all comes down to outcomes, he says. โ€œIf you define an outcome for a task, thatโ€™s the outcome youโ€™re holding that agent to.โ€ And that part isnโ€™t different to holding employees accountable to an outcome. โ€œBut you donโ€™t need to manage the agent,โ€ he adds. โ€œTheyโ€™re not people.โ€

Instead, the agent is orchestrated and you can plug and play them. โ€œIt needs to understand our business model and our business context, so you ground the agent to get the job done,โ€ he says.

For mission-critical functions, especially ones related to sensitive healthcare data, Vituity is building its own agents inside a HIPAA-certified LLM environment using the Workato agent development platform and the Microsoft agentic platform.

For other functions, especially ones having to do with public data, Vituity uses off-the-shelf agents, such as ones from Salesforce and Snowflake. The company is also using Claude with GitHub Copilot for coding. Nair can already see that agentic systems will change the way enterprise software works.

โ€œMost of the enterprise applications should get up to speed with MCP, the integration layer for standardization,โ€ he says. โ€œIf they donโ€™t get to it, itโ€™s going to become a challenge for them to keep selling their product.โ€

A company needs to be able to access its own data via an MCP connector, he says. โ€œAI needs data, and if they donโ€™t give you an MCP, you just start moving it all to a data warehouse,โ€ he adds.

Sharp learning curve

In addition to providing a way to store and organize your data, enterprise software vendors also offer logic and functionality, and AI will soon be able to handle that as well.

โ€œAll you need is a good workflow engine where you can develop new business processes on the fly, so it can orchestrate with other agents,โ€ Nair says. โ€œI donโ€™t think weโ€™re too far away, but weโ€™re not there yet. Until then, SaaS vendors are still relevant. The question is, can they charge that much money anymore.โ€

The costs of SaaS will eventually have to come down to the cost of inference, storage, and other infrastructure, but they canโ€™t survive the way theyโ€™re charging now he says. So SaaS vendors are building agents to augment or replace their current interfaces. But that approach itself has its limits. Say, for example, instead of using Salesforceโ€™s agent, a company can use its own agents to interact with the Salesforce environment.

โ€œItโ€™s already happening,โ€ Nair adds. โ€œMy SOC agent is pulling in all the log files from Salesforce. Theyโ€™re not providing me anything other than the security layer they need to protect the data that exists there.โ€

AI agents are set to change the dynamic between enterprises and software vendors in other ways, too. One major difference between software and agents is software is well-defined, operates in a particular way, and changes slowly, says Jinsook Han, chief of strategy, corporate development, and global agentic AI at Genpact.

โ€œBut we expect when the agent comes in, itโ€™s going to get smarter every day,โ€ she says. โ€œThe world will change dramatically because agents are continuously changing. And the expectations from the enterprises are also being reshaped.โ€

Another difference is agents can more easily work with data and systems where they are. Take for example a sales agent meeting with customers, says Anand Rao, AI professor at Carnegie Mellon University. Each salesperson has a calendar where all their meetings are scheduled, and they have emails, messages, and meeting recordings. An agent can simply access those emails when needed.

โ€œWhy put them all into Salesforce?โ€ Rao asks. โ€œIf the idea is to do and monitor the sale, it doesnโ€™t have to go into Salesforce, and the agents can go grab it.โ€

When Rao was a consultant having a conversation with a client, heโ€™d log it into Salesforce with a note, for instance, saying the client needs a white paper from the partner in charge of quantum.

With an agent taking notes during the meeting, it can immediately identify the action items and follow up to get the white paper.

โ€œRight now weโ€™re blindly automating the existing workflow,โ€ Rao says. โ€œBut why do we need to do that? Thereโ€™ll be a fundamental shift of how we see value chains and systems. Weโ€™ll get rid of all the intermediate steps. Thatโ€™s the biggest worry for the SAPs, Salesforces, and Workdays of the world.โ€

Another aspect of the agentic economy is instead of a human employee talking to a vendorโ€™s AI agent, a company agent can handle the conversation on the employeeโ€™s behalf. And if a company wants to switch vendors, the experience will be seamless for employees, since they never had to deal directly with the vendor anyway.

โ€œI think thatโ€™s something thatโ€™ll happen,โ€ says Ricardo Baeza-Yates, co-chair of theย  US technology policy committee at the Association for Computing Machinery. โ€œAnd it makes the market more competitive, and makes integrating things much easier.โ€

In the short term, however, it might make more sense for companies to use the vendorsโ€™ agents instead of creating their own.

โ€œI recommend people donโ€™t overbuild because everything is moving,โ€ says Bret Greenstein, CAIO at West Monroe Partners, a management consulting firm. โ€œIf you build a highly complicated system, youโ€™re going to be building yourself some tech debt. If an agent exists in your application and itโ€™s localized to the data in that application, use it.โ€

But over time, an agent thatโ€™s independent of the application can be more effective, he says, and thereโ€™s a lot of lock-in that goes into applications. โ€œItโ€™s going to be easier every day to build the agent you want without having to buy a giant license. โ€œThe effort to get effective agents is dropping rapidly, and the justification for getting expensive agents from your enterprise software vendors is getting less,โ€ he says.

The future of software

According to IDC, pure seat-based pricing will be obsolete by 2028, forcing 70% of vendors to figure out new business models.

With technology evolving as quickly as it is, JBGoodwin Realtors has already started to change its approach to buying tech, says Tull. It used to prefer long-term contracts, for example but thatโ€™s not the case anymore โ€œYou save more if you go longer, but Iโ€™ll ask for an option to re-sign with a cap,โ€ he says.

That doesnโ€™t mean SaaS will die overnight. Companies have made significant investments in their current technology infrastructure, says Patrycja Sobera, SVP of digital workplace solutions at Unisys.

โ€œTheyโ€™re not scrapping their strategies around cloud and SaaS,โ€ she says. โ€œTheyโ€™re not saying, โ€˜Letโ€™s abandon this and go straight to agentic.โ€™ Iโ€™m not seeing that at all.โ€

Ultimately, people are slow to change, and institutions are even slower. Many organizations are still running legacy systems. For example, the FAA has just come out with a bold plan to update its systems by getting rid of floppy disks and upgrading from Windows 95. They expect this to take four years.

But the center of gravity will move toward agents and, as it does, so will funding, innovation, green-field deployments, and the economics of the software industry.

โ€œThere are so many organizations and leaders who need to cross the chasm,โ€ says Sobera. โ€œYouโ€™re going to have organizations at different levels of maturity, and some will be stuck in SaaS mentality, but feeling more in control while some of our progressive clients will embrace the move. Weโ€™re also seeing those clients outperform their peers in revenue, innovation, and satisfaction.โ€

Before yesterdayMain stream

From oversight to intelligence: AIโ€™s impact on project management and business transformation

4 December 2025 at 05:00

For CIOs, the conversation around AI has moved from innovation to orchestration, and project management, long a domain of human coordination and control, is rapidly becoming the proving ground for how intelligent systems can reshape enterprise delivery and accelerate transformation.

In boardrooms across industries, CIOs face the same challenge of how to quantify AIโ€™s promise in operational terms: shorter delivery cycles, reduced overhead, and greater portfolio transparency. A 2025 Georgia Institute of Technology-sponsored study of 217 project management professionals and C-level tech leaders revealed that 73% of organizations have adopted AI in some form of project management.

Yet amid the excitement, the question of how AI will redefine the role of the project manager (PM) remains, as does how will the future framework for the business transformation program be defined.

A shift in the PMโ€™s role, not relevance

Across industries, project professionals are already seeing change. Early adopters in the study report project efficiency gains of up to 30%, but success depends less on tech and more on how leadership governs its use. The overwhelming majority found it highly effective in improving efficiency, predictive planning, and decision-making. But what does that mean for the associates running these projects?

Roughly one-third of respondents believed AI would allow PMs to focus more on strategic oversight, shifting from day-to-day coordination to guiding long-term outcomes. Another third predicted enhanced collaboration roles, where managers act as facilitators who interpret and integrate AI insights across teams. The rest envisioned PMs evolving into supervisors of AI systems themselves, ensuring that algorithms are ethical, accurate, and aligned with business goals.

These perspectives converge on a single point: AI will not replace PMs, but it will redefine their value. The PM of the next decade wonโ€™t simply manage tasks, theyโ€™ll manage intelligence and translate AI-driven insights into business outcomes.

Why PMOs canโ€™t wait

For project management offices (PMOs), the challenge is no longer whether to adopt AI but how. AI adoption is accelerating, with most large enterprises experimenting with predictive scheduling, automated risk reporting, and gen AI for documentation. But the integration is uneven.

Many PMOs still treat AI as an add-on, a set of tools rather than its strategic capability. This misses the point since AI is about augmenting judgment and automation. The organizations gaining a real competitive advantage are those embedding AI into their project methodologies, governance frameworks, and performance metrics with this five-point approach in mind.

1. Begin with pilot projects

Think small, scale fast. The most successful AI integrations begin with targeted use cases that automate project status reports, predict schedule slippage, or identify resource bottlenecks. These pilot projects create proof points, generate enthusiasm, and expose integration challenges early.

2. Measure value, not just activity

One common pitfall is adopting AI without clear performance metrics. PMOs should set tangible KPIs such as reduction in manual reporting time, improved accuracy in risk forecasts, shorter project cycle times, and higher stakeholder satisfaction. Communicating these outcomes across the organization is just as important as achieving them. Success stories build momentum, foster buy-in, and demystify AI for skeptical teams.

3. Upskill PMs

AI will only be as valuable as the people who use it. Nearly half of the surveyed professionals cited lack of a skilled workforce as a barrier to AI integration. Project managers donโ€™t need to become data scientists, but they must understand AI fundamentals, how algorithms work, where biases emerge, and what data quality means. In this evolving landscape, the most effective PMs will combine data literacy with human-centered leadership, including critical thinking, emotional intelligence, and communication.

4. Strengthen governance and ethics

Increasing AI raises pressing ethical questions, especially when algorithms influence project decisions. PMOs must take the lead in establishing AI governance frameworks that emphasize transparency, fairness, and human oversight. Embedding these principles into the PMOโ€™s charter doesnโ€™t just mitigate risk, it builds trust.

5. Evolve from PMO to BTO

The traditional PMO focuses on execution through scope, schedule, and cost. But AI-driven organizations are shifting toward business transformation offices (BTOs), which align projects directly with strategic value creation through process improvement in parallel. A PMO ensures projects are done right. A BTO ensures the right projects are done. A crucial element of this framework is the transition from a Waterfall to an Agile mindset. The evolution of project management has shifted from rigid plans to iterative, customer-centric, and collaborative methods, with hybrid methodologies becoming increasingly common. This Agile approach is vital for adapting to the rapid changes brought by AI and digital disruption.

The new PM career path

By 2030, AI could manage most routine project tasks, such as status updates, scheduling, and risk flagging, while human leaders focus on vision, collaboration, and ethics. This shift mirrors past revolutions in project management from the rise of Agile to digital transformation, but at an even faster pace. But as organizations adopt AI, the risk of losing the human element persists. Project management has always been about people and aligning interests, resolving conflicts, and inspiring teams. However, while AI can predict a delay, it canโ€™t motivate a team to overcome it. The PMโ€™s human ability to interpret nuance, build trust, and foster collaboration remains irreplaceable.

A call to action

AI represents the next frontier in enterprise project delivery, and the next decade will test how well PMOs, executives, and policymakers can navigate the evolution of transformation. To thrive, organizations must invest in people as much as in platforms, adopt ethical, transparent governance, foster continuous learning and experimentation, and measure success by outcomes rather than hype.

For CIOs, the mandate is clear: lead with vision, govern with integrity, and empower teams with intelligent tools. AI, after all, isnโ€™t a threat to the project management profession. Itโ€™s a catalyst for its reinvention, and when executed responsibly, AI-driven project management will not only deliver operational gains but also build more adaptive, human-centered organizations ready for the challenges ahead. By embracing it thoughtfully, PMs can elevate their roles from administrators to architects of change.

MS, ๋ฉ”ํƒ€๋ฒ„์Šค ์ „๋žต ์ถ•์†Œ๋กœ โ€˜๋ฉ”์‹œโ€™ ์•ฑ ์ง€์› ์ข…๋ฃŒยทยทยทํŒ€์ฆˆ ๋ชฐ์ž…ํ˜• ์ด๋ฒคํŠธ ๊ธฐ๋Šฅ ์ถœ์‹œ

4 December 2025 at 00:22

MS๋Š” ์ฝ”๋กœ๋‚˜19 ํŒฌ๋ฐ๋ฏน ๋‹น์‹œ, ์—…๋ฌด์— ๊ฐ€์ƒ ํ˜„์‹ค๊ณผ ๋ชฐ์ž…ํ˜• ํ™˜๊ฒฝ์„ ์ ์šฉํ•˜๋Š” ๋ฐ ๊ด€์‹ฌ์ด ๋†’์•„์ง€์ž ๋ฉ”์‹œ๋ฅผ ์ถœ์‹œํ–ˆ๋‹ค. ๋ฉ”์‹œ๋Š” ์œ ๋‹ˆํ‹ฐ ๊ธฐ๋ฐ˜์œผ๋กœ 3D ํ™˜๊ฒฝ์„ ์ œ์ž‘ํ•  ์ˆ˜ ์žˆ๋Š” ํ”Œ๋žซํผ์ด์ž, ๋™๋ฃŒ๊ฐ€ ๋ชฐ์ž…ํ˜• ๊ณต๊ฐ„์—์„œ ์•„๋ฐ”ํƒ€๋กœ ๋งŒ๋‚˜ ํ˜‘์—…ํ•  ์ˆ˜ ์žˆ๋Š” ์ „์šฉ ์•ฑ ํ˜•ํƒœ๋กœ ์ œ๊ณต๋ผ ์™”๋‹ค.

๊ทธ๋Ÿฌ๋‚˜ ์‹œ์žฅ ์ˆ˜์š”๋Š” ์ข€์ฒ˜๋Ÿผ ํ˜•์„ฑ๋˜์ง€ ์•Š์•˜๋‹ค. MS๋Š” ์ง€๋‚œ 1์ผ ๋ฉ”์‹œ ์•ฑ ์ œ๊ณต์„ ์ค‘๋‹จํ•œ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. ์ด์— ๋”ฐ๋ผ PC๋‚˜ ๋ฉ”ํƒ€์˜ ํ€˜์ŠคํŠธ ๊ฐ€์ƒํ˜„์‹ค ํ—ค๋“œ์…‹์—์„œ ๋ฉ”์‹œ ์•ฑ์„ ์ด์šฉํ•œ ์ด๋ฒคํŠธ ์ฐธ์„์ด๋‚˜ ํŒ€์ฆˆ ํšŒ์˜ ์ฐธ์—ฌ๊ฐ€ ๋” ์ด์ƒ ๋ถˆ๊ฐ€๋Šฅํ•ด์กŒ์œผ๋ฉฐ, ์›น์‚ฌ์ดํŠธ ์—ญ์‹œ ํ์‡„๋  ์˜ˆ์ •์ด๋‹ค.

์ด๋ฒˆ ์กฐ์น˜๋Š” MS๊ฐ€ ๋ฉ”ํƒ€๋ฒ„์Šค ๊ด€๋ จ ๋„๊ตฌ์—์„œ ์†์„ ๋–ผ๊ณ  ์žˆ์Œ์„ ๋ณด์—ฌ์ฃผ๋Š” ๋˜ ํ•˜๋‚˜์˜ ์‹ ํ˜ธ๋‹ค. ๋˜ํ•œ MS๋Š” ํ˜ผํ•ฉ ํ˜„์‹ค ํ—ค๋“œ์…‹ โ€˜ํ™€๋กœ๋ Œ์ฆˆโ€™์™€ ๊ด€๋ จํ•ด ๋ฏธ๊ตญ ์œก๊ตฐ๊ณผ ์•ฝ 22์–ต ๋‹ฌ๋Ÿฌ ๊ทœ๋ชจ์˜ ๊ณ„์•ฝ์„ ์ฒด๊ฒฐํ•˜๊ธฐ๋„ ํ–ˆ์ง€๋งŒ ์ง€๋‚œํ•ด ์ƒ์‚ฐ์„ ์ค‘๋‹จํ•œ ๋ฐ” ์žˆ๋‹ค.

MS๋Š” ์ด๋‚  ํŒ€์ฆˆ ํ˜‘์—… ์•ฑ์— ์ ์šฉํ•ด ์˜จ ๋ฉ”์‹œ ๊ธฐ๋ฐ˜ 3D ๋ชฐ์ž…ํ˜• ๊ฒฝํ—˜์—๋„ ๋ณ€๊ฒฝ ์‚ฌํ•ญ์„ ์ ์šฉํ•œ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. ํŒ€์ฆˆ์—์„œ ์†Œ๊ทœ๋ชจ ๊ทธ๋ฃน์ด ์•„๋ฐ”ํƒ€๋กœ ๋งŒ๋‚˜ ์ƒํ˜ธ์ž‘์šฉํ•  ์ˆ˜ ์žˆ์—ˆ๋˜ โ€˜๋ชฐ์ž…ํ˜• ๊ณต๊ฐ„(3D)โ€™ ๊ธฐ๋Šฅ์„ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์—์„œ ์ œ๊ฑฐํ•œ๋‹ค๋Š” ์„ค๋ช…์ด๋‹ค.

๋Œ€์‹  ํŒ€์ฆˆ ์‚ฌ์šฉ์ž๊ฐ€ ๊ต์œก ์„ธ์…˜, ๊ฐ€์ƒ ์ „์‹œํšŒ, ์ œํ’ˆ ์†Œ๊ฐœ ๋“ฑ ๋Œ€๊ทœ๋ชจ ๊ฐ€์ƒ ๋ชจ์ž„์„ ์ฃผ์ตœํ•˜๊ฑฐ๋‚˜ ์ฐธ์„ํ•  ์ˆ˜ ์žˆ๋Š” โ€˜ํŒ€์ฆˆ ๋ชฐ์ž…ํ˜• ์ด๋ฒคํŠธโ€™ ๊ธฐ๋Šฅ์ด ๊ณต๊ฐœ๋๋‹ค. ํ–‰์‚ฌ๋Š” ์•„์›ƒ๋ฃฉ ๋˜๋Š” ํŒ€์ฆˆ ์บ˜๋ฆฐ๋”์—์„œ ์˜ˆ์•ฝํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ๊ธฐ์—… ๋ธŒ๋žœ๋“œ์— ์ตœ์ ํ™”๋œ ํ™˜๊ฒฝ์ด๋‚˜ ์—…๋กœ๋“œํ•œ 3D ๋ชจ๋ธ๋กœ ์ž์œ ๋กญ๊ฒŒ ์‚ฌ์ „ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๋‹ค.

๋ชฐ์ž…ํ˜• ์ด๋ฒคํŠธ๋Š” PC, ๋งฅ, ๋ฉ”ํƒ€ ํ€˜์ŠคํŠธ ๊ธฐ๊ธฐ์—์„œ ์ด์šฉ ๊ฐ€๋Šฅํ•˜๋‹ค. ์ด ๊ธฐ๋Šฅ์„ ์ฃผ์ตœํ•˜๋ ค๋ฉด ํŒ€์ฆˆ ํ”„๋ฆฌ๋ฏธ์—„ ๋˜๋Š” ์ ํ•ฉํ•œ ํŒ€์ฆˆ ๋ผ์ด์„ ์Šค๊ฐ€ ํ•„์š”ํ•˜์ง€๋งŒ, ๊ณต๋™ ์ฃผ์ตœ์ž์™€ ์ฐธ์„์ž๋Š” ์ผ๋ฐ˜ ํŒ€์ฆˆ ๋ผ์ด์„ ์Šค๋งŒ์œผ๋กœ ์ฐธ์—ฌํ•  ์ˆ˜ ์žˆ๋‹ค.

๋ฉ”ํƒ€๋ฒ„์Šค ๊ฐœ๋…์ด ์‹œ์žฅ ์ „๋ฐ˜์—์„œ ํฐ ๋ฐ˜ํ–ฅ์„ ์ผ์œผํ‚ค์ง€๋Š” ๋ชปํ–ˆ์ง€๋งŒ, ์ผ๋ถ€ ๊ธฐ์—…์€ ์—ฌ์ „ํžˆ ๊ฐ€์ƒ ํ™˜๊ฒฝ์ด ์›๊ฒฉ ํ˜‘์—…์„ ๊ฐ•ํ™”ํ•  ๊ฐ€๋Šฅ์„ฑ์— ์ฃผ๋ชฉํ•˜๊ณ  ์žˆ๋‹ค.

๋ฉ”ํŠธ๋ฆฌ์ง€(Metrigy)์˜ ์‚ฌ์žฅ์ด์ž ์ˆ˜์„ ์• ๋„๋ฆฌ์ŠคํŠธ์ธ ์–ด์œˆ ๋ผ์ž๋Š”, ๋ณ„๋„ ์•ฑ์ด ์•„๋‹Œ ํŒ€์ฆˆ ๋‚ด๋ถ€์—์„œ ๊ธฐ์—…์ด ์ง์ ‘ ๊ฐ€์ƒ ์ด๋ฒคํŠธ๋ฅผ ๊ฐœ์ตœํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•œ MS์˜ ์ด๋ฒˆ ๋ณ€ํ™”๊ฐ€ ํ•ฉ๋ฆฌ์ ์ด๋ผ๊ณ  ํ‰๊ฐ€ํ–ˆ๋‹ค. ๋‹ค๋งŒ ์•ž์œผ๋กœ ๊ฐ€์ƒ ํ˜‘์—… ํ™˜๊ฒฝ์— ๋Œ€ํ•œ ์ˆ˜์š”๋Š” ์ œํ•œ์ ์ธ ๋ฒ”์œ„์— ๊ทธ์น  ๊ฒƒ์œผ๋กœ ์ „๋งํ–ˆ๋‹ค.

๋ฉ”ํŠธ๋ฆฌ์ง€ ์กฐ์‚ฌ์— ๋”ฐ๋ฅด๋ฉด ๊ฐ€์ƒํ˜„์‹ค๊ณผ ์ฆ๊ฐ•ํ˜„์‹ค ๋„์ž…์€ โ€˜๋А๋ฆฌ์ง€๋งŒ ๊พธ์ค€ํ•œโ€™ ์ฆ๊ฐ€์„ธ๋ฅผ ๋ณด์ด๊ณ  ์žˆ์œผ๋ฉฐ, 2024๋…„ ๋ง ์กฐ์‚ฌํ•œ ์•ฝ 400๊ฐœ ๊ธฐ์—… ๊ฐ€์šด๋ฐ 16.5%๊ฐ€ ์˜ฌํ•ด ๋ง๊นŒ์ง€ ํ•ด๋‹น ๊ธฐ์ˆ ์— ํˆฌ์žํ•  ๊ณ„ํš์ด๋ผ๊ณ  ๋‹ตํ–ˆ๋‹ค.

๋ผ์ž๋Š” โ€œ์‚ฌ์šฉ๋ก€๋Š” ์ผ๋ฐ˜์ ์ธ ํšŒ์˜๋ณด๋‹ค ๊ต์œก, ์ œํ’ˆ ์‹œ์—ฐ, ์—”์ง€๋‹ˆ์–ด๋ง๊ณผ ๋””์ž์ธ, ๊ณ ๊ฐ ์ฐธ์—ฌ์ฒ˜๋Ÿผ ๋ชฉ์ ์ด ๋šœ๋ ทํ•œ ์˜์—ญ์— ์ง‘์ค‘๋˜๋Š” ๊ฒฝํ–ฅ์ด ์žˆ๋‹ค. ์„ฑ์žฅ์€ ๋А๋ฆฌ์ง€๋งŒ ์ด์–ด์งˆ ๊ฒƒ์œผ๋กœ ์˜ˆ์ƒ๋˜์ง€๋งŒ, ์ด๋Ÿฌํ•œ ๊ฐ€์ƒํ˜„์‹ค ๋„๊ตฌ๊ฐ€ ์•ž์œผ๋กœ ์ฃผ๋ฅ˜ ์‹œ์žฅ์„ ๋„˜์–ด์„œ๋Š” ์ˆ˜์ค€์œผ๋กœ ํ™•๋Œ€๋  ๊ฒƒ์œผ๋กœ ๋ณด์ด์ง€๋Š” ์•Š๋Š”๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

AI ROI๊ฐ€ ๋ถ€์ง„ํ•œ ์ง„์งœ ์ด์œ , ๊ธฐ์ˆ ์ด ์•„๋‹ˆ๋ผ ๋ฆฌ๋”์‹ญ์ด๋‹ค

3 December 2025 at 20:06

AI์— ๋Œ€ํ•œ ๊ธฐ๋Œ€์™€ ๊ณผ์—ด๋œ ๊ด€์‹ฌ์ด ๊ทธ ์–ด๋–ค ๊ธฐ์ˆ  ๋ณ€ํ™”๋ณด๋‹ค ์ปค์กŒ๋‹ค๊ณ  ํ‰๊ฐ€๋œ๋‹ค. ์ด๋Ÿฐ ํ๋ฆ„์— ๋งž์ถฐ ํˆฌ์ž ๊ทœ๋ชจ๋„ ๋ฉ๋‹ฌ์•„ ์ปค์กŒ๋‹ค. ๊ฐ€ํŠธ๋„ˆ๋Š” 2025๋…„๊นŒ์ง€ ์ „ ์„ธ๊ณ„ AI ํˆฌ์ž ๊ทœ๋ชจ๊ฐ€ ์•ฝ 1์กฐ 5,000์–ต ๋‹ฌ๋Ÿฌ์— ์ด๋ฅผ ๊ฒƒ์œผ๋กœ ์ถ”์‚ฐํ–ˆ๋‹ค. ๊ฑฐ๋Œ€ํ•œ ํˆฌ์ž ๊ทœ๋ชจ์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ  ๋Œ€๋ถ€๋ถ„ ์กฐ์ง์€ ์•ฝ์†ํ–ˆ๋˜ ๊ฐ€์น˜์™€ ์‹ค์ œ๋กœ ์‹คํ˜„๋œ ๊ฐ€์น˜ ์‚ฌ์ด์˜ ๋งŒ์„ฑ์ ์ธ ๊ฒฉ์ฐจ์— ์—ฌ์ „ํžˆ ์‹œ๋‹ฌ๋ฆฌ๊ณ  ์žˆ๋‹ค.

๊ฐ€์žฅ ๋„๋ฆฌ ์ธ์šฉ๋˜๋Š” ๊ทผ๊ฑฐ๋Š” ์˜ฌํ•ด ์ดˆ MIT ๋ณด๊ณ ์„œ์—์„œ ๋‚˜์™”๋‹ค. ์ด ๋ณด๊ณ ์„œ๋Š” ์ƒ์„ฑํ˜• AI ํŒŒ์ผ๋Ÿฟ ํ”„๋กœ์ ํŠธ์˜ 95%๊ฐ€ ์‹คํŒจํ•œ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. ๋งฅํ‚จ์ง€ ์กฐ์‚ฌ์—์„œ๋„ ๊ฑฐ์˜ 80%์— ์ด๋ฅด๋Š” ๊ธฐ์—…์ด ์ƒ์„ฑํ˜• AI๋ฅผ ์‚ฌ์šฉํ•˜๊ณ  ์žˆ์ง€๋งŒ, ๊ฑฐ์˜ ๊ฐ™์€ ๋น„์œจ์˜ ๊ธฐ์—…์ด ์ˆ˜์ต์„ฑ์— ์˜๋ฏธ ์žˆ๋Š” ํšจ๊ณผ๋ฅผ ์–ป์ง€ ๋ชปํ–ˆ๋‹ค๊ณ  ์‘๋‹ตํ–ˆ๋‹ค.

AI๊ฐ€ ์–ด๋А ์ •๋„ ํšจ๊ณผ๋ฅผ ๋‚ด๊ณ  ์žˆ๋‹ค๋Š” ๊ทผ๊ฑฐ๋„ ์žˆ๋‹ค. 2025๋…„ ์‹œ์Šค์ฝ”์˜ โ€˜AI ์ค€๋น„๋„ ์ง€ํ‘œ(AI Readiness Index)โ€™์— ๋”ฐ๋ฅด๋ฉด, ์ „์ฒด ๊ธฐ์—…์˜ 13%๋Š” AI์—์„œ ์ง€์†์ ์œผ๋กœ ์ธก์ • ๊ฐ€๋Šฅํ•œ ์ˆ˜์ต์„ ์–ป๊ณ  ์žˆ๋‹ค๊ณ  ๋‹ตํ–ˆ๋‹ค. ๋น„๋ก ์†Œ์ˆ˜์ด์ง€๋งŒ ์„ ๋„ ๊ธฐ์—…์€ AI์—์„œ ๊ฐ€์น˜๋ฅผ ๋Œ์–ด๋‚ด๊ธฐ ์‹œ์ž‘ํ–ˆ๋‹ค๋Š” ์˜๋ฏธ๋‹ค. ์ด๋Ÿฐ ๊ฐ€์น˜์˜ ์ถœ์ฒ˜๋Š” ์ ์  ๊ธฐ์ˆ  ์ž์ฒด๊ฐ€ ์•„๋‹ˆ๋ผ ๋ฆฌ๋”์‹ญ์˜ ๋ช…ํ™•์„ฑ, ์ „๋žต์  ์ •๋ ฌ, ์‹คํ–‰ ์—ญ๋Ÿ‰์—์„œ ๋‚˜์˜ค๊ณ  ์žˆ๋‹ค. ์‹œ์Šค์ฝ”์˜ AI ์ค€๋น„๋„ ์ง€ํ‘œ๋Š” AI์—์„œ ๊ฐ€์น˜๋ฅผ ์‹คํ˜„ํ•œ ๊ธฐ์—…์˜ 99%๊ฐ€ ๋ณ€ํ™” ์ˆ˜์šฉ์„ ์ „์ œ๋กœ ํ•œ ๋ช…ํ™•ํ•œ ์ „๋žต์„ ๊ฐ–๊ณ  ์žˆ๊ณ , ์ž„์ง์›์ด ์ƒˆ๋กœ์šด ๊ธฐ์ˆ ์— ์ต์ˆ™ํ•ด์ง€๋„๋ก ์ง€์›ํ•˜๋Š” ๊ณต์‹ ํ”„๋กœ๊ทธ๋žจ์„ ์šด์˜ํ•˜๊ณ  ์žˆ๋‹ค๋Š” ์ ์„ ํ™•์ธํ–ˆ๋‹ค.

์˜ค๋Š˜๋‚  CEO์™€ CIO๋Š” ํ•œ ์„ธ๋Œ€๋ฅผ ๊ฐ€๋ฅด๋Š” ์ „ํ™˜์ ์— ์„œ ์žˆ๋‹ค. AI๋ฅผ ๋‹จ์ˆœํ•œ ๋น„์šฉ ์ ˆ๊ฐ ์ˆ˜๋‹จ์ด ์•„๋‹ˆ๋ผ ์—ญ๋Ÿ‰๊ณผ ์—ฌ๋ ฅ์„ ์ฐฝ์ถœํ•˜๊ณ  ํ˜์‹ ์„ ์ด‰์ง„ํ•˜๋ฉฐ ์ธ๊ฐ„ ์ค‘์‹ฌ์˜ ๊ฒฐ๊ณผ๋ฅผ ์ด๋„๋Š” ๋™๋ ฅ์œผ๋กœ ์žฌ์ •์˜ํ•ด์•ผ ํ•˜๋Š” ์‹œ๊ธฐ์ด๋‹ค. ์•ž์œผ๋กœ์˜ ๊ธธ์€ ์—…๋ฌด๋ฅผ ์ž˜๊ฒŒ ์ชผ๊ฐœ๊ณ , ์–ด๋””์— ์ž๋™ํ™”๋ฅผ ์ ์šฉํ•ด์•ผ ํ•˜๋Š”์ง€ ๋‹ค์‹œ ํ‰๊ฐ€ํ•˜๋ฉฐ, ์ธ์žฌ๊ฐ€ ์„ฑ์žฅ๊ณผ ๋ณ€ํ˜์— ์ง‘์ค‘ํ•  ์ˆ˜ ์žˆ๋„๋ก ๊ถŒํ•œ์„ ๋ถ€์—ฌํ•˜๋Š” ๋ฆฌ๋”์‹ญ์— ๋‹ฌ๋ ค ์žˆ๋‹ค.

์„ฑ๊ณต์„ ์œ„ํ•œ ์ค€๋น„ : AI์˜ ์•ฝ์†๊ณผ ํ˜„์‹ค

๋งŽ์€ CIO์—๊ฒŒ ๊ฐ€์žฅ ํฐ ๊ณผ์ œ๋Š” ์ตœ๊ณ  ๊ฒฝ์˜์ง„๊ณผ ์ด์‚ฌํšŒ๊ฐ€ โ€œAI๊ฐ€ ํ•„์š”ํ•˜๋‹คโ€๋Š” ์‚ฌ์‹ค์€ ์•Œ๊ณ  ์žˆ์ง€๋งŒ, ์ •์ž‘ ๋ฌด์—‡์„ ์œ„ํ•ด AI๊ฐ€ ํ•„์š”ํ•œ์ง€ ๋ช…ํ™•ํ•˜๊ฒŒ ์•Œ์ง€ ๋ชปํ•œ๋‹ค๋Š” ์ ์ด๋‹ค. ์ด ๋•Œ๋ฌธ์— AI๊ฐ€ ๋ชจ๋“  ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ณ  ์ƒ์‚ฐ์„ฑ์„ ๊ธ‰๋“ฑ์‹œ์ผœ ์ค„ ๋งŒ๋Šฅ ํ•ด๊ฒฐ์ฑ…์ด๋ผ๋Š” ๋น„ํ˜„์‹ค์ ์ธ ๊ธฐ๋Œ€๊ฐ€ ์ƒ๊ธด๋‹ค. ์ฒ˜์Œ๋ถ€ํ„ฐ ๋ชฉํ‘œ๊ฐ€ ๋น„ํ˜„์‹ค์ ์ด๋ฉด ํ”„๋กœ์ ํŠธ๋ฅผ ์„ฑ๊ณต์ ์œผ๋กœ ์™„๋ฃŒํ•˜๊ณ ๋„ ์‹คํŒจ๋กœ ๊ฐ„์ฃผํ•˜๋Š” ์ƒํ™ฉ์ด ๋ฐœ์ƒํ•œ๋‹ค. ์—ฌ๋Ÿฌ ์‚ฌ์—… ๋ถ€๋ฌธ์ด ์‚ฌ์—… ์„ฑ๊ณผ๋ฅผ ์ธก์ •ํ•  KPI๋ฅผ ๊ฐ–์ถ”์ง€ ๋ชปํ•œ ์ ๋„ ์ด๋Ÿฐ ๋ฌธ์ œ๋ฅผ ๋ถ€์ถ”๊ธด๋‹ค.

๋Œ€ํ‘œ์ ์ธ ์˜ˆ๊ฐ€ ์ปจํƒ์„ผํ„ฐ๋‹ค. ์ปจํƒ์„ผํ„ฐ์—์„œ๋Š” ์ƒ๋‹ด์‚ฌ ์ฝ”์นญ, ๊ฐ€์ƒ ์ƒ๋‹ด์‚ฌ, ๊ทผ๋ฌด ์Šค์ผ€์ค„๋ง, ํ†ตํ™” ์Šค์ฝ”์–ด๋ง, ์ฝœ ๋…ธํŠธ ์ž‘์„ฑ ๋“ฑ ๋‹ค์–‘ํ•œ ์˜์—ญ์— AI ์—์ด์ „ํŠธ๋ฅผ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋‹ค. ํ•˜์ง€๋งŒ ์ด๋Ÿฐ ํ™œ๋™์˜ ๊ฐ€์น˜๋ฅผ ์ธก์ •ํ•˜๊ธฐ๊ฐ€ ์–ด๋ ต๊ธฐ ๋•Œ๋ฌธ์— ๋งŽ์€ ๊ธฐ์—…์ด ๋‹จ์ˆœํžˆ ์ƒ๋‹ด์‚ฌ ์ˆ˜๋ฅผ ์ค„์—ฌ ๋น„์šฉ ์ ˆ๊ฐ์— ์ดˆ์ ์„ ๋งž์ถ”๋Š” ๋ฐฉ์‹์œผ๋กœ ๋Œ€์‘ํ•ด ์™”๋‹ค. ์ด๋Ÿฐ ์ ‘๊ทผ์€ ์‹ ์ค‘ํ•˜๊ฒŒ ์‹คํ–‰ํ•˜์ง€ ์•Š์œผ๋ฉด ์‰ฝ๊ฒŒ ์—ญํšจ๊ณผ๋ฅผ ๋‚ธ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, ๊ธˆ์œต ์„œ๋น„์Šค ํšŒ์‚ฌ ํด๋ผ๋ฅด๋‚˜(Klarna)๋Š” ์ƒ๋‹ด์‚ฌ ์•ฝ 700๋ช…์„ ๊ฐ์ถ•ํ•œ ๋’ค ๊ณ ๊ฐ ์„œ๋น„์Šค ์ ์ˆ˜๊ฐ€ ๊ธ‰๋ฝํ•˜์ž ์ธ๋ ฅ์„ ๋‹ค์‹œ ์ฑ„์šฉํ•ด์•ผ ํ–ˆ๋‹ค. ์ด ์‚ฌ๋ก€๋Š” ๊ธฐ์ˆ  ๋ฌธ์ œ๊ฐ€ ์•„๋‹ˆ๋ผ ๊ฒฝ์˜์ง„์ด ์˜ํ–ฅ๋„๋ฅผ ์ œ๋Œ€๋กœ ํŒŒ์•…ํ•˜๊ณ  ๊ณ„ํš์„ ์„ธ์šฐ์ง€ ๋ชปํ•œ ๋ฆฌ๋”์‹ญ ๋ฌธ์ œ์˜€๋‹ค.

๋ฌธ์ œ ์ง„๋‹จ : ๊ธฐ์ˆ ์  ํ•œ๊ณ„์ธ๊ฐ€, ๋ฆฌ๋”์‹ญ์˜ ๊ณต๋ฐฑ์ธ๊ฐ€

AI์™€ ์ž๋™ํ™” ๊ธฐ์ˆ ์—์„œ ROI๋ฅผ ์–ป์ง€ ๋ชปํ•˜๋Š” ๋ฌธ์ œ๋ฅผ ์ง„๋‹จํ•˜๋Š” ๊ณผ์ •์—์„œ ์‹ค์ œ ์›์ธ์„ ํ˜ผ๋™ํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค. ๋ณต์žกํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ํ†ตํ•ฉํ•˜๊ณ  ์ž๋™ํ™”๋ฅผ ํ™•์žฅํ•˜๋Š” ๊ธฐ์ˆ ์  ์ค€๋น„๊ฐ€ ๋ถ€์กฑํ•œ ๊ฒƒ์€ ์—ฌ์ „ํžˆ ํ•ด๊ฒฐํ•ด์•ผ ํ•  ๊ณผ์ œ์ด๋‹ค. ํ•˜์ง€๋งŒ ์‹ค์ œ๋กœ ๊ฐœ์„  ์กฐ์น˜๋ฅผ ์ทจํ•ด์•ผ ํ•  ์ง€์ ์€ ๊ธฐ์ˆ  ๊ทธ ์ž์ฒด๊ฐ€ ์•„๋‹ˆ๋ผ ๊ธฐ์ˆ  ํˆฌ์ž๋ฅผ ์–ด๋–ป๊ฒŒ ์ง€ํœ˜ํ•˜๊ณ  ์ด๋Œ์–ด ๊ฐ€๋Š”์ง€์— ๋” ๊ฐ€๊นŒ์šด ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค. ์ด๋Ÿฐ ์ƒํ™ฉ์€ AI๋ฅผ ์„ฑ๊ณต์ ์œผ๋กœ ๋„์ž…ํ•˜๋Š” ๋ฐ ๋ฐฉํ•ด๊ฐ€ ๋˜๋Š” ์žฅ์• ๋ฌผ์ด ์ฝ”๋“œ๋‚˜ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ ์„ค์ •์ด ์•„๋‹ˆ๋ผ ์ด์‚ฌํšŒ์™€ ๊ฒฝ์˜์ง„์— ๋” ๊ฐ€๊น๋‹ค๋Š” ์‚ฌ์‹ค์„ ์‹œ์‚ฌํ•œ๋‹ค. ๋งŽ์€ ์‹คํŒจ๊ฐ€ ์‹œ์žฅ์˜ ๊ณผ์—ด๊ณผ ์œ ํ–‰์— ๋ฐ€๋ ค ์‹ ์†ํ•จ์„ ์šฐ์„ ํ•˜๋Š” ๊ฒฝ์˜์ง„์˜ ์„ ํƒ์—์„œ ๋น„๋กฏ๋˜๊ณ , ์ง„์ •ํ•œ ๋น„์ฆˆ๋‹ˆ์Šค ๋ณ€ํ™”๋ฅผ ๊ณ ๋ฏผํ•˜์ง€ ์•Š์€ ์ฑ„ ์ถ”์ง„ํ•˜๊ธฐ ๋•Œ๋ฌธ์— ๋ฐœ์ƒํ•œ๋‹ค.

AI ํ”„๋กœ์ ํŠธ๋ฅผ ์‹œ์ž‘ํ•˜๊ธฐ ์ „์— ๋ฌด์—‡์„ ์ธก์ •ํ•˜๊ณ , ์–ด๋–ค ์ง€ํ‘œ๋กœ ํˆฌ์ž์ˆ˜์ต๋ฅ ์„ ๊ณ„์‚ฐํ• ์ง€์— ๋Œ€ํ•ด ๋น„์ฆˆ๋‹ˆ์Šค์™€ ๋ช…ํ™•ํžˆ ์ •๋ ฌํ•˜๋Š” ๊ณผ์ •๋„ ์ค‘์š”ํ•˜๋‹ค. ์ „๋žต์  ์šฐ์„ ์ˆœ์œ„์™€ ์—ฐ๊ณ„๋œ ๋ช…ํ™•ํ•œ ๋น„์ฆˆ๋‹ˆ์Šค ์ผ€์ด์Šค ์—†์ด ๊ณ ๋ฆฝ๋œ ๊ธฐ์ˆ  ์‹คํ—˜ ํ˜•ํƒœ๋กœ ์ถœ๋ฐœํ•œ ํ”„๋กœ์ ํŠธ๋Š” ๊ฐ€์น˜๊ฐ€ ๋ชจํ˜ธํ•ด ํšจ๊ณผ๋ฅผ ์ธก์ •ํ•˜๊ธฐ ์–ด๋ ต๊ณ , ์˜ˆ์‚ฐ ์กฐ์ • ์‹œ ์šฐ์„ ์ ์œผ๋กœ ์ž˜๋ ค ๋‚˜๊ฐ€๊ธฐ ์‰ฝ๋‹ค.

๋ฆฌ๋”์‹ญ์˜ ์ „ํ™˜์  : ๋น„์šฉ ์ ˆ๊ฐ์„ ๋„˜์–ด

๋น„์ฆˆ๋‹ˆ์Šค ์„ธ๊ณ„๋Š” ๋ฆฌ๋”์‹ญ ์ „ํ™˜์ ์— ์„œ ์žˆ๋‹ค. ์ธ๋ ฅ ํ˜์‹ ์ด ์ฒ˜์Œ์œผ๋กœ โ€˜์ธ๊ฐ„๋งŒ์˜ ๋ฌธ์ œโ€™๊ฐ€ ์•„๋‹Œ ์ง€์ ์ด๋‹ค. AI์™€ ์ž์œจ ์—์ด์ „ํŠธ๊ฐ€ ๋น ๋ฅด๊ฒŒ ํ™•์‚ฐ๋˜๋ฉด์„œ ๊ฒฝ์˜์ง„์€ ์กฐ์ง ์•ˆ์—์„œ ๊ฐ€์น˜๋ฅผ ์–ด๋–ป๊ฒŒ ๋งŒ๋“ค์–ด๋‚ด๊ณ  ์œ ์ง€ํ• ์ง€์— ๋Œ€ํ•ด ํ›จ์”ฌ ๋” ๋ณต์žกํ•œ ๊ฒฐ์ •์„ ๋‚ด๋ ค์•ผ ํ•œ๋‹ค. ์ด ๋ณ€ํ™”๋Š” ๋น„์šฉ์„ ์ค„์ด๋Š” ์ฐจ์›์„ ๋„˜์–ด, ์ธ๊ฐ„์˜ ์—ญ๋Ÿ‰๊ณผ AI ๊ธฐ์ˆ  ์—ญ๋Ÿ‰ ์‚ฌ์ด์˜ ๊ด€๊ณ„๋ฅผ ๋‹ค์‹œ ์„ค๊ณ„ํ•˜๊ณ , ์กฐ์ง ๋ฌธํ™”์™€ ํ”„๋กœ์„ธ์Šค๊ฐ€ ์ด๋Ÿฐ ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ์ธ๋ ฅ ํ™˜๊ฒฝ์— ์ ์‘ํ•  ์ˆ˜ ์žˆ๋„๋ก ๋งŒ๋“œ๋Š” ์ž‘์—…์„ ์š”๊ตฌํ•œ๋‹ค.

๊ธฐ์—… ์ธํ…”๋ฆฌ์ „์Šค ์ „๋ฌธ ์—…์ฒด ์•Œํฌํ…์…œ(r.Potential)์˜ CEO ๊ทธ๋ ‰ ์Šˆ๋ฉ”์ด์ปค๋Š” โ€œ๋งŽ์€ ๊ธฐ์—…์ด ๋” ๋งŽ์€ AI ๋ชจ๋ธ๊ณผ ์ปดํ“จํŠธ, ์—์ด์ „ํŠธ๋ฅผ ๊ตฌ์ถ•ํ•˜๋Š” ๋ฐ ๊ฒฝ์Ÿํ•˜๊ณ  ์žˆ์ง€๋งŒ, ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ AI ๋„์ž…์˜ ์‹ค์ œ ๋ณ‘๋ชฉ์€ ๊ณต๊ธ‰ ์ธก์ด ์•„๋‹ˆ๋ผ ๊ธฐ์—…์ด ์–ด๋””์— ์–ด๋–ป๊ฒŒ AI๋ฅผ ํ™œ์šฉํ•ด ์‹ค์ œ ์ผ์„ ์ฒ˜๋ฆฌํ• ์ง€ ๋ชจ๋ฅธ๋‹ค๋Š” ์ ์— ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ์Šˆ๋ฉ”์ด์ปค๋Š” โ€œ๋ˆ„๊ฐ€ ์–ด๋–ค ์ผ์„ ํ•ด์•ผ ํ•˜๋Š”์ง€์— ๋Œ€ํ•œ ์‹ค์ œ ์ˆ˜์š”๋ฅผ ํฌ์ฐฉํ•˜๊ณ , ์‚ฌ๋žŒ๊ณผ AI ์—ญ๋Ÿ‰์„ ํ˜„์‹ค์ ์œผ๋กœ ์กฐํ•ฉํ•œ ๊ตฌ์„ฑ์•ˆ์„ ๋งŒ๋“ค๋ฉฐ, ์ด๋ฅผ ์‹ค์ œ ๋น„์ฆˆ๋‹ˆ์Šค ์„ฑ๊ณผ์™€ ์—ฐ๊ณ„ํ•˜๋Š” ์ธ๊ฐ„-๋””์ง€ํ„ธ ์—…๋ฌด ์กฐ์ • ๊ณ„์ธต์ด ๋น ์ ธ ์žˆ๋‹ค๊ณ  ๋ณธ๋‹ค. ์ด ๋ฌธ์ œ๋ฅผ ํ’€์ง€ ๋ชปํ•˜๋ฉด ๋‹ค์Œ ์ž๋™ํ™” ๋ฌผ๊ฒฐ์€ ๊ธฐ์—…์„ ์žฌํŽธํ•˜๋Š” ์ˆ˜์ค€์„ ๋„˜์–ด, ์ผ ์ž์ฒด๋ฅผ ๋ถˆ์•ˆ์ •ํ•˜๊ฒŒ ๋งŒ๋“ค ๊ฒƒโ€์ด๋ผ๊ณ  ๊ฒฝ๊ณ ํ–ˆ๋‹ค.

์Šˆ๋ฉ”์ด์ปค์˜ ์ง€์ ์€ ๋งŽ์€ ๊ฒฝ์˜์ง„์ด ๊ฐ„๊ณผํ•˜๋Š” ์ง€์ ์„ ์ž˜ ๋ณด์—ฌ์ค€๋‹ค. AI์˜ ์ž ์žฌ๋ ฅ์€ ๋‹จ์ˆœํ•œ ๊ธฐ์ˆ  ํ˜น์€ ์šด์˜ ๊ณผ์ œ๊ฐ€ ์•„๋‹ˆ๋ผ ๋ฆฌ๋”์‹ญ์˜ ๋ณธ์งˆ์„ ๋ฌป๋Š” ๋ฌธ์ œ๋ผ๋Š” ์ ์ด๋‹ค. ์ธ๊ฐ„์ด ํ•ด์•ผ ํ•  ์—…๋ฌด์™€ ๊ธฐ๊ณ„๋กœ ์ž๋™ํ™”ํ•  ์—…๋ฌด ์‚ฌ์ด ๊ฒฝ๊ณ„๊ฐ€ ํ๋ ค์ง€๋Š” ์ƒํ™ฉ์—์„œ IT์™€ ๋น„์ฆˆ๋‹ˆ์Šค ๋ฆฌ๋”๋Š” ์‚ฌ๋žŒ, ๊ธฐ์ˆ , ๋ฌธํ™”์˜ ๊นŠ์€ ํ†ตํ•ฉ์„ ํ†ตํ•ด ๊ฐ€์น˜ ์ฐฝ์ถœ์„ ๊ทน๋Œ€ํ™”ํ•˜๋Š” ๋ฐ ์ง‘์ค‘ํ•ด์•ผ ํ•œ๋‹ค.

๊ทธ๋ ‡๊ธฐ ๋•Œ๋ฌธ์— ์ตœ๊ณ  ๊ฒฝ์˜์ง„์€ ํˆฌ์ž ํƒ€์ž„๋ผ์ธ์„ ๋‹ค์‹œ ๊ณ ๋ฏผํ•˜๊ณ , ๊ธฐ์ˆ  ๋ฐ ์‹œ์žฅ ๋ณ€ํ™”์— ๋งž์ถฐ ์กฐ์ง์˜ ์—ญ๋Ÿ‰์„ ํ™•์žฅํ•˜๋Š” ๋ฐฉ์‹์„ ์žฌ์„ค๊ณ„ํ•ด์•ผ ํ•œ๋‹ค. ์ด๋Ÿฐ ์ธ์  ์ž์› ๊ด€๋ฆฌ์˜ ๋ณ€ํ™”๋Š” ๋ฏธ๋ž˜ ์ธ๋ ฅ ๊ตฌ์กฐ๋ฅผ ์˜ˆ์ธกํ•˜๊ณ , ์‚ฌ๋žŒ๊ณผ ๊ธฐ๊ณ„ ๊ธฐ๋ฐ˜ ์ง€๋Šฅ์„ ์—ฐ๊ณ„ํ•ด ์ธ๋ ฅ์„ ๋ฐฐ์น˜ํ•˜๋Š” ๋Šฅ๋ ฅ์„ ํ•„์š”๋กœ ํ•œ๋‹ค. ํ˜์‹ ์€ ํ˜„์žฌ ์„ฑ๊ณผ๋ฅผ ์กฐ๊ธˆ ๋ณดํƒœ๋Š” ์ฐจ์›์ด ์•„๋‹ˆ๋ผ, ์กฐ์ง์ด ๊ด€๋ จ ์ธํ”„๋ผ์™€ ์ฒด๊ณ„๋ฅผ ๊ฐ–์ถ”๊ณ  ์Šค์Šค๋กœ๋ฅผ ์žฌ์ฐฝ์กฐํ•  ์ค€๋น„๊ฐ€ ๋˜์–ด ์žˆ๋Š”์ง€ ์—ฌ๋ถ€๋ฅผ ์ค‘์‹ฌ์— ๋‘ฌ์•ผ ํ•œ๋‹ค.

์—…๋ฌด์™€ ๊ฐ€์น˜ ์ฐฝ์ถœ์˜ ๋ถ„ํ•ด

AI์—์„œ ์ดํ•ด ๊ฐ€๋Šฅํ•œ ROI๋ฅผ ๋„์ถœํ•˜๋ ค๋ฉด ์—…๋ฌด๋ฅผ ๊ตฌ์„ฑํ•˜๋Š” ํ•ต์‹ฌ ์š”์†Œ๋ฅผ ์ •ํ™•ํžˆ ์ดํ•ดํ•ด์•ผ ํ•œ๋‹ค. ์–ด๋–ค ์ˆ˜์ต์ด๋“  ๊ธฐ์ˆ  ๋„์ž…๋ฟ ์•„๋‹ˆ๋ผ ํ”„๋กœ์„ธ์Šค ๋ณ€ํ™”๋ฅผ ํ•จ๊ป˜ ๊ณ ๋ คํ•ด์•ผ ํ•œ๋‹ค. ๋ชฉํ‘œ๋Š” ๋ฐ˜๋ณต ์ž‘์—…์—๋งŒ ์“ฐ์ด๋Š” ์˜์—ญ์ด ์•„๋‹ˆ๋ผ ์ธ๊ฐ„์˜ ํŒ๋‹จ, ๊ณต๊ฐ, ์ฐฝ์˜์„ฑ์ด ํ•„์š”ํ•œ ์˜์—ญ์—์„œ ์ธ๊ฐ„์˜ ๋…ธ๋ ฅ์„ ์ฆํญ์‹œํ‚ค๋Š” ๋ฐฉ์‹์œผ๋กœ AI๋ฅผ ํ™œ์šฉํ•˜๊ณ , ์‚ฌ๋žŒ์„ ๋ณด๋‹ค ๋†’์€ ๊ฐ€์น˜์˜ ๊ฐ๋… ๋ฐ ๋Œ€์™ธ ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜ ์—ญํ• ์— ๋ฐฐ์น˜ํ•ด ์ƒ์‚ฐ์„ฑ๊ณผ ๊ธฐ์—ฌ๋„๋ฅผ ๊ทน๋Œ€ํ™”ํ•˜๋Š” ๋ฐ ์žˆ๋‹ค.

์„ฑ๊ณต์˜ ํ•ต์‹ฌ์€ AI๋ฅผ ํ†ตํ•ด ์ธ์žฌ๊ฐ€ ๋งค์ถœ ์„ฑ์žฅ๊ณผ ํ˜์‹ ์„ ์ฃผ๋„ํ•  ์ˆ˜ ์žˆ๋„๋ก ๋•๋Š” ๋ฐฉํ–ฅ์œผ๋กœ ์ดˆ์ ์„ ์žฌ์กฐ์ •ํ•˜๋Š” ๊ฒƒ์ด๋‹ค. ๋ชฉํ‘œ๋Š” ์–ด๋–ค ์ง๋ฌด์—์„œ๋“  80%์— ๋‹ฌํ•˜๋Š” ์ž๋™ํ™” ๊ฐ€๋Šฅ ์—…๋ฌด๋ฅผ AI์— ๋งก๊ธฐ๊ณ , ๊ณ ์ˆ™๋ จ ์ธ๋ ฅ์ด ๋‚˜๋จธ์ง€ 20%์— ์ง‘์ค‘ํ•˜๋„๋ก ๋งŒ๋“œ๋Š” ๊ฒƒ์ด๋‹ค. ์ด ๋งˆ์ง€๋ง‰ 20%๊ฐ€ ์ƒˆ๋กœ์šด ๋งค์ถœ ์„ฑ์žฅ, ๊ณ ๊ฐ ์ถฉ์„ฑ๋„, ํ˜์‹ ์  ๋ŒํŒŒ๊ตฌ๋ฅผ ์ด๋„๋Š” ์˜์—ญ์ด๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค.

AI ์‹œ๋Œ€์—๋Š” ๋ชจ๋“  ์‚ฐ์—…์ด ์žฌํŽธ๋  ๊ฒƒ์ด๋ฉฐ, CIO์™€ CEO๊ฐ€ ํ•จ๊ป˜ ์ง„ํ™”ํ•˜์ง€ ์•Š์œผ๋ฉด AI ํˆฌ์ž๋Š” ํ—ˆ๊ณต์— ์‚ฌ๋ผ์ง€๋Š” ๋น„์šฉ์ด ๋œ๋‹ค. AI์—์„œ ์‹ค์งˆ์  ๊ฐ€์น˜๋ฅผ ์–ป๊ธฐ ์œ„ํ•œ ์—ด์‡ ๋Š” ๋ฆฌ๋”์‹ญ์ด ๋ฏธ๋ž˜ ์ง€ํ–ฅ์  ์‚ฌ๊ณ ๋ฅผ ๊ฐ–์ถ”๊ณ , ์ƒˆ๋กœ์šด ๊ธฐ์ˆ ๊ณผ ๋ณ€ํ™”๋ฅผ ๊ธฐ๊บผ์ด ์ˆ˜์šฉํ•˜๋А๋ƒ์— ๋‹ฌ๋ ค ์žˆ๋‹ค. ์ค‘์š”ํ•œ ๊ฒƒ์€ ๊ฐ€์žฅ ๋›ฐ์–ด๋‚œ ์ฝ”๋”๊ฐ€ ๋˜๋Š” ๊ฒƒ์ด ์•„๋‹ˆ๋ผ, ์˜ฌ๋ฐ”๋ฅธ ๋ฆฌ๋”์‹ญ ๊ตฌ์กฐ๋ฅผ ์„ค๊ณ„ํ•˜๋Š” ์ผ์ด๋‹ค. AI๋ฅผ ์‚ฌ๋žŒ ์ค‘์‹ฌ์˜ ๋ชฉํ‘œ๋ฅผ ํ™•์žฅํ•˜๋Š” ๋„๊ตฌ๋กœ ํ™œ์šฉํ•˜๊ณ , ๋‹จ์ˆœํžˆ ์šด์˜ ์ง€ํ‘œ์™€ ์Šคํ”„๋ ˆ๋“œ์‹œํŠธ๋ฅผ AI ๋ฐ์ดํ„ฐ๋กœ ์ฑ„์šฐ๋Š” ์ˆ˜๋‹จ์œผ๋กœ ์“ฐ์ง€ ์•Š๋Š” ๋ฆฌ๋”์‹ญ์ด ํ•„์š”ํ•˜๋‹ค. ์ด๋ฅผ ์œ„ํ•ด์„œ๋Š” ๋ชจ๋“  ์˜์‚ฌ๊ฒฐ์ • ๊ณผ์ •์— ์—„๊ฒฉํ•œ ์œค๋ฆฌ์™€ ๊ฑฐ๋ฒ„๋„Œ์Šค๋ฅผ ์‹ฌ๊ณ , ๊ฐ AI ํ”„๋กœ์ ํŠธ๊ฐ€ ์กฐ์ง์— ์–ด๋–ค ๋ชฉ์ ์„ ์ˆ˜ํ–‰ํ•˜๋Š”์ง€์— ๋Œ€ํ•ด ๋น„์ฆˆ๋‹ˆ์Šค ์ธก๋ฉด์—์„œ ํ™•์‹คํžˆ ํ•ฉ์˜ํ•ด ๋‘์–ด์•ผ ํ•œ๋‹ค.
dl-ciokorea@foundryco.com

AIโ€™s lack of ROI is down to leadership, not tech

3 December 2025 at 05:00

The increasing hype around AI has exceeded any other technology shift, perhaps ever. This has been met with a corresponding amount of investment. Gartner estimates worldwide spending on AI through 2025 will be nearly $1.5 trillion. Despite the staggering amount, most organizations continue to grapple with a chronic gap between promise and realized value.

The most widely recognized data point to support this comes from an MIT report from earlier this year that reveals 95% of gen AI pilots fail. A McKinsey study also found that nearly 80% of companies use gen AI, yet almost as many report no significant impact to the bottom line.

But thereโ€™s proof AI is working, if modestly. The 2025 Cisco AI Readiness Index shows 13% of all companies get consistently measurable returns from AI. So while this is a minority, leading organizations are starting to see value. But the origins of that value increasingly stem from leadership clarity, strategic alignment, and execution, not the technology itself. The Cisco AI Readiness Index measured this and found 99% of companies that have realized value from AI have a well-defined strategy that embraces change, and includes formal programs to help employees get comfortable with the new technology.

Todayโ€™s CEOs and CIOs face a generational inflection point. They must redefine success for AI not as a means of cost-cutting, but as a driver for capacity creation, innovation, and human-centric outcomes. The path forward requires breaking down work, reassessing where automation helps, and empowering talent to focus on growth and transformation.

Setting the stage: AI promise vs. reality

Among many CIOs, the biggest challenge is the C-Suite and board know they need AI but arenโ€™t sure what for. This creates unrealistic expectations that AI will be a panacea to all problems and send productivity skyrocketing. When the outcomes are unrealistic, a successful project may be deemed unsuccessful because the initial goals were incorrect. A contributing factor to this problem is that many business units donโ€™t have the KPIs to create the business metrics to measure.

An example of this is with contact centers where AI agents could be used for agent coaching, virtual agents, scheduling, scoring calls, note taking and more. Measuring the value of these can be difficult, so many businesses have defaulted to cutting costs by reducing agents. This can backfire if not done in a measured way. Klarna, for instance, eliminated about 700 agents, saw customer service scores tank, and then hired people back. This wasnโ€™t a technology problem but rather leadership didnโ€™t put the right plan in place to understand the impact.

Diagnosing the problem: Tech limitations or leadership gaps?

Issues with achieving ROI on AI and automation technology efforts are often confused with the diagnosis of the issue at hand. A lack of tech readiness concerning integrating complex data and scaling automation remains a challenge. More often than not, however, the issue that gets acted upon may be less with the technology itself and more with the way technology efforts are governed and led. This would indicate the obstacles that exist to technologyโ€™s successful implementation lie more with the board than the code and cloud setup. Failures happen because business leaders prioritize expediency driven by market hype instead of thinking about genuine business transformation.

Itโ€™s also important that prior to AI projects being kicked off, thereโ€™s clear business alignment and an understanding of the metrics being measured to calculate ROI. Projects launched as isolated technology experiments without a well-defined business case tied to a strategic priority are hard to measure and often vague in value, leading to projects that are easy to cut.

The leadership inflection point: Beyond cost cutting

The business world sits at a leadership inflection point where for the first time, workforce transformation will be more than just human. With the rapid adoption of AI and autonomous agents, leaders now face complex decisions about how value is derived and maintained within their organizations. This shift requires cutting costs but also reimagining the relationship between human skills and the technical capabilities of AI, which ensures organizational cultures and processes can adapt to this new era of hybrid workforces.

โ€œEveryoneโ€™s been racing to build more AI models, compute, and agents, but the real bottleneck to enterprise AI adoption isnโ€™t supply, itโ€™s that enterprises donโ€™t know where or how to use AI to do real work,โ€ says Greg Shewmaker, CEO of enterprise intelligence company r.Potential. โ€œWe believe the missing piece is the coordination layer between human and digital work, where you can capture actual workforce demand, generate realistic and deployable configurations of human and AI capabilities, and tie them to real business outcomes. If we donโ€™t get this right, the next wave of automation wonโ€™t just reshape companies, itโ€™ll destabilize work itself.โ€ย 

His point underscores what many executives miss: AIโ€™s promise isnโ€™t just a technological or operational challenge, itโ€™s an existential leadership one. As the boundaries blur between tasks suited for humans and those automated by machines, IT and business leaders will need to focus on maximizing value creation through deep integration of people, technology, and culture.โ€‹

So it becomes critical for the C-suite to reconsider timelines related to investments and expand capacity in accordance with tech and market needs. This shift in human capital management involves being able to forecast the future workforce and deploy human resources in sync with machine-based intelligence. Innovation should take precedence thatโ€™s less about adding to current performance and more about ensuring organizations can remain agile and ready to facilitate innovation in terms of related infrastructure and preparedness to reinvent themselves.

Breaking down work and value creation

Understanding the key components of work is essential to developing understandable ROI from AI. Any returns must consider the adoption of technology and the transformation of processes. The goal should be to leverage AI to amplify human effort in areas that require human judgment, empathy, and creativity rather than in areas where thereโ€™s only repetition of tasks, thereby assigning human resources to higher-value supervisory and human roles where they can be most productive and valuable.

The key to success is to refocus on enabling talent to drive revenue growth and innovation through AI. So the goal is to apply AI strategically to liberate highly-skilled people from working on the 80% of any job that can, should, and must be automated so they can focus on the last 20%, which drives new revenue growth, customer loyalty, and innovation breakthroughs.

The AI era will reshape every industry, and if CIOs and CEOs arenโ€™t evolving, the AI investment will be wasted. The key to realizing real value in AI is to ensure leadership is future-ready and embraces new skills and change. Itโ€™s not about being the best coder in the room but instilling the right leadership structure. This involves a leadership mentality that uses AI to further human-centric goals and not simply fill an operational spreadsheet with AI data. This requires strict ethics and governance modeled in every aspect of decision-making, and firm alignment on the business side where every AI project has a defined purpose for the organization.

Agentic AIโ€™s rise is making the enterprise architect role more fluid

3 December 2025 at 05:00

In a previous feature about enterprise architects, gen AI had emerged, but its impact on enterprise technology hadnโ€™t been felt. Today, gen AI has spawned a plethora of agentic AI solutions from the major SaaS providers, and enterprise architecture and the role of enterprise architect is being redrawn. So what do CIOs and their architects need to know?

Organizations, especially their CEOs, have been vocal of the need for AI to improve productivity and bring back growth, and analysts have backed the trend. Gartner, for example, forecasts that 75% of IT work will be completed by human employees using AI over the next five years, which will demand, it says, a proactive approach to identifying new value-creating IT work, like expanding into new markets, creating additional products and services, or adding features that boost margins.

If this radical change in productivity takes place, organizations will need a new plan for business processes and the tech that operates those processes. Recent history shows if organizations donโ€™t adopt new operating models, the benefits of tech investments canโ€™t be achieved.

As a result of agentic AI, processes will change, as well as the software used by the enterprise, and the development and implementation of the technology. Enterprise architects, therefore, are at the forefront of planning and changing the way software is developed, customized, and implemented.

In some quarters of the tech industry, gen AI is seen as a radical change to enterprise software, and to its large, well-known vendors. โ€œTo say AI unleashed will destroy the software industry is absurd, as it would require an AI perfection that even the most optimistic couldnโ€™t agree to,โ€ says Diego Lo Giudice, principal analyst at Forrester. Speaking at the One Conference in the fall, Lo Giudice reminded 4,000 business technology leaders that change is taking place, but itโ€™s built on the foundations of recent successes.

โ€œAgile has given better alignment, and DevOps has torn down the wall between developers and operations,โ€ he said. โ€œTheyโ€™re all trying to do the same thing, reduce the gap between an idea and implementation.โ€ Heโ€™s not denying AI will change the development of enterprise software, but like Agile and DevOps, AI will improve the lifecycle of software development and, therefore, the enterprise architecture. The difference is the speed of change. โ€œIn the history of development, thereโ€™s never been anything like this,โ€ adds Phil Whittaker, AI staff engineer at content management software provider Umbraco.

Complexity and process change

As the software development and customization cycle changes, and agentic applications become commonplace, enterprise architects will need to plan for increased complexity and new business processes. Existing business processes canโ€™t continue if agentic AI is taking on tasks currently done manually by staff.

Again, Lo Giudice adds some levity to a debate that can often become heated, especially in the wake of major redundancies by AI leaders such as AWS. โ€œThe view that everyone will get a bot that helps them do their job is naรฏve,โ€ he said at the One Conference. โ€œOrganizations will need to carry out a thorough analysis of roles and business processes to ensure they spend money and resources on deploying the right agents to the right tasks. Failure to do so will lead to agentic technology being deployed thatโ€™s not needed, canโ€™t cope with complex tasks, and increases the cloud costs of the business.

โ€œItโ€™s easy to build an agent that has access to really important information,โ€ says Tiago Azevedo, CIO for AI-powered low-code platform provider OutSystems. โ€œYou need segregation of data. When you publish an agent, you need to be able to control it, and thereโ€™ll be many agents, so costs will grow.โ€

The big difference, though, is deterministic and non-deterministic, says Whittaker. So non-deterministic requires guardrails of deterministic agents that produce the same output every time over the more random outcomes of non-deterministic agents. Defining business outcomes by deterministic and non-deterministic is a clear role for enterprise architecture. He adds that this is where AI can help organizations fill in gaps. Whittaker, whoโ€™s been an enterprise architect, says itโ€™ll be vital for organizations to experiment with AI to see how it can benefit their architecture and, ultimately, business outcomes.

โ€œThe path to greatness lies not in chasing hype or dismissing AIโ€™s potential, but in finding the golden middle ground where value is truly captured,โ€ write Gartner analysts Daryl Plummer and Alicia Mullery. โ€œAIโ€™s promise is undeniable, but realizing its full value is far from guaranteed. Our research reveals the sobering odds that only one in five AI initiatives achieve ROI, and just one in 50 deliver true transformation.โ€ Further research also finds just 32% of employees trust the organizationโ€™s leadership to drive transformation. โ€œAgents bring an additional component of complexity to architecture that makes the role so relevant,โ€ Azevedo adds.

In the past, enterprise architects were focused on frameworks. Whittaker points out that new technology models will need to be understood and deployed by architects to manage an enterprise that comprises employees, applications, databases, and agentic AI. He cites MCP as one as it provides a standard way to connect AI models to data sources, and simplifies the current tangle of bespoke integrations and RAG implementations. AI will also help architects with this new complexity. โ€œThere are tools for planning, requirements, creating epics, user stories, code generation, documenting code, and translating it,โ€ added Lo Giudice.

New responsibilities

Agentic AI is now a core feature of every major EA tool, says Stรฉphane Vanrechem, senior analyst at Forrester. โ€œThese agents automate data validation, capability mapping, and artifact creation, freeing architects to focus on strategy and transformation.โ€ He cites the technology of Celonis, SAP Signavio, and ServiceNow for their agentic integrations. Whittaker adds that the enterprise architect has become an important human in the loop to protect the organization and be responsible for the decisions and outcomes that agentic AI delivers.

Although some enterprise architects will see this as a collapse of their specialization, Whittaker thinks it broadens the scope of the role and makes them more T-shaped. โ€œI can go deep in different areas,โ€ he says. โ€œPigeon-holing people is never a great thing to do.โ€

Traditionally, architecture has suggested that something is planned, built, and then exists. The rise of agentic AI in the enterprise means the role of the enterprise architect is becoming more fluid as they continue to design and oversee construction. But the role will also involve continual monitoring and adjustment to the plan. Some call this orchestration, or perhaps itโ€™s akin to map reading. An enterprise architect may plan a route, but other factors will alter the course. And just like weather or a fallen tree, which can lead to a route deviation, so too will enterprise architects plan and then lead when business conditions change.

Again, this new way of being an enterprise architect will be impacted by technology. Lo Guidice believes thereโ€™ll be increased automation, and Azevedo sides with the orchestration view, saying agents are built and a catalogue of them is created across the organization, which is an opportunity for enterprise architects and CIOs to be orchestrators.

Whatever the job title, Whittaker says enterprise architecture is more important than ever. โ€œMore people will become enterprise architects as more software is written by AI,โ€ he says. โ€œThen itโ€™s an architectural role to coordinate and conduct the agents in front of you.โ€ He argues that as technologists allow agents and AI to do the development work for them, the responsibility of architecting how agents and processes function broadens and becomes the responsibility of many more technologists.

โ€œAI can create code for you, but itโ€™s your responsibility to make sure itโ€™s secure,โ€ he adds. Rather than developing the code, technology teams will become architecture teams, checking and accepting the technology that AI has developed, and then managing its deployment into the business processes.

With shadow AI already embedded in organizations, Whittakerโ€™s view shows the need for a team of enterprise architects that can help business align with the AI agents theyโ€™ve deployed, and at the same time protect customer data and cybersecurity posture.

AI agents are redrawing the enterprise, and at the same time replanning the role of enterprise architects.

ํด๋”๋ธ” ์‹œ์žฅ ํ™•์žฅ ๋‚˜์„ ๋‹คยทยทยท์‚ผ์„ฑ์ „์ž, ๋‘ ๋ฒˆ ์ ‘๋Š” โ€˜๊ฐค๋Ÿญ์‹œ Z ํŠธ๋ผ์ดํด๋“œโ€™ ์ถœ์‹œ

2 December 2025 at 02:00

์ด๋ฒˆ ์‹ ์ œํ’ˆ์€ ์‚ผ์„ฑ์ „์ž๊ฐ€ ์ฒซ ํด๋”๋ธ” ์Šค๋งˆํŠธํฐ ์ถœ์‹œ ์ดํ›„ ์ถ•์ ํ•ด ์˜จ ํ•˜๋“œ์›จ์–ด ์„ค๊ณ„์™€ ์ œ์กฐ ๊ธฐ์ˆ ์„ ๋ฐ”ํƒ•์œผ๋กœ ์™„์„ฑํ•œ ์ฐจ์„ธ๋Œ€ ์ œํ’ˆ์œผ๋กœ, ํ™”๋ฉด์„ ๋‘ ๋ฒˆ์— ๊ฑธ์ณ ์ ‘๊ณ  ํŽผ์น˜๋Š” ๊ตฌ์กฐ๋ฅผ ํ†ตํ•ด ์Šค๋งˆํŠธํฐ๊ณผ ํƒœ๋ธ”๋ฆฟ ์‚ฌ์ด์˜ ๊ฒฝ๊ณ„๋ฅผ ํ•œ์ธต ๋” ์œ ์—ฐํ•˜๊ฒŒ ํ–ˆ๋‹ค.

โ€˜๊ฐค๋Ÿญ์‹œ Z ํŠธ๋ผ์ดํด๋“œโ€™๋Š” ์™„์ „ํžˆ ํŽผ์น˜๋ฉด 10์ธ์น˜๊ธ‰ ๋Œ€ํ™”๋ฉด์œผ๋กœ ์ „ํ™˜๋ผ ๋ฌธ์„œ ์ž‘์—…, ๋ฉ€ํ‹ฐํƒœ์Šคํ‚น, ์ฝ˜ํ…์ธ  ๊ฐ์ƒ ๋“ฑ ํƒœ๋ธ”๋ฆฟ ์ˆ˜์ค€์˜ ํ™œ์šฉ์ด ๊ฐ€๋Šฅํ•˜๊ณ , ์ ‘์—ˆ์„ ๋•Œ๋Š” ๋ฐ”ํ˜• ์Šค๋งˆํŠธํฐ ํฌ๊ธฐ๋กœ ์ค„์–ด ํœด๋Œ€์„ฑ์„ ํ™•๋ณดํ•œ๋‹ค. ์ ‘์—ˆ์„ ๋•Œ 12.9mm, ํŽผ์ณค์„ ๋•Œ ๊ฐ€์žฅ ์–‡์€ ์ชฝ์˜ ๋‘๊ป˜๊ฐ€ 3.9mm๋กœ ์—ญ๋Œ€ ๊ฐค๋Ÿญ์‹œ Z ํด๋“œ ์‹œ๋ฆฌ์ฆˆ ์ค‘ ๊ฐ€์žฅ ์Šฌ๋ฆผํ•œ ๋””์ž์ธ์„ ๊ฐ–์ท„๋‹ค.

์ œํ’ˆ ์„ค๊ณ„๋Š” โ€˜์ธํด๋”ฉโ€™ ๊ธฐ๋ฐ˜์˜ 3๋‹จ ์ ‘ํž˜ ๊ตฌ์กฐ๋ฅผ ์ฑ„ํƒํ–ˆ์œผ๋ฉฐ, ์ ‘ํžˆ๋Š” ๊ณผ์ •์—์„œ ์ด์ƒ ์ง•ํ›„๊ฐ€ ๊ฐ์ง€๋  ๊ฒฝ์šฐ ์•Œ๋ฆผ๊ณผ ์ง„๋™์œผ๋กœ ์‚ฌ์šฉ์ž์—๊ฒŒ ์•ˆ๋‚ดํ•˜๋Š” ์ž๋™ ์•Œ๋žŒ ๊ธฐ๋Šฅ๋„ ํƒ‘์žฌํ–ˆ๋‹ค. ๋ฐฐํ„ฐ๋ฆฌ๋Š” ํด๋”๋ธ” ์‹œ๋ฆฌ์ฆˆ ์ค‘ ์ตœ๋Œ€ ์ˆ˜์ค€์ธ 5,600mAh ์šฉ๋Ÿ‰์„ ๋‹ด์•˜๊ณ , 3๊ฐœ์˜ ํŒจ๋„ ๊ฐ๊ฐ์— ์…€์„ ๊ณ ๋ฅด๊ฒŒ ๋ฐฐ์น˜ํ•˜๋Š” ๋ฐฉ์‹์œผ๋กœ ์ „๋ ฅ ๊ณต๊ธ‰์˜ ๊ท ํ˜•์„ ๋งž์ท„๋‹ค. ์ดˆ๊ณ ์† ์ถฉ์ „ ์ง€์›์œผ๋กœ ๋Œ€ํ™”๋ฉด ์‚ฌ์šฉ ํ™˜๊ฒฝ์—์„œ๋„ ์ถฉ์ „ ํŽธ์˜์„ฑ์„ ์œ ์ง€ํ–ˆ๋‹ค.

์„ฑ๋Šฅ๊ณผ ์‚ฌ์šฉ์„ฑ ์—ญ์‹œ ๋Œ€ํ™”๋ฉด ํด๋”๋ธ”์— ๋งž์ถฐ ๊ฐ•ํ™”๋๋‹ค. โ€˜๊ฐค๋Ÿญ์‹œ์šฉ ์Šค๋ƒ…๋“œ๋ž˜๊ณค 8 ์—˜๋ฆฌํŠธโ€™ ๋ชจ๋ฐ”์ผ ํ”„๋กœ์„ธ์„œ๋ฅผ ํƒ‘์žฌํ•ด ๊ณ ํ•ด์ƒ๋„ ๋ฉ€ํ‹ฐํƒœ์Šคํ‚น๊ณผ AI ๊ธฐ๋Šฅ ๊ตฌ๋™์— ํ•„์š”ํ•œ ์ฒ˜๋ฆฌ ๋Šฅ๋ ฅ์„ ํ™•๋ณดํ–ˆ์œผ๋ฉฐ, 2์–ต ํ™”์†Œ ๊ด‘๊ฐ ์นด๋ฉ”๋ผ๋ฅผ ์ ์šฉํ•ด ํด๋”๋ธ”์—์„œ๋„ ๊ณ ํ’ˆ์งˆ ์ดฌ์˜ ๊ฒฝํ—˜์„ ์ œ๊ณตํ•œ๋‹ค.

Samsung Trifold

Samsung

๋˜ํ•œ ์‚ฌ์šฉ์ž๋Š” ๋ฉ€ํ‹ฐ ์œˆ๋„์šฐ ๊ธฐ๋Šฅ์„ ํ†ตํ•ด ์ตœ๋Œ€ 3๊ฐœ์˜ ์•ฑ์„ ๋‚˜๋ž€ํžˆ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๋‹ค. ๊ฐ€๋ น ์›น ์ฝ˜ํ…์ธ ๋ฅผ ๋ณด๋ฉด์„œ ์š”์•ฝยท๋ฒˆ์—ญ ๊ฒฐ๊ณผ๋ฅผ ํ•จ๊ป˜ ํ™•์ธํ•˜๊ฑฐ๋‚˜ ๊ฑด๊ฐ• ๋ฐ์ดํ„ฐ๋ฅผ ํ•œ๋ˆˆ์— ์ •๋ฆฌํ•ด ๋ณด๋Š” ๋“ฑ ๋‹ค์–‘ํ•œ ์•ฑ์„ ๋Œ€ํ™”๋ฉด์— ์ตœ์ ํ™”๋œ ๋ฐฉ์‹์œผ๋กœ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋‹ค. ๋˜ํ•œ ํ™”๋ฉด์œผ๋กœ ๋ณด๊ณ  ์žˆ๋Š” ์ •๋ณด๋‚˜ ์นด๋ฉ”๋ผ ์˜์ƒ์„ AI์™€ ์‹ค์‹œ๊ฐ„์œผ๋กœ ๊ณต์œ ํ•ด ์งˆ๋ฌธํ•˜๊ณ  ๋‹ต๋ณ€์„ ๋ฐ›๋Š” ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ AI ๊ฒฝํ—˜๋„ ์ง€์›ํ•œ๋‹ค.

ํŠนํžˆ ์ด๋ฒˆ ์ œํ’ˆ์€ ์Šค๋งˆํŠธํฐ ์ตœ์ดˆ๋กœ ํƒœ๋ธ”๋ฆฟ ๋ฒ„์ „์˜ ์‚ผ์„ฑ ๋ฑ์Šค๋ฅผ ์ง€์›ํ•ด, ์™ธ๋ถ€ ๋ชจ๋‹ˆํ„ฐ ์—†์ด๋„ ๋ฐ์Šคํฌํ†ฑ์— ๊ฐ€๊นŒ์šด ์ž‘์—… ํ™˜๊ฒฝ์„ ๊ตฌํ˜„ํ•œ๋‹ค. ์—ฌ๋Ÿฌ ๊ฐœ์˜ ๊ฐ€์ƒ ์ž‘์—… ๊ณต๊ฐ„์„ ์˜ค๊ฐ€๋ฉฐ ๊ฐ ๊ณต๊ฐ„์— ์—ฌ๋Ÿฌ ์•ฑ์„ ๋™์‹œ์— ๋ฐฐ์น˜ํ•  ์ˆ˜ ์žˆ์–ด, ์ด๋™ ์ค‘์—๋„ PC ์ˆ˜์ค€์˜ ๋ฉ€ํ‹ฐํƒœ์Šคํ‚น์„ ๊ตฌํ˜„ํ•œ๋‹ค.

์‚ผ์„ฑ์ „์ž๋Š” โ€˜๊ฐค๋Ÿญ์‹œ Z ํŠธ๋ผ์ดํด๋“œโ€™๋ฅผ ๊ตญ๋‚ด์— ๋จผ์ € ์ถœ์‹œํ•œ ๋’ค ์ฃผ์š” ํ•ด์™ธ ์‹œ์žฅ์œผ๋กœ ์ˆœ์ฐจ ํ™•๋Œ€ํ•  ๊ณ„ํš์ด๋‹ค. ๊ตญ๋‚ด๋Š” 12์ผ๋ถ€ํ„ฐ ์‚ผ์„ฑ๋‹ท์ปด๊ณผ ์ฃผ์š” ์˜คํ”„๋ผ์ธ ๋งค์žฅ์—์„œ ํŒ๋งคํ•˜๋ฉฐ, ์ถœ์‹œ ์ „ ์ฒดํ—˜ ๊ณต๊ฐ„์„ ์šด์˜ํ•ด ์ œํ’ˆ ๊ฒฝํ—˜ ๊ธฐํšŒ๋ฅผ ์ œ๊ณตํ•œ๋‹ค. ์ œํ’ˆ์€ ๋‹จ์ผ ์ƒ‰์ƒ๊ณผ 16GB ๋ฉ”๋ชจ๋ฆฌ, 512GB ์Šคํ† ๋ฆฌ์ง€ ๊ตฌ์„ฑ์œผ๋กœ ์ถœ์‹œ๋˜๊ณ , ๊ตฌ๋งค ๊ณ ๊ฐ์„ ๋Œ€์ƒ์œผ๋กœ AI ์„œ๋น„์Šค ๊ตฌ๋…๊ถŒ, ์ฝ˜ํ…์ธ  ๊ตฌ๋… ํ˜œํƒ, ๋ณดํ˜ธํ•„๋ฆ„ ์„œ๋น„์Šค, ์ˆ˜๋ฆฌ๋น„ ํ• ์ธ ๋“ฑ ํด๋”๋ธ” ํ”„๋ฆฌ๋ฏธ์—„ ํ”„๋กœ๊ทธ๋žจ์„ ํ•จ๊ป˜ ์ œ๊ณตํ•  ์˜ˆ์ •์ด๋‹ค.

๋…ธํƒœ๋ฌธ ์‚ผ์„ฑ์ „์ž ๋Œ€ํ‘œ์ด์‚ฌ ์‚ฌ์žฅ์€ โ€œ๊ฐค๋Ÿญ์‹œ Z ํŠธ๋ผ์ดํด๋“œโ€™๋Š” ์ƒˆ๋กœ์šด ํผํŒฉํ„ฐ ๋ถ„์•ผ์—์„œ ์Œ“์•„์˜จ ์‚ผ์„ฑ์ „์ž์˜ ๋ฆฌ๋”์‹ญ์„ ๋ฐ”ํƒ•์œผ๋กœ ์ƒ์‚ฐ์„ฑ๊ณผ ํœด๋Œ€์„ฑ์˜ ๊ท ํ˜•์„ ์‹คํ˜„ํ•œ ์ œํ’ˆ์ด๋ฉฐ ์—…๋ฌดโˆ™์ฐฝ์˜์„ฑโˆ™์—ฐ๊ฒฐ์„ฑ ๋“ฑ ๋ชจ๋ฐ”์ผ ์ „๋ฐ˜์˜ ๊ฒฝํ—˜์„ ํ•œ์ธต ํ™•์žฅํ•  ๊ฒƒโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

ใ€Œ้‡ๅญใธใฎๅ‚™ใˆใ€ใŒๅ•ใ„็›ดใ™ไผๆฅญITใฎๅฏฟๅ‘ฝ่จญ่จˆโ€•Post-Quantum Cryptographyๆจ™ๆบ–ๅŒ–ใŒใ‚‚ใŸใ‚‰ใ™ใƒ‘ใƒฉใƒ€ใ‚คใƒ ใ‚ทใƒ•ใƒˆ

1 December 2025 at 08:20

ใ—ใ‹ใ—ใ€ใใฎๅˆฐๆฅใ‚’ๅพ…ใคใ“ใจใชใใ€ไธ–็•Œใฎใ‚ปใ‚ญใƒฅใƒชใƒ†ใ‚ฃใ‚ณใƒŸใƒฅใƒ‹ใƒ†ใ‚ฃใฏใ™ใงใซไธๅฏ้€†ใช่ปขๆ›็‚นใ‚’่ถ…ใˆใฆใ„ใ‚‹ใ€‚2020ๅนดไปฃๅ‰ๅŠใ€็ฑณๅ›ฝๆจ™ๆบ–ๆŠ€่ก“ๅฑ€๏ผˆNIST๏ผ‰ใซใ‚ˆใ‚‹Post-Quantum Cryptography๏ผˆPQC๏ผš่€้‡ๅญ่จˆ็ฎ—ๆฉŸๆš—ๅท๏ผ‰ใฎๆจ™ๆบ–ๅŒ–ใƒ—ใƒญใ‚ปใ‚นใŒๆœ€็ต‚ๆฎต้šŽใ‚’่ฟŽใˆใ€ๆ–ฐใŸใช้€ฃ้‚ฆๆจ™ๆบ–๏ผˆFIPS๏ผ‰ใจใ—ใฆ็ตๅฎŸใ—ใ‚ˆใ†ใจใ—ใฆใ„ใ‚‹ใ‹ใ‚‰ใ ใ€‚ใ“ใ‚Œใฏๅ˜ใชใ‚‹ๆŠ€่ก“ไป•ๆง˜ใฎๆ”นๅฎšใงใฏใชใ„ใ€‚ใ‚คใƒณใ‚ฟใƒผใƒใƒƒใƒˆใฎ้ปŽๆ˜ŽๆœŸใ‹ใ‚‰็พไปฃใซ่‡ณใ‚‹ใพใงใ€ใƒ‡ใ‚ธใ‚ฟใƒซใฎไฟก้ ผใ‚’ๆ นๅบ•ใงๆ”ฏใˆใฆใใŸRSAๆš—ๅทใ‚„ๆฅ•ๅ††ๆ›ฒ็ทšๆš—ๅทใจใ„ใฃใŸใ€Œ็พไปฃๆš—ๅทใฎ็ต‚็„‰ใ€ใจใ€ใใ‚Œใซไปฃใ‚ใ‚‹ใ€Œๆฌกไธ–ไปฃๆš—ๅทใธใฎ็งป่กŒใ€ใจใ„ใ†ใ€ๆ•ฐๅๅนดๅ˜ไฝใฎๅทจๅคงใชๅœฐๆฎปๅค‰ๅ‹•ใŒๅง‹ใพใฃใŸใ“ใจใ‚’ๆ„ๅ‘ณใ—ใฆใ„ใ‚‹ใ€‚Googleใ‚„Cloudflareใจใ„ใฃใŸใƒ†ใ‚ฏใƒŽใƒญใ‚ธใƒผใฎๅทจไบบใฏใ™ใงใซใƒ–ใƒฉใ‚ฆใ‚ถใ‚„ใƒใƒƒใƒˆใƒฏใƒผใ‚ฏใƒฌใƒ™ใƒซใงใฎๅฎŸ่ฃ…ๅฎŸ้จ“ใ‚’็นฐใ‚Š่ฟ”ใ—ใฆใŠใ‚Šใ€ๅคงๆ‰‹ใ‚ฏใƒฉใ‚ฆใƒ‰ใƒ™ใƒณใƒ€ใƒผใ‚‚ๆฐด้ขไธ‹ใงๅฏพๅฟœใ‚’้€ฒใ‚ใฆใ„ใ‚‹ใ€‚ๆœฌ็จฟใงใฏใ€ใพใ ่ฆ‹ใฌ่„…ๅจใงใ‚ใ‚‹ใฏใšใฎ้‡ๅญ่จˆ็ฎ—ๆฉŸๅฏพ็ญ–ใŒใ€ใชใœไปŠใ€ไผๆฅญใฎๅ–ซ็ทŠใฎ่ชฒ้กŒใจใ—ใฆๆตฎไธŠใ—ใฆใ„ใ‚‹ใฎใ‹ใ€ใใ—ใฆPQCใธใฎ็งป่กŒใŒไผๆฅญITใฎ้•ทๆœŸๆˆฆ็•ฅใซใฉใฎใ‚ˆใ†ใชๅค‰้ฉใ‚’่ฟซใ‚‹ใฎใ‹ใ‚’ใ€ๆŠ€่ก“็š„่ƒŒๆ™ฏใจ็ตŒๅ–ถ็š„ใƒชใ‚นใ‚ฏใฎไธก้ขใ‹ใ‚‰่ฉณ่ชฌใ™ใ‚‹ใ€‚

ใ€ŒHarvest Now, Decrypt Laterใ€ใฎ่„…ๅจใจๆจ™ๆบ–ๅŒ–ใฎๅŠ ้€Ÿ

ไธ€่ˆฌใซใ€้‡ๅญใ‚ณใƒณใƒ”ใƒฅใƒผใ‚ฟใซใ‚ˆใ‚‹ๆš—ๅท่งฃ่ชญใฎ่ฉฑ้กŒใŒๅ‡บใ‚‹ใจใ€ๅคšใใฎ็ตŒๅ–ถ่€…ใ‚„IT่ฒฌไปป่€…ใฏใ€Œใพใ ๅฎŸ็”จๅŒ–ใซใฏๆ™‚้–“ใŒใ‹ใ‹ใ‚‹ใ€ใ€Œ่‡ชๅˆ†ใŒ็พๅฝนใฎ้–“ใฏ้–ขไฟ‚ใชใ„ใ€ใจๆ‰ใˆใŒใกใงใ‚ใ‚‹ใ€‚็ขบใ‹ใซใ€Shorใฎใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใ‚’็”จใ„ใฆ็พๅœจใฎๅ…ฌ้–‹้ตๆš—ๅทใ‚’็พๅฎŸ็š„ใชๆ™‚้–“ใง็ ดใ‚‹ใŸใ‚ใซๅฟ…่ฆใชใ€ๅคง่ฆๆจกใ‹ใค่ชคใ‚Š่จ‚ๆญฃๆฉŸ่ƒฝใ‚’ๅ‚™ใˆใŸ้‡ๅญใ‚ณใƒณใƒ”ใƒฅใƒผใ‚ฟใฎๅฎŸ็พใฏใ€ไพ็„ถใจใ—ใฆๆŠ€่ก“็š„ใชใƒใƒผใƒ‰ใƒซใŒ้ซ˜ใ„ใ€‚ใ—ใ‹ใ—ใ€ไผๆฅญใŒ็›ด้ขใ—ใฆใ„ใ‚‹ใƒชใ‚นใ‚ฏใฎๅฎŸไฝ“ใฏใ€ๅฐ†ๆฅใฎ่งฃ่ชญ่ƒฝๅŠ›ใใฎใ‚‚ใฎใงใฏใชใใ€็พๅœจใฎใƒ‡ใƒผใ‚ฟใŒๆœชๆฅใฎๆ™‚็‚นใงๅฑ้™บใซๆ™’ใ•ใ‚Œใ‚‹ใจใ„ใ†ๆ™‚้–“่ปธใฎใ‚บใƒฌใซใ‚ใ‚‹ใ€‚ ใ‚ปใ‚ญใƒฅใƒชใƒ†ใ‚ฃๆฅญ็•Œใงใ€ŒHarvest Now, Decrypt Later๏ผˆไปŠๅŽ้›†ใ—ใ€ๅพŒใง่งฃ่ชญใ™ใ‚‹๏ผ‰ใ€ใจๅ‘ผใฐใ‚Œใ‚‹ๆ”ปๆ’ƒๆ‰‹ๆณ•ใธใฎๆ‡ธๅฟตใŒใ€ๅ„ๅ›ฝใฎๆจ™ๆบ–ๅŒ–ใ‚’ๆ€ฅใŒใ›ใ‚‹ๆœ€ๅคงใฎ่ฆๅ› ใจใชใฃใฆใ„ใ‚‹ใ€‚ๆ”ปๆ’ƒ่€…ใฏใ€ใŸใจใˆ็พๆ™‚็‚นใงใฏๆš—ๅทใ‚’่งฃ่ชญใงใใชใใจใ‚‚ใ€ใ‚ฟใƒผใ‚ฒใƒƒใƒˆใจใ™ใ‚‹ไผๆฅญใฎ้€šไฟกใƒ‡ใƒผใ‚ฟใ‚„ๆš—ๅทๅŒ–ใ•ใ‚ŒใŸใƒ•ใ‚กใ‚คใƒซใ‚’ไปŠใฎใ†ใกใซๅคง้‡ใซๅŽ้›†ใƒปไฟๅญ˜ใ—ใฆใŠใใ“ใจใŒๅฏ่ƒฝใงใ‚ใ‚‹ใ€‚ใใ—ใฆใ€ๅฐ†ๆฅ็š„ใซๅๅˆ†ใชๆ€ง่ƒฝใ‚’ๆŒใค้‡ๅญใ‚ณใƒณใƒ”ใƒฅใƒผใ‚ฟใŒ็™ปๅ ดใ—ใŸ็žฌ้–“ใซใ€้ŽๅŽปใซ่“„็ฉใ—ใŸใƒ‡ใƒผใ‚ฟใ‚’ไธ€ๆฐ—ใซ่งฃ่ชญใซใ‹ใ‘ใ‚‹ใฎใ ใ€‚ใ“ใฎใ‚ทใƒŠใƒชใ‚ชใซใŠใ„ใฆใ€ๆš—ๅทๅŒ–้€šไฟกใฎๅฎ‰ๅ…จๆ€งใฏใ€Œ้€šไฟกใ—ใฆใ„ใ‚‹็žฌ้–“ใ€ใ ใ‘ใงใฏๅฎŒ็ตใ—ใชใ„ใ€‚ใใฎๆƒ…ๅ ฑใŒๆŒใคๆฉŸๅฏ†ๆ€งใฎไฟๆŒๆœŸ้–“ใจใ€ๆš—ๅทๆŠ€่ก“ใŒ็ ดใ‚‰ใ‚Œใ‚‹ใพใงใฎๆœŸ้–“ใฎ็ซถ่ตฐใจใชใ‚‹ใ€‚

ใŸใจใˆใฐใ€ๅ›ฝๅฎถๆฉŸๅฏ†ใซ้–ขใ‚ใ‚‹ๅค–ไบคๆ–‡ๆ›ธใ€็Ÿฅ็š„่ฒก็”ฃใจใชใ‚‹ๆ–ฐ่–ฌใฎ็ ”็ฉถใƒ‡ใƒผใ‚ฟใ€ใ‚ใ‚‹ใ„ใฏๅ€‹ไบบใฎ้บไผๅญๆƒ…ๅ ฑใ‚„้•ทๆœŸใฎ้‡‘่ž่ณ‡็”ฃ่จ˜้Œฒใชใฉใฏใ€10ๅนดใ‹ใ‚‰ๆ•ฐๅๅนดใจใ„ใ†ๆฅตใ‚ใฆ้•ทใ„ๆœŸ้–“ใซใ‚ใŸใฃใฆๆฉŸๅฏ†ๆ€งใ‚’็ถญๆŒใ™ใ‚‹ๅฟ…่ฆใŒใ‚ใ‚‹ใ€‚ใ‚‚ใ—2030ๅนดไปฃใ‚ใ‚‹ใ„ใฏ2040ๅนดไปฃใซ้‡ๅญ่จˆ็ฎ—ๆฉŸใŒๅฎŸ็”จๅŒ–ใ•ใ‚Œใ‚‹ใจไปฎๅฎšใ™ใ‚Œใฐใ€ไปŠๆ—ฅ้€ไฟกใ•ใ‚Œใฆใ„ใ‚‹้•ทๆœŸไฟๅญ˜ใƒ‡ใƒผใ‚ฟใฎๅคšใใฏใ€ใ™ใงใซๅฑ้™บๆฐดๅŸŸใซๅ…ฅใฃใฆใ„ใ‚‹ใจ่จ€ใˆใ‚‹ใ ใ‚ใ†ใ€‚NISTใŒ้ธๅฎšใ—ใŸCRYSTALS-Kyber๏ผˆ้ตๅ…ฑๆœ‰ๆ–นๅผใ€ๆจ™ๆบ–ๅŒ–ๅ็งฐ๏ผšML-KEM๏ผ‰ใ‚„CRYSTALS-Dilithium๏ผˆ็ฝฒๅๆ–นๅผใ€ๆจ™ๆบ–ๅŒ–ๅ็งฐ๏ผšML-DSA๏ผ‰ใจใ„ใฃใŸๆ–ฐใ—ใ„ใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใฏใ€ใ“ใ†ใ—ใŸใ€Œๆœชๆฅใฎ่„…ๅจใซใ‚ˆใ‚‹็พๅœจใฎใƒชใ‚นใ‚ฏใ€ใ‚’ๅฐใ˜่พผใ‚ใ‚‹ใŸใ‚ใฎ้˜ฒๆณขๅ คใงใ‚ใ‚‹ใ€‚ใ“ใ‚Œใ‚‰ใฏๅพ“ๆฅใฎ็ด ๅ› ๆ•ฐๅˆ†่งฃใ‚„้›ขๆ•ฃๅฏพๆ•ฐๅ•้กŒใจใฏ็•ฐใชใ‚‹ใ€ๆ ผๅญๆš—ๅทใชใฉใฎๆ•ฐๅญฆ็š„้›ฃๅ•ใ‚’ๅฎ‰ๅ…จๆ€งใฎๆ นๆ‹ ใจใ—ใฆใŠใ‚Šใ€้‡ๅญ่จˆ็ฎ—ๆฉŸใซใ‚ˆใ‚‹ๆ”ปๆ’ƒใซ่€ใˆใ†ใ‚‹ใจ่€ƒใˆใ‚‰ใ‚Œใฆใ„ใ‚‹ใ€‚ใ™ใงใซใ“ใ‚Œใ‚‰ใฎใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใฏFIPS๏ผˆ้€ฃ้‚ฆๆƒ…ๅ ฑๅ‡ฆ็†ๆจ™ๆบ–๏ผ‰ใจใ—ใฆๆ–‡ๆ›ธๅŒ–ใŒ้€ฒใฟใ€TLS๏ผˆTransport Layer Security๏ผ‰ใ‚„VPNใ€้›ปๅญ็ฝฒๅใจใ„ใฃใŸ็คพไผšใ‚คใƒณใƒ•ใƒฉใฎๆทฑๅฑคใธใฎ็ต„ใฟ่พผใฟใŒๅ‰ๆใจใชใ‚Šใคใคใ‚ใ‚‹ใ€‚ใคใพใ‚Šใ€PQCใฏ้ ใ„ๆœชๆฅใฎๆŠ€่ก“ใงใฏใชใใ€ใ™ใงใซๅฎŸ่ฃ…ใƒ•ใ‚งใƒผใ‚บใซๅ…ฅใฃใŸใ€Œ็พไปฃใฎๆŠ€่ก“ใ€ใชใฎใงใ‚ใ‚‹ใ€‚

ใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใฎใ€Œใƒชใƒ—ใƒฌใƒผใ‚นใ€ใ‚’่ถ…ใˆใŸใ‚ทใ‚นใƒ†ใƒ ๅŸบ็›คใธใฎ่กๆ’ƒ

PQCใธใฎ็งป่กŒใŒไผๆฅญITใซใจใฃใฆๆฅตใ‚ใฆๅŽ„ไป‹ใชใฎใฏใ€ใใ‚ŒใŒๅ˜ใชใ‚‹ใ‚ฝใƒ•ใƒˆใ‚ฆใ‚งใ‚ขใฎใ‚ขใƒƒใƒ—ใƒ‡ใƒผใƒˆใ‚„ใ€่จญๅฎšใƒ•ใ‚กใ‚คใƒซใฎๆ›ธใๆ›ใˆ็จ‹ๅบฆใงใฏๆธˆใพใชใ„ๅฏ่ƒฝๆ€งใŒ้ซ˜ใ„ใจใ„ใ†็‚นใซใ‚ใ‚‹ใ€‚ใ‹ใคใฆDESใ‹ใ‚‰AESใธใ€ใ‚ใ‚‹ใ„ใฏSHA-1ใ‹ใ‚‰SHA-2ใธใจๆš—ๅทใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใŒ็งป่กŒใ—ใŸ้š›ใ‚‚็›ธๅฟœใฎๅŠดๅŠ›ใ‚’่ฆใ—ใŸใŒใ€ไปŠๅ›žใฎPQC็งป่กŒใฏใใ‚Œใ‚‰ใจใฏๆฏ”่ผƒใซใชใ‚‰ใชใ„ใปใฉใ‚ทใ‚นใƒ†ใƒ ๅŸบ็›คใธใฎ็‰ฉ็†็š„ใƒป่ซ–็†็š„ใชใ‚คใƒณใƒ‘ใ‚ฏใƒˆใŒๅคงใใ„ใ€‚ ใใฎๆœ€ๅคงใฎ่ฆๅ› ใฏใ€้ตใ‚ตใ‚คใ‚บใจ็ฝฒๅใƒ‡ใƒผใ‚ฟใ‚ตใ‚คใ‚บใฎ่‚ฅๅคงๅŒ–ใงใ‚ใ‚‹ใ€‚ใŸใจใˆใฐใ€้ตๅ…ฑๆœ‰ใƒกใ‚ซใƒ‹ใ‚บใƒ ใงใ‚ใ‚‹ML-KEMใฏใ€็พๅœจไธปๆตใฎๆฅ•ๅ††ๆ›ฒ็ทšๆš—ๅท๏ผˆECDHใชใฉ๏ผ‰ใจๆฏ”่ผƒใ—ใฆใ€้ต้•ทใ‚„ๆš—ๅทๆ–‡ใฎใ‚ตใ‚คใ‚บใŒๆก้•ใ„ใซๅคงใใใชใ‚‹ๅ‚พๅ‘ใŒใ‚ใ‚‹ใ€‚ๅŒๆง˜ใซใ€้›ปๅญ็ฝฒๅใซ็”จใ„ใ‚‰ใ‚Œใ‚‹ML-DSAใ‚‚ใ€ๅพ“ๆฅใฎRSAใ‚„ECDSAใซๆฏ”ในใฆ็ฝฒๅใ‚ตใ‚คใ‚บใŒๅข—ๅคงใ™ใ‚‹ใ€‚ๆœ€ๆ–ฐใฎ้ซ˜ๆ€ง่ƒฝใ‚ตใƒผใƒใƒผใ‚„ๅคชใ„ๅธฏๅŸŸใ‚’ๆŒใคใƒใƒƒใ‚ฏใƒœใƒผใƒณๅ›ž็ทšใงใ‚ใ‚Œใฐใ€ใ“ใฎ็จ‹ๅบฆใฎใ‚ชใƒผใƒใƒผใƒ˜ใƒƒใƒ‰ใฏ่จฑๅฎน็ฏ„ๅ›ฒใ‹ใ‚‚ใ—ใ‚Œใชใ„ใ€‚ใ—ใ‹ใ—ใ€ใƒชใ‚ฝใƒผใ‚นใŒๅŽณใ—ใๅˆถ้™ใ•ใ‚ŒใŸ็’ฐๅขƒใซใŠใ„ใฆใฏใ€ใ“ใฎใ€Œ้‡ใ•ใ€ใŒ่‡ดๅ‘ฝ็š„ใชใƒœใƒˆใƒซใƒใƒƒใ‚ฏใจใชใ‚Šๅพ—ใ‚‹ใ€‚

ๅ…ทไฝ“็š„ใซๅฝฑ้ŸฟใŒๆ‡ธๅฟตใ•ใ‚Œใ‚‹ใฎใฏใ€IoT๏ผˆInternet of Things๏ผ‰ใ‚„OT๏ผˆOperational Technology๏ผ‰ใฎ้ ˜ๅŸŸใงใ‚ใ‚‹ใ€‚ๅทฅๅ ดๅ†…ใฎใ‚ปใƒณใ‚ตใƒผใ€่‡ชๅ‹•่ปŠใฎๅˆถๅพกใƒฆใƒ‹ใƒƒใƒˆใ€ใ‚นใƒžใƒผใƒˆใƒกใƒผใ‚ฟใƒผใ€ใ‚ใ‚‹ใ„ใฏๅŒป็™‚็”จๅŸ‹ใ‚่พผใฟใƒ‡ใƒใ‚คใ‚นใชใฉใฏใ€ๆฅตใ‚ใฆ้™ใ‚‰ใ‚ŒใŸใƒกใƒขใƒชใจ่จˆ็ฎ—่ƒฝๅŠ›ใงๅ‹•ไฝœใ—ใฆใŠใ‚Šใ€้€šไฟกๅธฏๅŸŸใ‚‚็‹ญใ„ๅ ดๅˆใŒๅคšใ„ใ€‚ใ“ใ“ใซใ‚ตใ‚คใ‚บใฎๅคงใใชPQCใ‚’ใใฎใพใพๅฐŽๅ…ฅใ—ใ‚ˆใ†ใจใ™ใ‚Œใฐใ€ใƒ‘ใ‚ฑใƒƒใƒˆๅˆ†ๅ‰ฒใซใ‚ˆใ‚‹้…ๅปถใฎๅข—ๅคงใ€ใƒกใƒขใƒชไธ่ถณใซใ‚ˆใ‚‹ๅ‹•ไฝœไธๅฎ‰ๅฎšใ€ใ‚ใ‚‹ใ„ใฏใƒใƒณใƒ‰ใ‚ทใ‚งใ‚คใ‚ฏๅ‡ฆ็†ใซใ‚ˆใ‚‹ใƒใƒƒใƒ†ใƒชใƒผๆถˆ่ฒปใฎๆฟ€ๅข—ใจใ„ใฃใŸๅ•้กŒใŒ้ก•ๅœจๅŒ–ใ™ใ‚‹ๆใ‚ŒใŒใ‚ใ‚‹ใ€‚ ใ•ใ‚‰ใซใ€ๆ—ขๅญ˜ใฎใƒ—ใƒญใƒˆใ‚ณใƒซใ‚„ใƒ‡ใƒผใ‚ฟใƒ•ใ‚ฉใƒผใƒžใƒƒใƒˆใŒใ€ใ“ใ‚Œใปใฉๅคงใใช้ตใ‚„็ฝฒๅใ‚’ๆ ผ็ดใ™ใ‚‹ใ“ใจใ‚’ๆƒณๅฎšใ—ใฆใ„ใชใ„ใ‚ฑใƒผใ‚นใ‚‚ๅคšใ€…ใ‚ใ‚‹ใ€‚X.509่จผๆ˜Žๆ›ธใซPQCใฎๅ…ฌ้–‹้ตใ‚„็ฝฒๅใ‚’ๅŸ‹ใ‚่พผใ‚“ใ ็ตๆžœใ€่จผๆ˜Žๆ›ธใฎใ‚ตใ‚คใ‚บใŒ่‚ฅๅคงๅŒ–ใ—ใ€ๅพ“ๆฅใฎUDPใƒ™ใƒผใ‚นใฎ้€šไฟกใงใƒ‘ใ‚ฑใƒƒใƒˆใ‚ตใ‚คใ‚บๅˆถ้™ใซๆŠต่งฆใ—ใŸใ‚Šใ€ๅคใ„ใƒŸใƒ‰ใƒซใ‚ฆใ‚งใ‚ขใŒใƒใƒƒใƒ•ใ‚กใ‚ชใƒผใƒใƒผใƒ•ใƒญใƒผใ‚’่ตทใ“ใ—ใŸใ‚Šใ™ใ‚‹ใƒชใ‚นใ‚ฏใ‚‚ๆŒ‡ๆ‘˜ใ•ใ‚Œใฆใ„ใ‚‹ใ€‚

ใพใŸใ€็งป่กŒๆœŸ็‰นๆœ‰ใฎ่ค‡้›‘ใ•ใจใ—ใฆใ€Œใƒใ‚คใƒ–ใƒชใƒƒใƒ‰ๆš—ๅทใ€ใฎ้‹็”จใŒๆŒ™ใ’ใ‚‰ใ‚Œใ‚‹ใ€‚PQCใฎใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใฏๆฏ”่ผƒ็š„ๆ–ฐใ—ใ„ใŸใ‚ใ€ๅฐ†ๆฅ็š„ใซๆœช็Ÿฅใฎ่„†ๅผฑๆ€งใŒ่ฆ‹ใคใ‹ใ‚‹ๅฏ่ƒฝๆ€งใ‚’ๅฎŒๅ…จใซใฏๅฆๅฎšใงใใชใ„ใ€‚ใใฎใŸใ‚ใ€็งป่กŒๆœŸ้–“ไธญใฏใ€้•ทๅนดใฎๅฎŸ็ธพใŒใ‚ใ‚‹ๅพ“ๆฅใฎๆฅ•ๅ††ๆ›ฒ็ทšๆš—ๅทใจใ€ๆ–ฐใ—ใ„PQCใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใ‚’็ต„ใฟๅˆใ‚ใ›ใฆไบŒ้‡ใซ้ตๅ…ฑๆœ‰ใ‚’่กŒใ†ใ€Œใƒใ‚คใƒ–ใƒชใƒƒใƒ‰ๆ–นๅผใ€ใŒๆŽจๅฅจใ•ใ‚Œใฆใ„ใ‚‹ใ€‚ใ“ใ‚Œใฏๅฎ‰ๅ…จๆ€งใซใŠใ‘ใ‚‹ไฟ้™บใจใ—ใฆใฏๅˆ็†็š„ใ ใŒใ€ใ‚ทใ‚นใƒ†ใƒ ้‹็”จๅดใ‹ใ‚‰่ฆ‹ใ‚Œใฐใ€็ฎก็†ใ™ในใ้ตใฎ็จฎ้กžใŒๅข—ใˆใ€ๅ‡ฆ็†่ฒ ่ทใŒๅข—ๅคงใ—ใ€ใƒˆใƒฉใƒ–ใƒซใ‚ทใƒฅใƒผใƒ†ใ‚ฃใƒณใ‚ฐใŒ่ค‡้›‘ๅŒ–ใ™ใ‚‹ใ“ใจใ‚’ๆ„ๅ‘ณใ™ใ‚‹ใ€‚ๆ—ขๅญ˜ใฎRSAใ‚„ECDSAใฎใฟใซๅฏพๅฟœใ—ใŸใƒฌใ‚ฌใ‚ทใƒผๆฉŸๅ™จใจใ€PQCๅฏพๅฟœใฎๆœ€ๆ–ฐๆฉŸๅ™จใŒๆททๅœจใ™ใ‚‹็’ฐๅขƒใ‚’ใ€ใ‚ปใ‚ญใƒฅใƒชใƒ†ใ‚ฃใƒใƒชใ‚ทใƒผใฎไธ€่ฒซๆ€งใ‚’ไฟใกใชใŒใ‚‰ใฉใ†็ตฑๅˆ็ฎก็†ใ—ใฆใ„ใใฎใ‹ใ€‚ไผๆฅญใฏใ€ใƒใƒƒใƒˆใƒฏใƒผใ‚ฏๆฉŸๅ™จใฎ่ฒทใ„ๆ›ฟใˆใ‚ตใ‚คใ‚ฏใƒซใ‚„ใ‚ขใƒ—ใƒชใ‚ฑใƒผใ‚ทใƒงใƒณใฎๆ”นไฟฎ่จˆ็”ปใ‚’ๅซใ‚ใŸใ€้•ทๆœŸ็š„ใชใƒญใƒผใƒ‰ใƒžใƒƒใƒ—ใฎ็ญ–ๅฎšใ‚’่ฟซใ‚‰ใ‚Œใ‚‹ใ“ใจใซใชใ‚‹ใ€‚

ๆš—ๅทใƒฉใ‚คใƒ•ใ‚ตใ‚คใ‚ฏใƒซ็ฎก็†๏ผˆCLM๏ผ‰ใจใ€Œใ‚ฏใƒชใƒ—ใƒˆใƒปใ‚ขใ‚ธใƒชใƒ†ใ‚ฃใ€ใฎ็ขบ็ซ‹

ใ“ใฎใ‚ˆใ†ใชๆŠ€่ก“็š„ใƒป้‹็”จ็š„ใช่ชฒ้กŒใ‚’ๅ‰ใซใ—ใฆใ€ไผๆฅญใฏใฉใฎใ‚ˆใ†ใชๆˆฆ็•ฅใ‚’ๆŒใคในใใชใฎใ‹ใ€‚ๆœ€ใ‚‚้‡่ฆใช่ฆ–็‚นใฏใ€PQCๅฏพๅฟœใ‚’ๅ˜ใชใ‚‹ใ€Œ20XXๅนดๅ•้กŒใ€ใฎใ‚ˆใ†ใชๆœŸ้™ไป˜ใใฎๅฏพๅ‡ฆ็™‚ๆณ•ใจใ—ใฆๆ‰ใˆใ‚‹ใฎใงใฏใชใใ€็ต„็น”ๅ…จไฝ“ใฎใ€Œๆš—ๅทใฎ็ฎก็†่ƒฝๅŠ›๏ผˆใ‚ฏใƒชใƒ—ใƒˆใƒปใ‚ขใ‚ธใƒชใƒ†ใ‚ฃ๏ผ‰ใ€ใ‚’ๆŠœๆœฌ็š„ใซๅผทๅŒ–ใ™ใ‚‹ๆฉŸไผšใจๆ‰ใˆใ‚‹ใ“ใจใงใ‚ใ‚‹ใ€‚ ใ‚ฏใƒชใƒ—ใƒˆใƒปใ‚ขใ‚ธใƒชใƒ†ใ‚ฃใจใฏใ€ไฝฟ็”จใ—ใฆใ„ใ‚‹ๆš—ๅทใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใซๅฑๆฎ†ๅŒ–๏ผˆๅฎ‰ๅ…จๆ€งใŒๆใชใ‚ใ‚Œใ‚‹ใ“ใจ๏ผ‰ใ‚„่„†ๅผฑๆ€งใŒ็™บ่ฆ‹ใ•ใ‚ŒใŸ้š›ใซใ€ใ‚ทใ‚นใƒ†ใƒ ๅ…จไฝ“ใธใฎๅฝฑ้Ÿฟใ‚’ๆœ€ๅฐ้™ใซๆŠ‘ใˆใคใคใ€่ฟ…้€Ÿใ‹ใคใ‚นใƒ ใƒผใ‚บใซๆ–ฐใ—ใ„ๅฎ‰ๅ…จใชใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใธๅˆ‡ใ‚Šๆ›ฟใˆใ‚‹่ƒฝๅŠ›ใ‚’ๆŒ‡ใ™ใ€‚ใ“ใ‚Œใพใงๅคšใใฎไผๆฅญใ‚ทใ‚นใƒ†ใƒ ใงใฏใ€ๆš—ๅทใ‚ขใƒซใ‚ดใƒชใ‚บใƒ ใฏไธ€ๅบฆๅฎŸ่ฃ…ใ•ใ‚Œใ‚Œใฐใ€ใ‚ทใ‚นใƒ†ใƒ ใŒๅปƒๆฃ„ใ•ใ‚Œใ‚‹ใพใงๅกฉๆผฌใ‘ใซใ•ใ‚Œใ‚‹ใ“ใจใŒไธ€่ˆฌ็š„ใงใ‚ใฃใŸใ€‚ใ‚ขใƒ—ใƒชใ‚ฑใƒผใ‚ทใƒงใƒณใฎใ‚ณใƒผใƒ‰ใฎไธญใซใƒใƒผใƒ‰ใ‚ณใƒผใƒ‰ใ•ใ‚Œใฆใ„ใŸใ‚Šใ€ใƒใƒผใƒ‰ใ‚ฆใ‚งใ‚ขใƒใƒƒใƒ—ใซ็„ผใไป˜ใ‘ใ‚‰ใ‚Œใฆใ„ใŸใ‚Šใ—ใฆใ€ๅฎนๆ˜“ใซๅค‰ๆ›ดใงใใชใ„ๆง‹้€ ใซใชใฃใฆใ„ใ‚‹ใ“ใจใŒๅคšใ‹ใฃใŸใฎใงใ‚ใ‚‹ใ€‚PQCใธใฎ็งป่กŒใฏใ€ใ“ใ†ใ—ใŸ็กฌ็›ด็š„ใชๆง‹้€ ใ‚’ๆ‰“็ ดใ—ใ€ๆš—ๅท้ƒจๅ“ใ‚’ใ„ใคใงใ‚‚ไบคๆ›ๅฏ่ƒฝใชใ€Œ้ƒจๅ“ใ€ใจใ—ใฆ็–Ž็ตๅˆๅŒ–ใ™ใ‚‹ใ‚ขใƒผใ‚ญใƒ†ใ‚ฏใƒใƒฃใธใฎ่ปขๆ›ใ‚’ไฟƒใ™ใ‚‚ใฎใงใ‚ใ‚‹ใ€‚

ๅ…ทไฝ“็š„ใช็ฌฌไธ€ๆญฉใจใ—ใฆๆฑ‚ใ‚ใ‚‰ใ‚Œใ‚‹ใฎใฏใ€่‡ช็คพใฎใ‚ทใ‚นใƒ†ใƒ ่ณ‡็”ฃใซใŠใ‘ใ‚‹ๆš—ๅทไพๅญ˜้–ขไฟ‚ใฎๅฎŒๅ…จใชๆฃšๅธใ—ใงใ‚ใ‚‹ใ€‚ใ“ใ‚Œใ‚’่ฟ‘ๅนดใงใฏใ€ŒCBOM๏ผˆCryptography Bill of Materials๏ผšๆš—ๅท้ƒจๅ“่กจ๏ผ‰ใ€ใจๅ‘ผใถๅ‹•ใใ‚‚ใ‚ใ‚‹ใ€‚ใฉใฎใ‚ตใƒผใƒใƒผใฎใฉใฎใƒฉใ‚คใƒ–ใƒฉใƒชใงใ€ใฉใฎใƒใƒผใ‚ธใƒงใƒณใฎOpenSSLใŒๅ‹•ใ„ใฆใ„ใ‚‹ใฎใ‹ใ€‚็‹ฌ่‡ช้–‹็™บใฎใ‚ขใƒ—ใƒชใ‚ฑใƒผใ‚ทใƒงใƒณๅ†…ใงใ€ใฉใฎใ‚ˆใ†ใชๆš—ๅท้–ขๆ•ฐใŒๅ‘ผใณๅ‡บใ•ใ‚Œใฆใ„ใ‚‹ใฎใ‹ใ€‚ๅค–้ƒจใจๆŽฅ็ถšใ™ใ‚‹VPN่ฃ…็ฝฎใ‚„ใ€ใ‚ฏใƒฉใ‚ฆใƒ‰ใ‚ตใƒผใƒ“ใ‚นใจใฎAPI้€ฃๆบใซใŠใ„ใฆใ€ใฉใฎๆš—ๅทใ‚นใ‚คใƒผใƒˆใŒไฝฟใ‚ใ‚Œใฆใ„ใ‚‹ใฎใ‹ใ€‚ใใ—ใฆไฝ•ใ‚ˆใ‚Šใ€ใใ‚Œใžใ‚Œใฎใ‚ทใ‚นใƒ†ใƒ ใงๆ‰ฑใ‚ใ‚Œใฆใ„ใ‚‹ใƒ‡ใƒผใ‚ฟใฎไฟๅญ˜ๆœŸ้–“ใฏใฉใ‚Œใใ‚‰ใ„ใชใฎใ‹ใ€‚ใ“ใ‚Œใ‚‰ใ‚’็ถฒ็พ…็š„ใซๅฏ่ฆ–ๅŒ–ใ—ใชใ„้™ใ‚Šใ€ใฉใ“ใ‹ใ‚‰ๆ‰‹ใ‚’ใคใ‘ใ‚‹ในใใ‹ใฎๅ„ชๅ…ˆ้ †ไฝใ™ใ‚‰ๆฑบใ‚ใ‚‹ใ“ใจใŒใงใใชใ„ใ€‚ ็‰นใซใ€ๆƒ…ๅ ฑใฎใ€Œๅฏฟๅ‘ฝใ€ใจๆš—ๅทใฎใ€Œๅฏฟๅ‘ฝใ€ใฎใ‚ฎใƒฃใƒƒใƒ—ใŒๅคงใใ„้ ˜ๅŸŸใ“ใใŒใ€ๆœ€ๅ„ชๅ…ˆใงๅฏพ็ญ–ใ‚’่ฌ›ใ˜ใ‚‹ในใใƒ›ใƒƒใƒˆใ‚นใƒใƒƒใƒˆใจใชใ‚‹ใ€‚ใŸใจใˆใฐใ€ๆณ•ๅฎšไฟๅญ˜ๆœŸ้–“ใŒ้•ทใ„ๆ–‡ๆ›ธ็ฎก็†ใ‚ทใ‚นใƒ†ใƒ ใ‚„ใ€่ฃฝๅ“ๅฏฟๅ‘ฝใŒ20ๅนดใซๅŠใถใ‚คใƒณใƒ•ใƒฉ่จญๅ‚™ใฎๅˆถๅพกใ‚ทใ‚นใƒ†ใƒ ใชใฉใฏใ€ๆฑŽ็”จ็š„ใชITๆฉŸๅ™จใ‚ˆใ‚Šใ‚‚ใฏใ‚‹ใ‹ใซๆ—ฉใ„ๆฎต้šŽใงPQCๅฏพๅฟœใ€ใ‚ใ‚‹ใ„ใฏใƒใ‚คใƒ–ใƒชใƒƒใƒ‰ๆง‹ๆˆใธใฎ็งป่กŒ่จˆ็”ปใ‚’็ซ‹ใฆใ‚‹ๅฟ…่ฆใŒใ‚ใ‚‹ใ ใ‚ใ†ใ€‚้€†ใซใ€ๆ•ฐๆ™‚้–“ใงไพกๅ€คใ‚’ๅคฑใ†ไธ€ๆ™‚็š„ใชใƒญใ‚ฐใƒ‡ใƒผใ‚ฟใ‚„ใ€็ŸญๆœŸ้–“ใง็ ดๆฃ„ใ•ใ‚Œใ‚‹ใ‚ญใƒฃใƒƒใ‚ทใƒฅใƒ‡ใƒผใ‚ฟใงใ‚ใ‚Œใฐใ€็›ดใกใซใ‚ณใ‚นใƒˆใ‚’ใ‹ใ‘ใฆPQCใ‚’ๅฐŽๅ…ฅใ™ใ‚‹ๅฟ…่ฆๆ€งใฏไฝŽใ„ใ‹ใ‚‚ใ—ใ‚Œใชใ„ใ€‚

็ตๅฑ€ใฎใจใ“ใ‚ใ€Post-Quantumๆ™‚ไปฃใซใŠใ‘ใ‚‹ใ‚ปใ‚ญใƒฅใƒชใƒ†ใ‚ฃๆˆฆ็•ฅใจใฏใ€ๆฅใ‚‹ในใ้‡ๅญใ‚ณใƒณใƒ”ใƒฅใƒผใ‚ฟใฎ่„…ๅจใซๆ€ฏใˆใ‚‹ใ“ใจใงใฏใชใใ€่‡ช็คพใŒๅฎˆใ‚‹ในใๆƒ…ๅ ฑใฎไพกๅ€คใจๆ™‚้–“่ปธใ‚’ๅ†ๅฎš็พฉใ—ใ€ใใ‚Œใ‚’ๅฎˆใ‚‹ใŸใ‚ใฎๆŠ€่ก“ๅŸบ็›คใ‚’ใ€Œๆ›ดๆ–ฐๅฏ่ƒฝใช็Šถๆ…‹ใ€ใซไฟใก็ถšใ‘ใ‚‹ใƒ—ใƒญใ‚ปใ‚นใใฎใ‚‚ใฎใงใ‚ใ‚‹ใจ่จ€ใˆใ‚‹ใ€‚้‡ๅญๆŠ€่ก“ใฎ้€ฒๅฑ•้€Ÿๅบฆใฏใ€ไธ€ไผๆฅญใฎใ‚ณใƒณใƒˆใƒญใƒผใƒซใ‚’่ถ…ใˆใŸๅค–้ƒจ่ฆๅ› ใงใ‚ใ‚‹ใ€‚ใ—ใ‹ใ—ใ€่‡ช็คพใฎใ‚ทใ‚นใƒ†ใƒ ใŒๅคใ„ๆš—ๅทๆŠ€่ก“ใจๅฟƒไธญใ™ใ‚‹ใฎใ‹ใ€ใใ‚Œใจใ‚‚ๆ–ฐใ—ใ„ๆŠ€่ก“ใ‚’ๆŸ”่ปŸใซๅ—ใ‘ๅ…ฅใ‚Œใ‚‰ใ‚Œใ‚‹ไฝ“่ณชใซๅค‰ใ‚ใ‚‹ใฎใ‹ใฏใ€็ตŒๅ–ถใฎๆ„ๆ€ๆฑบๅฎšใซใ‹ใ‹ใฃใฆใ„ใ‚‹ใ€‚PQCๆจ™ๆบ–ๅŒ–ใจใ„ใ†ๆณขใฏใ€ๆš—ๅทใจใ„ใ†ใ€Œ่ฆ‹ใˆใชใ„ใ‚คใƒณใƒ•ใƒฉใ€ใ‚’็ตŒๅ–ถใ‚ขใ‚ธใ‚งใƒณใƒ€ใธใจๆŠผใ—ไธŠใ’ใ€้™็š„ใ ใฃใŸใ‚ปใ‚ญใƒฅใƒชใƒ†ใ‚ฃ้‹็”จใ‚’ๅ‹•็š„ใงใ—ใชใ‚„ใ‹ใชใ‚‚ใฎใธใจ้€ฒๅŒ–ใ•ใ›ใ‚‹ใŸใ‚ใฎใ€ๅผทๅŠ›ใช่งฆๅช’ใจใ—ใฆๆฉŸ่ƒฝใ—ใฆใ„ใ‚‹ใฎใงใ‚ใ‚‹ใ€‚

6 strategies for CIOs to effectively manage shadow AI

28 November 2025 at 05:00

As employees experiment with gen AI tools on their own, CIOs are facing a familiar challenge with shadow AI. Although itโ€™s often well-intentioned innovation, it can create serious risks around data privacy, compliance, and security.

According to 1Passwordโ€™s 2025 annual report, The Access-Trust Gap, shadow AI increases an organizationโ€™s risk as 43% of employees use AI apps to do work on personal devices, while 25% use unapproved AI apps at work.

Despite these risks, experts say shadow AI isnโ€™t something to do away with completely. Rather, itโ€™s something to understand, guide, and manage. Here are six strategies that can help CIOs encourage responsible experimentation while keeping sensitive data safe.

1. Establish clear guardrails with room to experiment

Managing shadow AI begins with getting clear on whatโ€™s allowed and what isnโ€™t. Danny Fisher, chief technology officer at West Shore Home, recommends that CIOs classify AI tools into three simple categories:ย approved, restricted, and forbidden.

โ€œApproved tools are vetted and supported,โ€ he says. โ€œRestricted tools can be used in a controlled space with clear limits, like only using dummy data. Forbidden tools, which are typically public or unencrypted AI systems, should be blocked at the network or API level.โ€

Matching each type of AI use with a safe testing space, such as an internal OpenAI workspace or a secure API proxy, lets teams experiment freely without risking company data, he adds.

Jason Taylor, principal enterprise architect at LeanIX, an SAP company, says clear rules are essential in todayโ€™s fast-moving AI world.

โ€œBe clear which tools and platforms are approved and which ones arenโ€™t,โ€ he says. โ€œAlso be clear which scenarios and use cases are approved versus not, and how employees are allowed to work with company data and information when using AI like, for example, one-time upload as opposed to cut-and-paste or deeper integration.โ€

Taylor adds that companies should also create a clear list that explains which types of data are or arenโ€™t safe to use, and in what situations. A modern data loss prevention tool can help by automatically finding and labeling data, and enforcing least-privilege or zero-trust rules on who can access what.

Patty Patria, CIO at Babson College, notes itโ€™s also important for CIOs to establish specific guardrails for no-code/low-code AI tools and vibe-coding platforms.

โ€œThese tools empower employees to quickly prototype ideas and experiment with AI-driven solutions, but they also introduce unique risks when connecting to proprietary or sensitive data,โ€ she says.

To deal with this, Patria says companies should set up security layers that let people experiment safely on their own but require extra review and approval whenever someone wants to connect an AI tool to sensitive systems.

โ€œFor example, weโ€™ve recently developed clear internal guidance for employees outlining when to involve the security team for application review and when these tools can be used autonomously, ensuring both innovation and data protection are prioritized,โ€ she says. โ€œWe also maintain a list of AI tools we support, and which we donโ€™t recommend if theyโ€™re too risky.โ€

2. Maintain continuous visibility and inventory tracking

CIOs canโ€™t manage what they canโ€™t see. Experts say maintaining an accurate, up-to-date inventory of AI tools is one of the most important defenses against shadow AI.

โ€œThe most important thing is creating a culture where employees feel comfortable sharing what they use rather than hiding it,โ€ says Fisher. His team combines quarterly surveys with a self-service registry where employees log the AI tools they use. IT then validates those entries through network scans and API monitoring.

Ari Harrison, VP of IT at branding manufacturer Bamko, says his team takes a layered approach to maintaining visibility.

โ€œWe maintain a living registry of connected applications by pulling from Google Workspaceโ€™s connected-apps view and piping those events into our SIEM [security information and event management system],โ€ he says. โ€œMicrosoft 365 offers similar telemetry, and cloud access security broker tools can supplement visibility where needed.โ€

That layered approach gives Bamko a clear map of which AI tools are touching corporate data, who authorized them, and what permissions they have.

Mani Gill, SVP of product at cloud-based iPaaS Boomi, argues that manual audits are no longer enough.

โ€œEffective inventory management requires moving beyond periodic audits to continuous, automated visibility across the entire data ecosystem,โ€ he says, adding that good governance policies ensure all AI agents, whether approved or built into other tools, send their data in and out through one central platform. This gives organizations instant, real-time visibility into what each agent is doing, how much data itโ€™s using, and whether itโ€™s following the rules.

Tanium chief security advisor Tim Morris agrees that continuous discovery across every device and application is key. โ€œAI tools can pop up overnight,โ€ he says. โ€œIf a new AI app or browser plugin appears in your environment, you should know about it immediately.โ€

3. Strengthen data protection and access controls

When it comes to securing data from shadow AI exposure, experts point to the same foundation:ย data loss prevention (DLP), encryption, and least privilege.

โ€œUse DLP rules to block uploads of personal information, contracts, or source code to unapproved domains,โ€ Fisher says. He also recommends masking sensitive data before it leaves the organization, and turning on logging and audit trails to track every prompt and response in approved AI tools.

Harrison echoes that approach, noting that Bamko focuses on the security controls that matter most in practice: Outbound DLP and content inspection to prevent sensitive data from leaving; OAuth governance to keep third-party permissions to least privilege; and access limits that restrict uploads of confidential data to only approved AI connectors within its productivity suite.

In addition, the company treats broad permissions, such as read and write access to documents or email, as high-risk and requires explicit approval, while narrow, read-only permissions can move faster, Harrison adds.

โ€œThe goal is to allow safe day-to-day creativity while reducing the chance of a single click granting an AI tool more power than intended,โ€ he says.

Taylor adds that security must be consistent across environments. โ€œEncrypt all sensitive data at rest, in use, and in motion, employ least-privilege and zero-trust policies for data access permissions, and ensure DLP systems can scan for, tag, and protect sensitive data.โ€

He notes that companies should ensure these controls work the same on desktop, mobile, and web, and keep checking and updating them as new situations come up.

4. Clearly define and communicate risk tolerance

Defining risk tolerance is as much about communication as it is about control. Fisher advises CIOs to tie risk tolerance to data classification instead of opinion. His team uses a simple color-coded system: green for low-risk activities, such as marketing content; yellow for internal documents that must use approved tools; and red for customer or financial data that canโ€™t be used with AI systems.

โ€œRisk tolerance should be grounded in business value and regulatory obligation,โ€ says Morris. Like Fisher, Morris recommends classifying AI use into clear categories, whatโ€™s permitted, what needs approval, and whatโ€™s prohibited, and communicating that framework through leadership briefings, onboarding, and internal portals.

Patria says Babsonโ€™s AI Governance Committee plays a key role in this process. โ€œWhen potential risks emerge, we bring them to the committee for discussion and collaboratively develop mitigation strategies,โ€ she says. โ€œIn some cases, weโ€™ve decided to block tools for staff but permit them for classroom use. That balance helps manage risk without stifling innovation.โ€

5. Foster transparency and a culture of trust

Transparency is the key to managing shadow AI well. Employees need to know whatโ€™s being monitored and why.

โ€œTransparency means employees always know whatโ€™s allowed, whatโ€™s being monitored, and why,โ€ Fisher says. โ€œPublish your governance approach on the company intranet and include real examples of both good and risky AI use. Itโ€™s not about catching people. Youโ€™re building confidence that utilizing AI is safe and fair.โ€

Taylor recommends publishing a list of officially sanctioned AI offerings and keeping it updated. โ€œBe clear about the roadmap for delivering capabilities that arenโ€™t yet available,โ€ he says, and provide a process to request exceptions or new tools. That openness shows governance exists to support innovation, not hinder it.

Patria says in addition to technical controls and clear policies, establishing dedicated governance groups, like the AI Governance Committee, can greatly enhance an organizationโ€™s ability to manage shadow AI risks.

โ€œWhen potential risks emerge, such as concerns about tools like DeepSeek and Fireflies.AI, we collaboratively develop mitigation strategies,โ€ she says.

This governance group not only looks at and handles risks, but explains its decisions and the reasons behind them, helping create transparency and shared responsibility, Patria adds.

Morris agrees. โ€œTransparency means there are no surprises. Employees should know which AI tools are approved, how decisions are made, and where to go with questions or new ideas,โ€ he says.

6. Build continuous, role-based AI training

Training is one of the most effective ways to prevent accidental misuse of AI tools. The key is be succinct, relevant, and recurring.

โ€œKeep training short, visual, and role-specific,โ€ says Fisher. โ€œAvoid long slide decks and use stories, quick demos, and clear examples instead.โ€

Patria says Babson integrates AI risk awareness into annual information security training, and sends periodic newsletters about new tools and emerging risks.

โ€œRoutine training sessions are offered to ensure employees understand approved AI tools and emerging risks, while departmental AI champions are encouraged to facilitate dialogue and share practical experiences, highlighting both the benefits and potential pitfalls of AI adoption,โ€ she adds.

Taylor recommends embedding training in-browser, so employees learn best practices directly in the tools theyโ€™re using. โ€œCutting and pasting into a web browser or dragging and dropping a presentation seems innocuous until your sensitive data has left your ecosystem,โ€ he says.

Gill notes that training should connect responsible use with performance outcomes.

โ€œEmployees need to understand that compliance and productivity work together,โ€ he says. โ€œApproved tools deliver faster results, better data accuracy, and fewer security incidents compared with shadow AI. Role-based, ongoing training can demonstrate how guardrails and governance protect both data and efficiency, ensuring that AI accelerates workflows rather than creating risk.โ€

Responsible AI use is good business

Ultimately, managing shadow AI isnโ€™t just about reducing risk, itโ€™s about supporting responsible innovation. CIOs who focus on trust, communication, and transparency can turn a potential problem into a competitive advantage.

โ€œPeople generally donโ€™t try and buck the system when the system is giving them what theyโ€™re looking for, especially when thereโ€™s more friction for the user in taking the shadow AI approach,โ€ says Taylor.

Morris concurs. โ€œThe goal isnโ€™t to scare people but to make them think before they act,โ€ he says. โ€œIf they know the approved path is easy and safe, theyโ€™ll take it.โ€

Thatโ€™s the future CIOs should work toward: a place where people can innovate safely, feel trusted to experiment, and keep data protected because responsible AI use isnโ€™t just compliance, itโ€™s good business.

2026๋…„์— ์ฃผ๋ชฉํ•ด์•ผ ํ•  10๋Œ€ IT ๊ธฐ์ˆ  ์—ญ๋Ÿ‰

27 November 2025 at 21:55

์ƒ์„ฑํ˜• AI๊ฐ€ ๊ธฐ์—…์˜ AI ์ „๋žต ์žฌํŽธ์„ ์ด๋Œ๋ฉด์„œ IT ๊ธฐ์ˆ  ์ธ๋ ฅ ์‹œ์žฅ๋„ ์žฌํŽธ๋˜๊ณ  ์žˆ๋‹ค. ๊ธฐ์—…์€ AI ์—ญ๋Ÿ‰์„ ๊ฐ–์ถ˜ ์ง€์›์ž์™€ ์žฌ์ง์ž์—๊ฒŒ ์šฐ์„ ์ˆœ์œ„๋ฅผ ๋‘๊ณ  ์žˆ๋‹ค. ์ธ๋””๋“œ์˜ โ€˜2025 ํ…Œํฌ ์ธ์žฌ ๋ณด๊ณ ์„œ(Tech Talent Report)โ€™์— ๋”ฐ๋ฅด๋ฉด AI ๊ด€๋ จ ์กฐ์ง ์žฌํŽธ์˜ ์˜ํ–ฅ์„ ๊ฐ€์žฅ ํฌ๊ฒŒ ๋ฐ›์€ ์ƒ์œ„ 4๊ฐœ ์—ญํ• ์€ ์†Œํ”„ํŠธ์›จ์–ด ์—”์ง€๋‹ˆ์–ดยท๊ฐœ๋ฐœ์ž, QA ์—”์ง€๋‹ˆ์–ด, ํ”„๋กœ๋•ํŠธ ๋งค๋‹ˆ์ €, ํ”„๋กœ์ ํŠธ ๋งค๋‹ˆ์ €์˜€๋‹ค. ํ˜„์žฌ ๊ธฐ์—…์€ ์‚ฌ์ด๋ฒ„๋ณด์•ˆ, ๋ฐ์ดํ„ฐ ๋ถ„์„ยท๋ฐ์ดํ„ฐ ์• ๋„๋ฆฌํ‹ฑ์Šค, AIํŒ€ ๊ตฌ์ถ•ยท๊ด€๋ฆฌ ์—ญ๋Ÿ‰์„ ๋ณด์œ ํ•œ ์ „๋ฌธ๊ฐ€์— ์˜ˆ์‚ฐ๊ณผ ๋…ธ๋ ฅ์„ ์ง‘์ค‘ํ•˜๊ณ  ์žˆ๋‹ค.

IT ์—ญํ• ์˜ ์šฐ์„ ์ˆœ์œ„๊ฐ€ ๋ฐ”๋€Œ๋ฉด์„œ ๊ตฌ์ง์ž๊ฐ€ ์ด๋ ฅ์„œ์— ๋‹ด์•„์•ผ ํ•  IT ๊ธฐ์ˆ  ์—ญ๋Ÿ‰๋„ ๋‹ฌ๋ผ์กŒ๋‹ค. ๊ธฐ์—…์€ ์ด์ œ ์ดˆ๊ธ‰ IT ์ง๋ฌด๋ผ๋„ ์ตœ์†Œํ•œ ๊ธฐ๋ณธ์ ์ธ ํ”„๋กฌํ”„ํŠธ ์—”์ง€๋‹ˆ์–ด๋ง ์—ญ๋Ÿ‰์„ ๊ฐ–์ถ”๊ธฐ๋ฅผ ๊ธฐ๋Œ€ํ•œ๋‹ค. ๊ทธ๋ณด๋‹ค ๋†’์€ ์ˆ˜์ค€์—์„œ๋Š” AI ๋„๊ตฌ์™€ ์ „๋žต์„ ๊ฐ๋…ํ•˜๊ณ , ๋„์ž…ํ•˜๊ณ , ๋ณด์•ˆ ์ˆ˜์ค€์„ ํ™•๋ณดํ•˜๊ณ , ์šด์˜๊นŒ์ง€ ์ฑ…์ž„์งˆ ์ˆ˜ ์žˆ๋Š” IT ์ „๋ฌธ๊ฐ€๋ฅผ ์ฐพ๊ณ  ์žˆ๋‹ค.

์ธ๋””๋“œ ๋ฐ์ดํ„ฐ์— ๋”ฐ๋ฅด๋ฉด 2024๋…„๊ณผ 2025๋…„ ์‚ฌ์ด ์ฑ„์šฉ ๊ณต๊ณ ์— ์š”๊ตฌ ์กฐ๊ฑด์œผ๋กœ ํฌํ•จ๋œ ํšŸ์ˆ˜๋ฅผ ๊ธฐ์ค€์œผ๋กœ ๋‹ค์Œ 10๊ฐ€์ง€ IT ๊ธฐ์ˆ  ์—ญ๋Ÿ‰์˜ ์„ ํ˜ธ๋„๊ฐ€ ๊ฐ€์žฅ ํฌ๊ฒŒ ๋†’์•„์กŒ๋‹ค.

AI

AI ์—ญ๋Ÿ‰ ์ˆ˜์š”๊ฐ€ ๊ฐ€์žฅ ํฐ ํญ์œผ๋กœ ๋Š˜์–ด๋‚œ ๊ฒƒ์€ ์ „ํ˜€ ๋†€๋ผ์šด ์ผ์ด ์•„๋‹ˆ๋‹ค. ์—ฌ๋Ÿฌ ์‚ฐ์—…๊ณผ ์ง๋ฌด์—์„œ AI ๋„์ž… ๊ฒฝ์Ÿ์ด ๋ฒŒ์–ด์ง€๋ฉด์„œ ๊ธฐ์—…์€ AI๋ฅผ ์–ด๋–ป๊ฒŒ๋“  ํ™œ์šฉํ•˜๋ ค ๋ถ„์ฃผํ•˜๊ฒŒ ์›€์ง์ด๊ณ  ์žˆ๋‹ค. 2024๋…„์—๋Š” AI ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๊ฐ€ 500๋งŒ ๊ฑด์„ ์›ƒ๋Œ์•˜๊ณ , 2025๋…„์—๋Š” 1๋…„ ์ƒˆ 400๋งŒ ๊ฑด ์ด์ƒ ๋Š˜์–ด๋‚ฌ๋‹ค. ์ด์ œ ๊ธฐ์ˆ  ๋ถ„์•ผ ๋ฐ–์—์„œ ์ผํ•˜๋Š” ์ง€์›์ž๋ผ๋„ ํ”„๋กฌํ”„ํŠธ ์—”์ง€๋‹ˆ์–ด๋ง, ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ, ํ”„๋กœ๊ทธ๋ž˜๋ฐยท์ฝ”๋”ฉ์šฉ AI ํ™œ์šฉ ๋“ฑ ์ผ์ • ์ˆ˜์ค€์˜ AI ์—ญ๋Ÿ‰์„ ๊ฐ–์ถฐ์•ผ ํ•˜๋Š” ์ƒํ™ฉ์ด๋‹ค.

ํŒŒ์ด์ฌ

ํŒŒ์ด์ฌ์€ ๋ฐ์ดํ„ฐ ๋ถ„์„, ์›น ๊ฐœ๋ฐœ, ์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ, ๊ณผํ•™ ์ปดํ“จํŒ…, AIยท๋จธ์‹ ๋Ÿฌ๋‹(ML) ๋ชจ๋ธ ๊ตฌ์ถ• ๋“ฑ ์—ฌ๋Ÿฌ ๋ถ„์•ผ์—์„œ ํ™œ์šฉ๋˜๋Š” ํ”„๋กœ๊ทธ๋ž˜๋ฐ ์–ธ์–ด์ด๋‹ค. ์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ์ž, ์›น ๊ฐœ๋ฐœ์ž, ๋ฐ์ดํ„ฐ ์‚ฌ์ด์–ธํ‹ฐ์ŠคํŠธ, ๋ฐ์ดํ„ฐ ์• ๋„๋ฆฌ์ŠคํŠธ, ML ์—”์ง€๋‹ˆ์–ด, ์‚ฌ์ด๋ฒ„๋ณด์•ˆ ๋ถ„์„๊ฐ€, ํด๋ผ์šฐ๋“œ ์—”์ง€๋‹ˆ์–ด ๋“ฑ ๋‹ค์–‘ํ•œ IT ์ง๋ฌด๊ฐ€ ํญ๋„“๊ฒŒ ์‚ฌ์šฉํ•˜๋Š” ๋‹ค๋ชฉ์  ์–ธ์–ด์ด๋‹ค. ๊ธฐ์—… ํ™˜๊ฒฝ์—์„œ์˜ ํ™œ์šฉ ํญ์ด ๋„“๊ธฐ ๋•Œ๋ฌธ์— ๊พธ์ค€ํžˆ ์ˆ˜์š” ์ƒ์œ„๊ถŒ์„ ์œ ์ง€ํ•˜๋Š” ๊ธฐ์ˆ  ์—ญ๋Ÿ‰์ด๋‹ค. 2024๋…„์—๋Š” ํŒŒ์ด์ฌ ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๊ฐ€ 1,500๋งŒ ๊ฑด์„ ์กฐ๊ธˆ ๋„˜์—ˆ๊ณ , 2025๋…„์—๋Š” 1,800๋งŒ ๊ฑด์— ์•ฝ๊ฐ„ ๋ชป ๋ฏธ์น˜๋Š” ์ˆ˜์ค€๊นŒ์ง€ ๋Š˜์–ด๋‚ฌ๋‹ค. ๋” ๋งŽ์€ ์กฐ์ง์ด ์ฝ”๋”ฉ์— AI๋ฅผ ํ™œ์šฉํ•˜๊ณ  ์žˆ์ง€๋งŒ, ์—ฌ์ „ํžˆ ๋ณต์žกํ•œ ์ฝ”๋“œ๋ฅผ ์ง์ ‘ ์ž‘์„ฑํ•˜๊ณ , AI๊ฐ€ ์ƒ์„ฑํ•œ ์ฝ”๋“œ์˜ ํ”„๋กฌํ”„ํŠธ๋ฅผ ๋‹ค๋“ฌ๊ณ  ํ’ˆ์งˆ์„ ๊ฒ€์ฆํ•  ์ˆ˜ ์žˆ๋Š” ์ˆ™๋ จ ๊ฐœ๋ฐœ์ž ์—ญ๋Ÿ‰์ด ํ•„์š”ํ•˜๋‹ค.

์•Œ๊ณ ๋ฆฌ์ฆ˜

๋งŽ์€ ๊ธฐ์—…์ด ์ฝ”๋”ฉ๊ณผ ํ”„๋กœ๊ทธ๋ž˜๋ฐ ํšจ์œจํ™”๋ฅผ ์œ„ํ•ด AI๋ฅผ ๋„์ž…ํ•˜๋ฉด์„œ, ์ด๋Ÿฐ ๊ณผ์ •์„ ์ด๋Œ๊ณ  ์ œ์–ดํ•˜๋Š” ๊ธฐ์ค€์ด ๋˜๋Š” ์•Œ๊ณ ๋ฆฌ์ฆ˜์— ๋Œ€ํ•œ ์˜์กด๋„๋„ ๋†’์•„์กŒ๋‹ค. ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์ค‘์‹ฌ์˜ ์‚ฌ๊ณ ์—๋Š” ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค์™€ ํ”„๋กœ๊ทธ๋ž˜๋ฐ์— ๋Œ€ํ•œ ๊นŠ์€ ์ดํ•ด, ๊ณ ์ฐจ์› ๋น„ํŒ์  ์‚ฌ๊ณ , ๋ฌธ์ œ ํ•ด๊ฒฐ ๋Šฅ๋ ฅ์ด ํ•„์š”ํ•˜๋‹ค. ์•Œ๊ณ ๋ฆฌ์ฆ˜ ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๋Š” 2024๋…„์—๋Š” ์•ฝ 18๋งŒ ๊ฑด์— ๋ถˆ๊ณผํ–ˆ์ง€๋งŒ, 2025๋…„์—๋Š” 200๋งŒ ๊ฑด์ด ๋„˜์—ˆ๋‹ค. AI๊ฐ€ ์ดˆ๊ธ‰ ์—…๋ฌด ์ƒ๋‹น ๋ถ€๋ถ„์„ ๋„˜๊ฒจ๋ฐ›์œผ๋ฉด์„œ, ๊ธฐ์—…์€ AI ์‹œ์Šคํ…œ์„ ์„ค๊ณ„ยท์ง€๋„ํ•˜๊ณ  ํšจ์œจ์ ์ธ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๋Š” ๊ณ ๊ธ‰ ์ธ๋ ฅ์„ ์ฐพ๊ณ  ์žˆ๋‹ค.

CI/CD

CI/CD ์—ญ๋Ÿ‰์€ AI ๋„์ž… ์ดํ›„ ์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ ๋ผ์ดํ”„์‚ฌ์ดํด์„ ํšจ์œจํ™”ํ•˜๋Š” ๋ฐ ํ•„์ˆ˜ ๊ธฐ์ˆ ๋กœ ๋– ์˜ฌ๋ž๋‹ค. CI/CD ์—ญ๋Ÿ‰์„ ๊ฐ–์ถ˜ ์ „๋ฌธ๊ฐ€๋Š” ์ž๋™ํ™”ยท์Šคํฌ๋ฆฝํŒ… ๋„๊ตฌ๋ฅผ ๊ตฌ์ถ•ํ•˜๊ณ , ์ปจํ…Œ์ด๋„ˆํ™”, ํด๋ผ์šฐ๋“œ ํ†ตํ•ฉ, ์ž๋™ํ™” ํ…Œ์ŠคํŠธ ๊ฐ™์€ ๊ฐœ๋…์„ ์ดํ•ดํ•˜๊ณ  ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋‹ค. 2024๋…„์—๋Š” CI/CD ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๊ฐ€ 700๋งŒ ๊ฑด์— ์กฐ๊ธˆ ๋ชป ๋ฏธ์ณค๊ณ , 2025๋…„์—๋Š” 900๋งŒ ๊ฑด์„ ๋„˜๋Š” ์ˆ˜์ค€๊นŒ์ง€ ์ฆ๊ฐ€ํ–ˆ๋‹ค.

๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ

๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ๋Š” ๊ธฐ์—…์˜ IT ์†”๋ฃจ์…˜์„ ๊ตฌ์ถ•ยท๋ฐฐํฌยท๊ด€๋ฆฌํ•˜๊ธฐ ์œ„ํ•ด ๋„๋ฆฌ ์‚ฌ์šฉํ•˜๋Š” ํด๋ผ์šฐ๋“œ ํ”Œ๋žซํผ์œผ๋กœ, ๊ตฌ๊ธ€์€ ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ ์—ญ๋Ÿ‰๊ณผ ์ง€์‹์„ ์ธ์ฆํ•˜๋Š” ์—ฌ๋Ÿฌ ์ž๊ฒฉ์ฆ์„ ์ œ๊ณตํ•œ๋‹ค. ์ตœ๊ทผ ๋ช‡ ๋…„ ๋™์•ˆ ๋งŽ์€ ์กฐ์ง์ด ์—…๋ฌด ๋„๊ตฌ, ์„œ๋น„์Šค, ๋ฐ์ดํ„ฐ ์ €์žฅ์†Œ๋ฅผ ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ ๊ธฐ๋ฐ˜ ์„œ๋น„์Šค๋กœ ์ด์ „ํ•ด ์™”๋‹ค. ํŠนํžˆ ๋Œ€๊ทœ๋ชจ ๋ฐ์ดํ„ฐ๋ฅผ ์ €์žฅํ•˜๊ณ  ์ฒ˜๋ฆฌํ•ด์•ผ ํ•˜๋Š” AI ๊ฐœ๋ฐœ ํ™˜๊ฒฝ์—์„œ ํด๋ผ์šฐ๋“œ ๋„๊ตฌ๋Š” ํ•„์ˆ˜ ์ธํ”„๋ผ๋กœ ์ž๋ฆฌ ์žก์•˜๋‹ค. 2024๋…„์—๋Š” ๊ตฌ๊ธ€ ํด๋ผ์šฐ๋“œ ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๊ฐ€ ์•ฝ 350๋งŒ ๊ฑด์ด์—ˆ์ง€๋งŒ, 2025๋…„์—๋Š” 530๋งŒ ๊ฑด์„ ์›ƒ๋„๋Š” ์ˆ˜์ค€๊นŒ์ง€ ๋Š˜์–ด๋‚ฌ๋‹ค.

AWS

AWS๋Š” ํ˜„์žฌ ๊ฐ€์žฅ ๋„๋ฆฌ ์“ฐ์ด๋Š” ํด๋ผ์šฐ๋“œ ํ”Œ๋žซํผ์ด๋‹ค. ์—ฌ๋Ÿฌ ์‚ฐ์—…์—์„œ ํด๋ผ์šฐ๋“œ ์ „๋žต์˜ ํ•ต์‹ฌ ์ถ•์„ ๋‹ด๋‹นํ•˜๋Š” ๋งŒํผ, ํ”Œ๋žซํผ์ด ์ œ๊ณตํ•˜๋Š” ๋ฐฉ๋Œ€ํ•œ ์„œ๋น„์Šค๋ฅผ ์ œ๋Œ€๋กœ ํ™œ์šฉํ•˜๊ธฐ ์œ„ํ•œ AWS ์—ญ๋Ÿ‰ ์ˆ˜์š”๋„ ํฌ๋‹ค. ํด๋ผ์šฐ๋“œ ์—”์ง€๋‹ˆ์–ด, ๋ฐ๋ธŒ์˜ต์Šค ์—”์ง€๋‹ˆ์–ด, ์†”๋ฃจ์…˜ ์•„ํ‚คํ…ํŠธ, ๋ฐ์ดํ„ฐ ์—”์ง€๋‹ˆ์–ด, ์‚ฌ์ด๋ฒ„๋ณด์•ˆ ๋ถ„์„๊ฐ€, ์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ์ž, ๋„คํŠธ์›Œํฌ ๊ด€๋ฆฌ์ž ๋“ฑ ์ˆ˜๋งŽ์€ IT ์ง๋ฌด์—์„œ ๊ณตํ†ต์œผ๋กœ ์š”๊ตฌ๋˜๋Š” ์—ญ๋Ÿ‰์ด๋‹ค. 2024๋…„์—๋„ AWS ์—ญ๋Ÿ‰์€ ๋†’์€ ์ธ๊ธฐ๋ฅผ ์œ ์ง€ํ•˜๋ฉฐ 1,200๋งŒ ๊ฑด์ด ์กฐ๊ธˆ ๋„˜๋Š” ์ฑ„์šฉ ๊ณต๊ณ ์— ์š”๊ตฌ์‚ฌํ•ญ์œผ๋กœ ํฌํ•จ๋๊ณ , 2025๋…„์—๋Š” 1,370๋งŒ ๊ฑด์„ ๋„˜์—ˆ๋‹ค.

๋ถ„์„ ์—ญ๋Ÿ‰

AI๊ฐ€ ์ดˆ๊ธ‰ยท๋ฐ˜๋ณต ์—…๋ฌด ์ƒ๋‹น ๋ถ€๋ถ„์„ ๋Œ€์ฒดํ•˜๋ฉด์„œ IT ์ „๋ฌธ๊ฐ€์—๊ฒŒ๋Š” ๋” ๋†’์€ ์ˆ˜์ค€์˜ ๋ถ„์„์  ์‚ฌ๊ณ  ์—ญ๋Ÿ‰์ด ์š”๊ตฌ๋˜๊ณ  ์žˆ๋‹ค. AI๊ฐ€ ๋‚ด๋ฆฌ๋Š” ํŒ๋‹จ์ด๋‚˜ ์ƒ์„ฑ ๊ฒฐ๊ณผ๊ฐ€ ํ•ญ์ƒ ์™„๋ฒฝํ•˜์ง€ ์•Š๊ธฐ ๋•Œ๋ฌธ์—, ํŠนํžˆ ์ˆซ์ž์™€ ๋ฐ์ดํ„ฐ ์˜์—ญ์—์„œ๋Š” AI ํ™˜๊ฐ๊ณผ ์˜ค๋ฅ˜๋ฅผ ๊ฐ€๋ ค๋‚ผ ์ˆ˜ ์žˆ๋Š” ์ธ๊ฐ„์˜ ์‹œ์„ ๊ณผ ๋ถ„์„ ๋Šฅ๋ ฅ์ด ํ•„์š”ํ•˜๋‹ค. ๋ถ„์„ ์—ญ๋Ÿ‰์€ ์ด๋ฏธ ์˜ค๋ž˜์ „๋ถ€ํ„ฐ ์กฐ์ง์— ํ•ต์‹ฌ์œผ๋กœ ์ž๋ฆฌํ•œ ๋Šฅ๋ ฅ์ด๋‹ค. 2024๋…„์—๋Š” ๋ถ„์„ ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๊ฐ€ 1,900๋งŒ ๊ฑด์„ ์กฐ๊ธˆ ๋„˜์—ˆ๊ณ , 2025๋…„์—๋Š” 2,100๋งŒ ๊ฑด์„ ์›ƒ๋„๋Š” ์ˆ˜์ค€์œผ๋กœ ๋Š˜์–ด๋‚ฌ๋‹ค.

์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ

AI ์˜์กด๋„๊ฐ€ ๋†’์•„์ง€๋ฉด์„œ ๊ธฐ์—…์˜ ๋ณด์•ˆ ์ทจ์•ฝ ์ง€์ ๋„ ํ•จ๊ป˜ ๋Š˜์–ด๋‚˜๊ณ  ์žˆ๋‹ค. ๋” ๋งŽ์€ ์ œํ’ˆ๊ณผ ์„œ๋น„์Šค๋ฅผ ์˜จ๋ผ์ธ์œผ๋กœ ์ „ํ™˜ํ•˜๊ณ  AI๋ฅผ ํ†ตํ•ฉํ•˜๋Š” ๊ณผ์ •์—์„œ ๊ณต๊ฒฉ์ž๊ฐ€ ๋…ธ๋ฆด ์ˆ˜ ์žˆ๋Š” ์ง€์ ์ด ๋งŽ์•„์ง„๋‹ค๋Š” ์˜๋ฏธ์ด๋‹ค. ์‚ฌ์ด๋ฒ„๋ณด์•ˆ ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๋Š” 2024๋…„ ์•ฝ 240๋งŒ ๊ฑด์ด์—ˆ๋‹ค๊ฐ€ 2025๋…„์—๋Š” 400๋งŒ ๊ฑด์„ ๋„˜๋Š” ์ˆ˜์ค€๊นŒ์ง€ ์ฆ๊ฐ€ํ–ˆ๋‹ค. AI๋ฅผ ๋ณด์•ˆ ์†”๋ฃจ์…˜์— ํ†ตํ•ฉํ•˜๋“ , AI๋ฅผ ์•…์šฉํ•œ ์ƒˆ๋กœ์šด ๊ณ ๋„ํ™” ๊ณต๊ฒฉ์„ ๋ง‰๋“ , ๋ณด์•ˆ์€ AI ๋„์ž…์„ ์ถ”์ง„ํ•˜๋Š” ์กฐ์ง์ด ์ตœ์šฐ์„ ์œผ๋กœ ๊ณ ๋ คํ•˜๋Š” ๋ถ„์•ผ์ด๋‹ค.

์†Œํ”„ํŠธ์›จ์–ด ๋ฌธ์ œ ํ•ด๊ฒฐ ์—ญ๋Ÿ‰

๋” ๋งŽ์€ ์กฐ์ง์ด ๊ธฐ๋ณธ์ ์ธ ์ฝ”๋“œ์™€ ์Šคํฌ๋ฆฝํŠธ ์ž‘์„ฑ์—๋Š” AI๋ฅผ ํ™œ์šฉํ•˜๊ณ  ์žˆ์ง€๋งŒ, ์ตœ์ข… ๊ฒฐ๊ณผ๋ฌผ์—์„œ ๊ฒฐํ•จ, ๋ณด์•ˆ ๋ฌธ์ œ, ์ด์ƒ ์ง•ํ›„๋ฅผ ์ฐพ์•„๋‚ด๋Š” ์ผ์€ ์—ฌ์ „ํžˆ ์ธ๊ฐ„ IT ์ „๋ฌธ๊ฐ€์˜ ๋ชซ์ด๋‹ค. ์†Œํ”„ํŠธ์›จ์–ด ํŠธ๋Ÿฌ๋ธ”์ŠˆํŒ… ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๋Š” 2024๋…„ 900๋งŒ ๊ฑด์ด ์กฐ๊ธˆ ๋„˜์—ˆ๊ณ , 2025๋…„์—๋Š” 1,100๋งŒ ๊ฑด์— ์กฐ๊ธˆ ๋ชป ๋ฏธ์น˜๋Š” ์ˆ˜์ค€๊นŒ์ง€ ์ฆ๊ฐ€ํ–ˆ๋‹ค. ์ด ๋ถ„์•ผ์—๋Š” ์†Œํ”„ํŠธ์›จ์–ด ๋ฌธ์ œ๋ฅผ ํŒŒ์•…ํ•˜๊ณ  ๊ณ ๊ฐยท์‚ฌ์šฉ์ž์˜ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•œ ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜ ๋Šฅ๋ ฅ, ๋ฌธ์ œ ํ•ด๊ฒฐ ๋Šฅ๋ ฅ, ๋น„ํŒ์  ์‚ฌ๊ณ , ๊ธฐ์ˆ  ์—ญ๋Ÿ‰์ด ๋ชจ๋‘ ํ•„์š”ํ•˜๋‹ค.

๋จธ์‹ ๋Ÿฌ๋‹

๋จธ์‹ ๋Ÿฌ๋‹์€ AI ๊ฐœ๋ฐœ์˜ ํ•ต์‹ฌ ๊ธฐ์ˆ ๋กœ, AI๋ฟ๋งŒ ์•„๋‹ˆ๋ผ ์ž์—ฐ์–ด์ฒ˜๋ฆฌ์— ๋Œ€ํ•œ ๋†’์€ ์ˆ˜์ค€์˜ ์ „๋ฌธ์„ฑ์ด ์š”๊ตฌ๋œ๋‹ค. ๊ธฐ์—…์€ AI ๋„์ž…๊ณผ ํ–ฅํ›„ ํ™•์‚ฐ์„ ๋’ท๋ฐ›์นจํ•  ์ˆ˜ ์žˆ๋Š” ML ์—ญ๋Ÿ‰ ๋ณด์œ  ์ „๋ฌธ๊ฐ€๋ฅผ ํ™•๋ณดํ•˜๋Š” ๋ฐ ์ฃผ๋ ฅํ•˜๊ณ  ์žˆ๋‹ค. ๋จธ์‹ ๋Ÿฌ๋‹ ์—ญ๋Ÿ‰์„ ์š”๊ตฌํ•œ ์ฑ„์šฉ ๊ณต๊ณ ๋Š” 2024๋…„ ์•ฝ 370๋งŒ ๊ฑด์—์„œ 2025๋…„์—๋Š” 500๋งŒ ๊ฑด์„ ๋„˜์—ˆ๋‹ค. ๊ธฐ์—…์ด AI ํ”„๋กœ์„ธ์Šค๋ฅผ ์ ๊ทน ๋„์ž…ํ•˜๊ณ , AI ์‹œ์Šคํ…œ์„ ์ง€์›ยท์œ ์ง€ํ•  ์ธ์žฌ๋ฅผ ์ฐพ๋Š” ํ๋ฆ„์ด ์ด์–ด์ง€๋Š” ํ•œ ๋จธ์‹ ๋Ÿฌ๋‹ ์—ญ๋Ÿ‰์„ ๊ฐ–์ถ˜ IT ์ „๋ฌธ๊ฐ€๋Š” ๊ณ„์† ๋†’์€ ์ˆ˜์š”๋ฅผ ์œ ์ง€ํ•  ๊ฒƒ์œผ๋กœ ๋ณด์ธ๋‹ค.
dl-ciokorea@foundryco.com

์• ํ”Œ, ๊ณ ๊ฐ€ ํ—ค๋“œ์…‹์œผ๋กœ ํ™•์žฅํ˜„์‹ค ๊ธฐ์—… ์ˆ˜์š” ๋…ธ๋ฆฐ๋‹ค

27 November 2025 at 20:03

๊ตฌ๊ธ€์€ ์ž์‚ฌ์˜ ์ฒซ ์ฆ๊ฐ•ํ˜„์‹ค ์‹œ๋„์˜€๋˜ ๊ตฌ๊ธ€ ๊ธ€๋ž˜์Šค๋ฅผ ๊ณต์‹์ ์œผ๋กœ ์ค‘๋‹จํ–ˆ๋‹ค. ๋‹ค๋งŒ, ๋Œ€์ค‘์ ์œผ๋กœ๋Š” ์‹คํŒจํ•œ ์ œํ’ˆ์œผ๋กœ ์—ฌ๊ฒจ์กŒ์Œ์—๋„, ๋‘ ํ•ด ์ „๊นŒ์ง€๋งŒ ํ•ด๋„ ์—ฌ์ „ํžˆ ๊ธฐ์—…์šฉ ๋ชจ๋ธ๋กœ ํŒ๋งค๋˜๊ณ  ์žˆ์—ˆ๋‹ค.

๋งˆ์ดํฌ๋กœ์†Œํ”„ํŠธ๋„ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ ํ™€๋กœ๋ Œ์ฆˆ๋ฅผ ๊ธฐ์—… ์ค‘์‹ฌ์œผ๋กœ ๋ฐฉํ–ฅ ์ „ํ™˜ํ–ˆ๋‹ค๊ฐ€ ๊ฒฐ๊ตญ์€ ์™„์ „ํžˆ ์ทจ์†Œํ–ˆ๋‹ค.

๋ฐ˜๋ฉด, ์• ํ”Œ์€ ๊ณ ๊ฐ€์˜ ๋น„์ „ ํ”„๋กœ ํ—ค๋“œ์…‹์œผ๋กœ ์ฒ˜์Œ๋ถ€ํ„ฐ ๊ธฐ์—… ์‹œ์žฅ์„ ๊ณต๋žตํ•ด ์™”๋‹ค. ์ฒซ ๊ณต๊ฐœ ์ดํ›„ ์—ด๋ฆฐ WWDC์—์„œ๋Š” ๋น„์ „ ์šด์˜์ฒด์ œ ์ „์šฉ ์„ธ์…˜์„ ๋งˆ๋ จํ–ˆ๊ณ , ํ”Œ๋žซํผ์˜ ์ฒซ ๋ฒˆ์งธ ์ฃผ์š” ์—…๋ฐ์ดํŠธ์—์„œ๋Š” ๋ชจ๋ฐ”์ผ ๊ธฐ๊ธฐ ๊ด€๋ฆฌ ๊ธฐ๋Šฅ๋„ ๋„์ž…ํ–ˆ๋‹ค.

์ด๋Ÿฌํ•œ ์ „๋žต์€ ๋‹น์—ฐํ•œ ์ˆ˜์ˆœ์œผ๋กœ ๋ณผ ์ˆ˜ ์žˆ๋‹ค. ํ™•์žฅํ˜„์‹ค ๊ธฐ๊ธฐ๋Š” ๊ฐ€๊ฒฉ๊ณผ ํ™œ์šฉ ๋ฐฉ์‹ ๋ชจ๋‘์—์„œ ๊ธฐ์—… ์‹œ์žฅ์— ์ ํ•ฉํ•˜๋‹ค๊ณ  ์—ฌ๊ฒจ์ง€๊ณ  ์žˆ์œผ๋ฉฐ, ๊ฐ€์ƒํ˜„์‹ค ์‹œ์žฅ์€ ์„ฑ๊ณต์„ ์ขŒ์šฐํ•˜๋Š” ํ•ต์‹ฌ ์š”์†Œ๋กœ ์ธ์‹๋˜๊ณ  ์žˆ๋‹ค.

๋‹ค๋งŒ, 3,499๋‹ฌ๋Ÿฌ์— ๋‹ฌํ•˜๋Š” ์• ํ”Œ ๋น„์ „ ํ”„๋กœ๋Š” ๋ชจ๋“  ๊ธฐ์—…์— ์ ํ•ฉํ•˜์ง€ ์•Š์„ ์ˆ˜๋„ ์žˆ๋‹ค. ํ•„์š”ํ•˜๋”๋ผ๋„ ๋ชจ๋“  ์ง์›์„ ๋Œ€์ƒ์œผ๋กœ ํ•œ ์ œํ’ˆ์€ ์•„๋‹ˆ๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์ œํ’ˆ ์„ค๊ณ„, ์‹œ์ œํ’ˆ ์ œ์ž‘, ๋ฐ์ดํ„ฐ ์‹œ๊ฐํ™”, ๊ต์œก, ์‚ฐ์—… ํ˜„์žฅ ํ™œ์šฉ, ์˜๋ฃŒ ์‹œ์ˆ , ํŠน์ • ๊ณ ๊ฐ ์‘๋Œ€ ๋ถ„์•ผ ๋“ฑ์—์„œ ์‹ค์งˆ์ ์ธ ํšจ๊ณผ๋ฅผ ๋‚ผ ์ˆ˜ ์žˆ๋Š” ์˜์—ญ์€ ๋ถ„๋ช… ์กด์žฌํ•œ๋‹ค.

์ด๋Ÿฌํ•œ ๊ฐ€์น˜๋ฅผ ์‹คํ˜„ํ•˜๋ ค๋ฉด, ์ผ๋ฐ˜์ ์ธ ๊ธฐ์ˆ  ๋„์ž…๋ณด๋‹ค ๊ธฐ์—… ์ •๋ณด๊ธฐ์ˆ  ๋ถ€์„œ๊ฐ€ ๋” ์„ธ๋ฐ€ํ•˜๊ฒŒ ์ ‘๊ทผํ•ด์•ผ ํ•˜๋Š”๋ฐ, ๋ถˆ๊ฐ€๋Šฅํ•œ ๊ณผ์ œ๋Š” ์•„๋‹ˆ๋‹ค. ๋น„์ „ ํ”„๋กœ๋ฅผ ๊ธฐ์—…์šฉ ๊ธฐ๊ธฐ๋กœ ๊ณ ๋ คํ•  ๋•Œ์˜ ์ ‘๊ทผ ๋ฐฉ์‹๊ณผ ๋„์ž… ์ „ ๊ฒ€ํ† ํ•ด์•ผ ํ•  ์š”์†Œ๋ฅผ ์‚ดํŽด๋ณด์ž.

๋น„์ „ ํ”„๋กœ, ๋‹จ์ˆœ ์ „์‹œ์šฉ ๊ธฐ๊ธฐ ์•„๋ƒ

๊ฐ€์žฅ ๋จผ์ € ๋„˜์–ด์•ผ ํ•  ํฐ ๊ณผ์ œ๋Š” ๊ณต๊ฐ„ ์ปดํ“จํŒ… ๊ธฐ์ˆ ์„ ์—…๋ฌด์šฉ์œผ๋กœ ์ธ์‹ํ•˜๊ฒŒ ๋งŒ๋“œ๋Š” ๊ฒƒ์ด๋‹ค. ์—ฌ๊ธฐ์—๋Š” ์‚ฌ์šฉ์ž๊ฐ€ ์–ผ๊ตด์— ์žฅ์ฐฉํ•œ ์ปดํ“จํ„ฐ์— ์ต์ˆ™ํ•ด์ง€๋Š” ๊ฒƒ๋ฟ ์•„๋‹ˆ๋ผ, ์‚ฌ๋ฌด์‹ค์—์„œ ๋ˆ„๊ตฐ๊ฐ€ ํ•ด๋‹น ๊ธฐ๊ธฐ๋ฅผ ์ฐฉ์šฉํ•œ ๋ชจ์Šต์„ ์ž์—ฐ์Šค๋Ÿฝ๊ฒŒ ๋ฐ›์•„๋“ค์ด๋Š” ๊ฒƒ๊นŒ์ง€ ํฌํ•จ๋œ๋‹ค. ๋” ์ค‘์š”ํ•œ ๊ฒƒ์€ ๋น„์ „ ํ”„๋กœ๋ฅผ ๋‹จ์ˆœํ•œ ๊ณ ๊ฐ€ ์ „์ž ๊ธฐ๊ธฐ๊ฐ€ ์•„๋‹Œ ์—…๋ฌด ๋„๊ตฌ๋กœ ์ธ์‹ํ•˜๋Š” ๊ฒƒ์ด๋‹ค. ๊ทธ๋Ÿฌ๋ ค๋ฉด ๋น„์ „ ํ”„๋กœ๊ฐ€ ์‹ค์งˆ์ ์œผ๋กœ ํšŒ์‚ฌ์— ๊ธฐ์—ฌํ•˜๋Š” ๊ฐ€์น˜๋ฅผ ์ž…์ฆํ•ด์•ผ ํ•˜๋ฉฐ, ์‚ฌ์šฉ์ž๋„ ๊ทธ ๊ฐ€์น˜๋ฅผ ์ดํ•ดํ•ด์•ผ ํ•œ๋‹ค.

๋”ฐ๋ผ์„œ ๋น„์ „ ํ”„๋กœ๋ฅผ ์ •๋ณด๊ธฐ์ˆ  ๋„๊ตฌ๋กœ ๋„์ž…ํ•˜๋ ค๋ฉด, ์™œ ์ด ์žฅ๋น„๊ฐ€ ์ตœ์ ์˜ ํ•ด๊ฒฐ์ฑ…์ธ์ง€ ์‚ฌ์šฉ์ž์—๊ฒŒ ์ถฉ๋ถ„ํžˆ ์„ค๋ช…ํ•ด์•ผ ํ•œ๋‹ค.

๋ช…ํ™•ํ•œ ๋ชฉํ‘œ์™€ ๊ธฐ๋Œ€ ์„ฑ๊ณผ ์„ค์ •

๋น„์ „ ์šด์˜์ฒด์ œ ๊ธฐ๋ฐ˜ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜๊ณผ ์›Œํฌํ”Œ๋กœ๋ฅผ ๊ตฌ์ถ•ํ•  ๋•Œ๋Š” ๋ช…ํ™•ํ•œ ์ „๋žต์ด ํ•„์š”ํ•˜๋‹ค. ๋”ฐ๋ผ์„œ ๋น„์ „ ํ”„๋กœ๋ฅผ ์ง์›์šฉ์œผ๋กœ ๋„์ž…ํ•˜๋ ค๋Š” ๊ธฐ์—…์€ ๊ตฌ์ฒด์ ์ธ ๋ชฉํ‘œ์™€ ์„ฑ๊ณผ ์ธก์ • ๋ฐฉ์‹์„ ์‚ฌ์ „์— ์„ค์ •ํ•ด์•ผ ํ•œ๋‹ค.

์‚ฐ์—…๊ตฐ๊ณผ ๊ธฐ์—… ํŠน์„ฑ์— ๋”ฐ๋ผ ๋ฐฉ์‹์€ ๋‹ค์–‘ํ•˜๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด, ์ง€์—ญ์ด ๋‹ค๋ฅธ ๋””์ž์ธ ํŒ€์ด ๊ฐ€์ƒ ์›Œํฌ์ˆ์—์„œ ์‹ค์‹œ๊ฐ„์œผ๋กœ ํ˜‘์—…ํ•˜๋Š” ํ™˜๊ฒฝ์„ ๊ตฌํ˜„ํ•˜๊ฑฐ๋‚˜, ์›๊ฒฉ ๊ณต์žฅ ์žฅ๋น„ ์กฐ์ž‘ ๊ต์œก์„ ์ง„ํ–‰ํ•˜๊ฑฐ๋‚˜, ๋ฐ์ดํ„ฐ ๋ถ„์„์„ ์œ„ํ•œ ์‹œ๊ฐํ™” ํ™˜๊ฒฝ์„ ์ œ๊ณตํ•˜๊ฑฐ๋‚˜, ์˜๋ฃŒ ์ „๋ฌธ๊ฐ€๊ฐ€ ์›๊ฒฉ์œผ๋กœ ์ง„๋ฃŒ๋ฅผ ์ง€์›ํ•˜๋Š” ํ˜•ํƒœ ๋“ฑ์ด ์žˆ๋‹ค.

์–ด๋–ค ์‚ฌ๋ก€์— ํ•ด๋‹นํ•˜๋“ , ๊ธฐ์—…์€ ๊ธฐ์กด ์ปดํ“จํŒ… ๋„๊ตฌ๋ณด๋‹ค ๊ณต๊ฐ„ ์ปดํ“จํŒ…์ด ์–ด๋–ค ์ด์ ์„ ์ œ๊ณตํ•˜๋Š”์ง€๋ฅผ ๋ช…ํ™•ํžˆ ํ•ด์•ผ ํ•œ๋‹ค. ํŠนํžˆ ๊ณ ๊ฐ€์˜ ํ•˜๋“œ์›จ์–ด๊ฐ€ ์ˆ˜๋ฐ˜๋˜๋Š” ๋งŒํผ ๊ทธ ํ•„์š”์„ฑ์€ ๋” ๊ฐ•์กฐ๋˜์–ด์•ผ ํ•œ๋‹ค.

์‚ฌ์šฉ์ž์™€ ๊ธด๋ฐ€ํžˆ ํ˜‘๋ ฅํ•˜๋ผ

๋น„์ „ ํ”„๋กœ์˜ ์„ฑ๊ณต์ ์ธ ์ •์ฐฉ์€ ๊ธฐ๊ธฐ๋ฅผ ์ง์ ‘ ์‚ฌ์šฉํ•˜๋Š” ์ง์›, ํŒ€ ๋ฆฌ๋”, ๊ด€๋ฆฌ์ž์™€์˜ ํ˜‘๋ ฅ์— ๋‹ฌ๋ ค ์žˆ๋‹ค. ๊ธฐ์ˆ ์„ ๋จผ์ € ์š”๊ตฌํ•œ ์ชฝ์ด ์ง์›์ด๋“ , ์ •๋ณด์ฑ…์ž„์ž๋“  ์ƒ๊ด€์—†์ด, ๋ชจ๋“  ์ดํ•ด๊ด€๊ณ„์ž๊ฐ€ ์ž์‹ ์˜ ์—…๋ฌด๊ฐ€ ๋ฌด์—‡์ด๋ฉฐ ์ด ํ”Œ๋žซํผ์ด ์–ด๋–ป๊ฒŒ ์ด๋ฅผ ์ง€์›ํ•  ์ˆ˜ ์žˆ๋Š”์ง€๋ฅผ ํ•จ๊ป˜ ๋…ผ์˜ํ•ด์•ผ ํ•œ๋‹ค.

ํŠนํžˆ ๋„์ž… ์ดˆ๊ธฐ ๋‹จ๊ณ„์—๋Š” ์ •๊ธฐ์ ์ธ ์ ๊ฒ€ ์ผ์ •์„ ๋งˆ๋ จํ•˜๊ณ , ํ”ผ๋“œ๋ฐฑ ์ˆ˜์ง‘ ๋ฐฉ์‹(์†Œํ”„ํŠธ์›จ์–ด๋‚˜ ๋Œ€๋ฉด ๋ณด๊ณ  ํฌํ•จ)๋„ ๋ช…ํ™•ํžˆ ํ•ด์•ผ ํ•œ๋‹ค. ๋ฌด์—‡๋ณด๋‹ค ์ค‘์š”ํ•œ ๊ฒƒ์€ ์‚ฌ์šฉ์ž ํ”ผ๋“œ๋ฐฑ์„ ๋ฐ˜์˜ํ•ด ์‹ค์ œ ๊ฐœ์„  ์กฐ์น˜๋กœ ์—ฐ๊ฒฐํ•˜๋Š” ๊ฒƒ์ด๋‹ค.

๋ธŒ๋ผ์šฐ์ €๋„ ๋Œ€์•ˆ

์–ด๋–ค ๊ธฐ๊ธฐ๋“  ํ•ต์‹ฌ ๊ฐ€์น˜๋Š” ํ•˜๋“œ์›จ์–ด ์ž์ฒด๋ณด๋‹ค ์‹คํ–‰๋˜๋Š” ์†Œํ”„ํŠธ์›จ์–ด์—์„œ ๋‚˜์˜จ๋‹ค. ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์€ ๊ณต๊ฐœ ์•ฑ์ด ๋  ์ˆ˜๋„ ์žˆ๊ณ , ๊ธฐ์—… ์ „์šฉ ๋‚ด๋ถ€ ์•ฑ์ผ ์ˆ˜๋„ ์žˆ๋‹ค. ์ผ๋ฐ˜์ ์œผ๋กœ๋Š” ์ด ๋‘˜์„ ํ˜ผํ•ฉํ•ด ์‚ฌ์šฉํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค.

์ž์ฒด ๊ฐœ๋ฐœ ์•ฑ์€ ์‚ฌ์šฉ์ž์˜ ํ•„์š”๋‚˜ ์›Œํฌํ”Œ๋กœ์— ๋งž์ถฐ ์ •๋ฐ€ํ•˜๊ฒŒ ๊ตฌํ˜„ํ•  ์ˆ˜ ์žˆ๋Š” ์žฅ์ ์ด ์žˆ์ง€๋งŒ, ๋งŽ์€ ๊ฐœ๋ฐœ ์‹œ๊ฐ„๊ณผ ๋ฆฌ์†Œ์Šค๊ฐ€ ์†Œ์š”๋œ๋‹ค. ์ผ๋ถ€ ํŠน์ˆ˜ ๋ชฉ์ ์—๋Š” ์ด ๋ฐฉ๋ฒ•์ด ์œ ์ผํ•œ ์„ ํƒ์ผ ์ˆ˜ ์žˆ๋‹ค.

๊ทธ๋Ÿฌ๋‚˜ ๊ธฐ์กด ์†”๋ฃจ์…˜์ด ์žˆ๋Š”์ง€ ๋จผ์ € ํ™•์ธํ•˜๋Š” ๊ฒƒ์ด ๋” ๋น ๋ฅด๊ณ  ๋น„์šฉ๋„ ์ ˆ๊ฐ๋  ์ˆ˜ ์žˆ๋‹ค. ๊ธฐ์—…์šฉ ๋””๋ฐ”์ด์Šค์™€ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ, ์ •๋ณด๊ธฐ์ˆ  ๋ถ€์„œ๋Š” ํ˜„์žฌ ์‚ฌ์šฉ ์ค‘์ธ ๋ชจ๋ฐ”์ผ ๊ธฐ๊ธฐ ๊ด€๋ฆฌ ์‹œ์Šคํ…œ์„ ํ™œ์šฉํ•ด ํผ๋ธ”๋ฆญ ์•ฑ๊ณผ ๊ธฐ์—… ๋‚ด๋ถ€ ์•ฑ์„ ํ๋ ˆ์ด์…˜ํ•  ์ˆ˜ ์žˆ์–ด์•ผ ํ•œ๋‹ค.

์‚ฌ์šฉ์ž๋กœ๋ถ€ํ„ฐ ์•ฑ ๋Œ€์•ˆ ์ œ์•ˆ์ด ๋“ค์–ด์˜ค๋Š” ๊ฒฝ์šฐ๋„ ์žˆ์œผ๋ฏ€๋กœ, ๊ฐ ์ œ์•ˆ์ด ๋” ๋‚˜์€ ํ•ด๋ฒ•์ด ๋  ์ˆ˜ ์žˆ๋Š”์ง€ ๊ฒ€ํ† ํ•˜๋Š” ๊ณผ์ •์ด ๋ฐ˜๋“œ์‹œ ํ•„์š”ํ•˜๋‹ค.

๋˜ํ•œ, ์ „์ฒด ๊ธฐ๋Šฅ์„ ๊ฐ–์ถ˜ ์•ฑ์ด ๋ฐ˜๋“œ์‹œ ํ•„์š”ํ•œ ๊ฒƒ์€ ์•„๋‹ˆ๋‹ค. ๋น„์ „ ์šด์˜์ฒด์ œ์˜ ์‚ฌํŒŒ๋ฆฌ ์›น ๋ธŒ๋ผ์šฐ์ €๋Š” ๋ชฐ์ž…ํ˜• ์ฝ˜ํ…์ธ  ์ œ๊ณต์ด ๊ฐ€๋Šฅํ•˜๋ฉฐ, ๋‹จ์ˆœํ•œ ๋ธŒ๋ผ์šฐ์ € ๊ธฐ๋ฐ˜ ๋„๊ตฌ๋‚˜ ์œ„์ ฏ๋„ ์œ ํšจํ•œ ์†”๋ฃจ์…˜์ด ๋  ์ˆ˜ ์žˆ๋‹ค.

๊ณต๋™ ์ž‘์—… ๊ธฐ๋Šฅ ํ™œ์šฉ

์…ฐ์–ดํ”Œ๋ ˆ์ด๋Š” ๋น„์ „ ์šด์˜์ฒด์ œ์—์„œ ๊ฐ€์žฅ ๊ฐ•๋ ฅํ•œ ํ˜‘์—… ๊ธฐ๋Šฅ ์ค‘ ํ•˜๋‚˜๋กœ, ์ œํ’ˆ ๊ฐœ๋ฐœ์ด๋‚˜ ์—”์ง€๋‹ˆ์–ด๋ง ๋ถ€์„œ์˜ ์›๊ฒฉ ํ˜‘์—… ํ”Œ๋žซํผ์œผ๋กœ ๋งค์šฐ ์œ ์šฉํ•˜๋‹ค.

์—ฌ๋Ÿฌ ์‚ฌ์šฉ์ž๊ฐ€ ๋™์ผํ•œ ๊ฐ€์ƒ ๊ณต๊ฐ„์—์„œ ํ•˜๋‚˜์˜ ์›Œํฌํ”Œ๋กœยท์—…๋ฌดยท์•ฑ์„ ์‹ค์‹œ๊ฐ„์œผ๋กœ ํ•จ๊ป˜ ์‚ฌ์šฉํ•˜๋ฉฐ ์ฝ˜ํ…์ธ ๋ฅผ ๋ณด๊ณ  ์ˆ˜์ •ํ•  ์ˆ˜ ์žˆ๋‹ค. ์ด๋Ÿฌํ•œ ๊ธฐ๋Šฅ์€ ํ˜„์žฅ ๋˜๋Š” ์›๊ฒฉ ํ˜‘์—…์„ ๋ชจ๋‘ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•œ๋‹ค.

์ด ๊ธฐ๋Šฅ์€ ๋‹ค๋ฅธ ์• ํ”Œ ๊ธฐ๊ธฐ์™€๋„ ์—ฐ๋™๋˜์–ด, ๋น„์ „ ํ”„๋กœ๋ฅผ ์‚ฌ์šฉํ•˜์ง€ ์•Š๋Š” ์ง์›๋„ ๋น„๋ชฐ์ž…ํ˜• ์ฝ˜ํ…์ธ ์— ์ ‘๊ทผํ•  ์ˆ˜ ์žˆ๋‹ค. ๋ชฐ์ž…ํ˜• ํ˜‘์—…์ด ํ•„์š”ํ•œ ์ง์›์€ ํ•ด๋‹น ๊ธฐ๋Šฅ์„ ํ™œ์šฉํ•˜๊ณ , ์ผ๋ถ€ ์‚ฌ์–‘์ด๋‚˜ ๋ฐ์ดํ„ฐ๋งŒ ํ•„์š”ํ•œ ๋™๋ฃŒ๋Š” ๋งฅ, ์•„์ดํŒจ๋“œ, ์•„์ดํฐ์„ ํ†ตํ•ด ์ฐธ์—ฌํ•  ์ˆ˜ ์žˆ๋‹ค.

๋‹จ, ๋น„์ „ ์šด์˜์ฒด์ œ๋Š” ํŠน์ • ์ž‘์—… ๊ณต๊ฐ„์—์„œ ์ฝ˜ํ…์ธ  ๊ณต์œ ๋ฅผ ์ œํ•œํ•˜๋Š” ๊ธฐ๋Šฅ๋„ ํฌํ•จํ•˜๊ณ  ์žˆ๋‹ค. ์ด๋Š” ํ”„๋ผ์ด๋ฒ„์‹œ ๋ณดํ˜ธ, ์ง€์—ญ ๊ทœ์ œ, ๊ธฐ์—… ๋‚ด๋ถ€ ๋ฐ์ดํ„ฐ ๋ณด์•ˆ ์ •์ฑ… ๋“ฑ๊ณผ ๊ด€๋ จ๋œ ์กฐ์น˜๋‹ค.

๋„์ž… ์ ˆ์ฐจ๋Š” ๋‹จ๊ณ„์ ์œผ๋กœ

๋น„์ „ ์šด์˜์ฒด์ œ๋Š” ์ „ํ˜€ ์ƒˆ๋กœ์šด ์‚ฌ์šฉ์ž ๊ฒฝํ—˜์„ ์ œ๊ณตํ•˜๋Š” ํ”Œ๋žซํผ์ด๊ธฐ ๋•Œ๋ฌธ์—, ์ดˆ๊ธฐ ๋„์ž… ์‹œ์—๋Š” ์˜ค๋ฆฌ์—”ํ…Œ์ด์…˜๊ณผ ๊ต์œก์„ ๋Œ€๋ฉด์œผ๋กœ ์ง„ํ–‰ํ•˜๋Š” ๊ฒƒ์ด ๊ฐ€์žฅ ํšจ๊ณผ์ ์ด๋‹ค. ์‹œ์„  ์ถ”์ ์„ ํ†ตํ•œ ์„ ํƒ, ์‹ค์ œ ๊ณต๊ฐ„์— ์ฝ˜ํ…์ธ  ๋ฐฐ์น˜, ๊ฐ€์ƒ ์ฝ˜ํ…์ธ  ์กฐ์ž‘ ๋“ฑ์€ ์ต์ˆ™ํ•˜์ง€ ์•Š์€ ์‚ฌ์šฉ์ž์—๊ฒŒ ์ง„์ž… ์žฅ๋ฒฝ์ด ๋  ์ˆ˜ ์žˆ๋‹ค.

์ „๋‹ด ์ง€์› ์ธ๋ ฅ ๋ฐฐ์น˜

์ง€์› ์ธ๋ ฅ ์ผ๋ถ€๋ฅผ ๋น„์ „ ์šด์˜์ฒด์ œ ์ „๋‹ด์œผ๋กœ ์ง€์ •ํ•˜๋Š” ๊ฒƒ์ด ๋ฐ”๋žŒ์งํ•˜๋‹ค. ์ง€์› ์ธ๋ ฅ๋„ ์ง์ ‘ ๊ธฐ๊ธฐ๋ฅผ ์‚ฌ์šฉํ•ด๋ณผ ์ˆ˜ ์žˆ์–ด์•ผ ํ•˜๋ฉฐ, ์‹ค์ œ๋กœ ๊ธฐ๊ธฐ๋ฅผ ๊ฒฝํ—˜ํ•˜์ง€ ์•Š๊ณ ๋Š” ์‚ฌ์šฉ์ž ๋ฌธ์ œ๋ฅผ ์ œ๋Œ€๋กœ ํŒŒ์•…ํ•˜๊ฑฐ๋‚˜ ํ•ด๊ฒฐํ•˜๊ธฐ ์–ด๋ ต๋‹ค.

๋น„์šฉ์ด ๋ถ€๋‹ด๋œ๋‹ค๋ฉด ํŒ€๋‹น ํ•œ ๋Œ€์˜ ๊ธฐ๊ธฐ๋งŒ ๋ฐฐ์ •ํ•ด๋„ ์ถฉ๋ถ„ํ•˜๋‹ค. ์ง์ ‘ ๊ธฐ๊ธฐ๋ฅผ ์‚ฌ์šฉํ•ด๋ณธ ์ง€์› ๋‹ด๋‹น์ž๋Š” ๋ฌธ์ œ๋ฅผ ๋น ๋ฅด๊ฒŒ ์žฌํ˜„ํ•˜๊ณ  ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ๋Š” ์—ญ๋Ÿ‰์ด ๋†’์•„์ง„๋‹ค.

๋Š์ž„์—†๋Š” ๋ณ€ํ™”์— ๋Œ€๋น„ํ•˜๋ผ

ํ˜ผํ•ฉํ˜„์‹ค ๊ธฐ์ˆ ์€ ์ธ๊ณต์ง€๋Šฅ ๋‹ค์Œ์œผ๋กœ ๋ณ€ํ™” ์†๋„๊ฐ€ ๋น ๋ฅธ ๊ธฐ์ˆ  ๋ถ„์•ผ๋‹ค. ๋ฉ”ํƒ€, ์•ˆ๋“œ๋กœ์ด๋“œ ๊ธฐ๋ฐ˜ ํ™•์žฅํ˜„์‹ค ์ƒํƒœ๊ณ„๋„ ๋น„์ „ ํ”„๋กœ์™€ ๊ฒฝ์Ÿํ•˜๊ธฐ ์œ„ํ•ด ์ ๊ทน์ ์œผ๋กœ ์›€์ง์ด๊ณ  ์žˆ๋‹ค. ์‹œ์žฅ์€ ์—ฌ์ „ํžˆ ์œ ๋™์ ์ด๋ฉฐ, ์• ํ”Œ๊ณผ ์ง์ ‘ ๊ด€๋ จ ์—†๋Š” ๋‰ด์Šค๋ผ๋„ ๋น„์ „ ์šด์˜์ฒด์ œ์˜ ํ–ฅํ›„ ๊ธฐํš์— ์ฐธ๊ณ ๊ฐ€ ๋  ์ˆ˜ ์žˆ๋‹ค.
dl-ciokorea@foundryco.com

The 10 hottest IT skills for 2026

27 November 2025 at 05:00

Gen AI has reshaped the IT skills market as companies restructure for AI strategies, and prioritize candidates and employees with AI skills. Data from Indeedโ€™s 2025 Tech Talent Report show that the top four roles affected by AI-related restructuring include software engineers and developers, QA engineers, product managers, and project managers. Companies are now focusing their efforts and hiring budgets on professionals with skills in cybersecurity, data analytics and analysis, and building or managing AI teams.

This reprioritization of IT roles has also created a shift in the most in-demand IT skills that jobseekers will want to have on their rรฉsumรฉs. Organizations now expect candidates to have basic prompt engineering skills at minimum, even for entry-level IT roles. And beyond that, theyโ€™re looking for IT professionals who can help oversee, implement, secure, and manage AI tools and strategies.

Data from Indeed reveal these are the 10 IT skills that grew the most desirable between 2024 and 2025, based on how many times they appeared as a requirement in a job posting year over year.

AI

Itโ€™s no surprise that AI is at the top of the list for one of the most in-demand skills based on growth in tech job postings listed since 2024. Companies are scrambling to adopt AI as it rapidly finds its way into every industry and career path. In 2024, there were just over 5 million job postings that required AI skills, and in 2025, that number grew by more than 4 million. So candidates, even for those working outside of tech, are now expected to have some level of AI skills, whether itโ€™s prompt engineering, natural language processing, or using AI for programming and coding.

Python

Python is a programming language used in several fields, including data analysis, web development, software programming, scientific computing, and for building AI and ML models. Itโ€™s a versatile language used by a wide range of IT professionals such as software developers, web developers, data scientists, data analysts, ML engineers, cybersecurity analysts, cloud engineers, and more. Its widespread use in the enterprise makes it a steady entry on any in-demand skill list. In 2024, there were just over 15 million job listings requiring Python skills, and that grew to just under 18 million in 2025. Although more organizations are relying on AI for coding, they still need skilled professionals who understand key programming languages to write more complex code, and to help with prompt and QA code written by AI.

Algorithms

As more companies embrace AI and its ability to streamline coding and programming, organizations are also becoming more reliant on algorithms to help guide and dictate those processes. Algorithmic thinking requires a complex understanding of databases and programming, high-value critical thinking, and problem solving. Algorithm skills were listed as a requirement on around 180,000 job postings in 2024, which jumped to over 2 million in 2025. AI has taken over more of the entry-level work, leaving organizations looking for higher-skilled professionals who can help build and guide AI systems, and who understand how to build efficient algorithms.

CI/CD

Continuous integration and continuous delivery or deployment skills have grown in demand in the wake of AI implementation to help streamline the software development lifecycle. Professionals with CI/CD skills can handle tasks such as building tools used for automation and scripting, and have a strong understanding of concepts such as containerization, cloud integration, and automated testing. In 2024, there were just under 7 million job listings that looked for CI/CD skills and that number jumped to just over 9 million in 2025.

Google Cloud

Google Cloud is a popular platform to build, deploy, and manage IT solutions for an organization, with several certifications offered by Google to certify your professional skills with, and knowledge of, Google Cloud. Organizations have adopted the cloud in recent years, moving tools, services, and data storage to solutions hosted by Googleโ€™s cloud services. Cloud tools are critical for AI development, allowing for more versatile and agile storage solutions to host the large data sets required to train and run AI tools. Google Cloud skills were a requirement for around 3.5 million job listings in 2024, but that rose to just over 5.3 million in 2025.

AWS

Amazon Web Services is the most widely used cloud platform today. Central to cloud strategies across nearly every industry,ย AWS skillsย are in high demand as organizations look to make the most of the platformโ€™s wide range of offerings. Itโ€™s a common skill for cloud engineers, DevOps engineers, solutions architects, data engineers, cybersecurity analysts, software developers, network administrators, and many more IT roles. In 2024, AWS skills were still popular and were listed as a requirement on just over 12 million job listings, which jumped to over 13.7 million in 2025.

Analysis Skills

AI has taken a lot of entry-level and rote work off the table for IT professionals, which has created more room for higher-level skills such as analytical thinking. Since AI still doesnโ€™t create perfect outputs with every prompt, companies need a human eye and analytical mind to catch AI hallucinations and errors, especially when it comes to numbers and data. Analysis skills have been critical for organizations for a while now; in 2024, just over 19 million job listings required analysis skills, a number that number jumped to just over 21 million in 2025.

Cybersecurity

An increased reliance on AI has created more vulnerabilities for organizations. As they take more products and services online and integrate AI, more opportunities are created for security attacks. Cybersecurity skills were a requirement on around 2.4 million job listings in 2024, which grew to just over 4 million in 2025. Whether organizations look to integrate AI into cybersecurity solutions or help prevent new sophisticated attacks that use AI to breach systems, security is a top priority for organizations as they move forward with AI.

Software troubleshooting

Although organizations are increasingly using AI to write basic codes and scripts to build software tools, organizations still need human IT professionals to identify flaws, security issues, and other potential anomalies in the final product. Software troubleshooting skills were listed as a requirement on just over 9 million job listings in 2024, but this year, that number grew to just under 11 million. Itโ€™s an area of IT that requires communication, problem-solving, critical thinking, and technical skills to identify software issues and troubleshoot problems for clients and customers.

Machine Learning

ML is fundamental to AI development and requires a strong expertise of not only AI but also natural language processing. Organizations are seeking professionals with ML skills to support AI initiatives, and the future of AI adoption in the enterprise. In 2024, there were around 3.7 million job listings that looked for ML skills, while that jumped to over 5 million in 2025. IT professionals with ML skills will continue to be in demand as companies embrace AI processes and look for professionals to help support and maintain AI systems.

AI์˜ ROI๋ฅผ ๋†’์ด๋Š” CIO์˜ 5๋‹จ๊ณ„ ์ฒดํฌ๋ฆฌ์ŠคํŠธ

27 November 2025 at 00:26

์˜ฌํ•ด ์ดˆ MIT๋Š” โ€œ์กฐ์ง์˜ 95%๊ฐ€ AI ํˆฌ์ž์—์„œ ์•„๋ฌด๋Ÿฐ ์ˆ˜์ต์„ ์–ป์ง€ ๋ชปํ•˜๊ณ  ์žˆ๋‹คโ€๋Š” ์กฐ์‚ฌ ๊ฒฐ๊ณผ๋ฅผ ๋ฐœํ‘œํ–ˆ๋‹ค. ๋ฏธ๊ตญ ๋‚ด ์ƒ์„ฑํ˜• AI ๊ด€๋ จ ๋‚ด๋ถ€ ํ”„๋กœ์ ํŠธ์—๋งŒ 300์–ต ๋‹ฌ๋Ÿฌ ์ด์ƒ์ด ํˆฌ์ž…๋œ ์ƒํ™ฉ์ด์—ˆ๋‹ค. ์™œ ์ด๋ ‡๊ฒŒ ๋งŽ์€ AI ํ”„๋กœ์ ํŠธ๊ฐ€ ๊ธฐ๋Œ€๋งŒํผ์˜ ROI๋ฅผ ๋‚ด์ง€ ๋ชปํ• ๊นŒ? IT ์ปจ์„คํŒ… ํšŒ์‚ฌ ์ฝ”๊ทธ๋‹ˆ์ „ํŠธ(Cognizant)์˜ ๊ธ€๋กœ๋ฒŒ CIO ๋‹ ๋ผ๋งˆ์‚ฌ๋ฏธ๋Š” โ€œAI๊ฐ€ ๋น„์ฆˆ๋‹ˆ์Šค ๊ฐ€์น˜์™€ ๋ช…ํ™•ํžˆ ์—ฐ๊ฒฐ๋˜์ง€ ์•Š๊ธฐ ๋•Œ๋ฌธโ€์ด๋ผ๋ฉฐ, โ€œ๊ธฐ์ˆ ์ ์œผ๋กœ ์ธ์ƒ์ ์ด์ง€๋งŒ ์‹ค์ œ ๋ฌธ์ œ ํ•ด๊ฒฐ์ด๋‚˜ ์‹ค์งˆ์  ์„ฑ๊ณผ๋กœ ์ด์–ด์ง€์ง€ ์•Š๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค.

IT ์ฑ…์ž„์ž๋Š” ์ข…์ข… ์œ ํ–‰์— ํœฉ์“ธ๋ ค ๋น„์ฆˆ๋‹ˆ์Šค ์„ฑ๊ณผ๋ฅผ ๊ณ ๋ คํ•˜์ง€ ์•Š์€ ์ฑ„ AI ์‹คํ—˜์— ๋›ฐ์–ด๋“ ๋‹ค. ์•„์‚ฌ๋‚˜(Asana)์˜ CIO ์‚ฌ์ผ“ ์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” โ€œ๋งŽ์€ ๊ธฐ์—…์ด ๋น„์ฆˆ๋‹ˆ์Šค ๊ฒฐ๊ณผ๋ณด๋‹ค ๋ชจ๋ธ์ด๋‚˜ ํŒŒ์ผ๋Ÿฟ ํ”„๋กœ์ ํŠธ๋ถ€ํ„ฐ ์‹œ์ž‘ํ•œ๋‹ค. ํŒ€์ด ์›Œํฌํ”Œ๋กœ์šฐ๋ฅผ ๋‹ค์‹œ ์„ค๊ณ„ํ•˜๊ฑฐ๋‚˜ ์†์ต ์ฑ…์ž„์ž๋ฅผ ์ง€์ •ํ•˜์ง€ ์•Š์€ ์ฑ„ ๊ณ ๋ฆฝ๋œ ์‹œ์—ฐ๋งŒ ์ง„ํ–‰ํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ œํ’ˆ ์ค‘์‹ฌ ์‚ฌ๊ณ ์˜ ๋ถ€์žฌ, ๋ถ€์‹คํ•œ ๋ฐ์ดํ„ฐ ๊ด€๋ฆฌ, ๋ถ€์žฌํ•œ ๊ฑฐ๋ฒ„๋„Œ์Šค, ๊ทธ๋ฆฌ๊ณ  AI ํ™œ์šฉ์„ ์žฅ๋ คํ•˜์ง€ ์•Š๋Š” ์กฐ์ง ๋ฌธํ™”๊ฐ€ ๊ฒน์น˜๋ฉด ๋ถ€์ •์ ์ธ ๊ฒฐ๊ณผ๋กœ ์ด์–ด์ง„๋‹ค. ์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” โ€œํ”„๋กœ์„ธ์Šค๋ฅผ ๋ฐ”๊พธ์ง€ ์•Š์œผ๋ฉด AI๋Š” ํ˜„์žฌ์˜ ๋น„ํšจ์œจ์„ ๋” ๋น ๋ฅด๊ฒŒ ๋ฐ˜๋ณตํ•  ๋ฟโ€์ด๋ผ๊ณ  ๊ฒฝ๊ณ ํ–ˆ๋‹ค.

์กฐ์ง์ด ๋ถ€์ •์ ์ธ AI ROI๋ฅผ ํ”ผํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ํšจ๊ณผ์ ์ธ ๋ณ€ํ™” ๊ด€๋ฆฌ๊ฐ€ ํ•ต์‹ฌ์ด๋‹ค. ๋‹ค์Œ์€ CIO๊ฐ€ ๋ฐ”๋กœ ์‹ค์ฒœํ•  ์ˆ˜ ์žˆ๋Š” ๋‹ค์„ฏ ๊ฐ€์ง€ ๋ณ€ํ™” ๊ด€๋ฆฌ ์ง€์นจ์ด๋‹ค. ์ด ์ง€์นจ์„ ์ง€ํ‚จ๋‹ค๋ฉด, ๋”ฐ๋ฅด๋ฉด ์กฐ์ง์€ ์•ˆํ‹ฐ ํŒจํ„ด์—์„œ ๊ตํ›ˆ์„ ์–ป๊ณ , ์„ฑ๊ณต์ ์ธ AI ํ”„๋กœ์ ํŠธ๋ฅผ ์ž…์ฆํ•  ์ˆ˜ ์žˆ๋Š” ์ธก์ • ์ง€ํ‘œ๋ฅผ ๋ฐœ๊ฒฌํ•  ์ˆ˜ ์žˆ์„ ๊ฒƒ์ด๋‹ค.

1. ๋น„์ฆˆ๋‹ˆ์Šค ๋ชฉํ‘œ๋ฅผ ๋ช…ํ™•ํžˆ ํ•˜๊ณ  ๋ฆฌ๋”์‹ญ ์ •๋ ฌ์„ ํ†ตํ•ด AI ํ”„๋กœ์ ํŠธ๋ฅผ ์ด๋ˆ๋‹ค

AI ํ”„๋กœ์ ํŠธ๋Š” ์ตœ๊ณ ๊ฒฝ์˜์ง„์˜ ํ›„์›๊ณผ ๋ช…ํ™•ํ•œ ๋น„์ „ ์—†์ด๋Š” ์„ฑ๊ณตํ•˜๊ธฐ ์–ด๋ ต๋‹ค. CMIT ์†”๋ฃจ์…˜์˜ ์‚ฌ์žฅ ๊ฒธ ์ˆ˜์„ vCIO ์• ๋ค ๋กœํŽ˜์ฆˆ๋Š” โ€œ๊ฐ•๋ ฅํ•œ ๋ฆฌ๋”์‹ญ์€ AI ํˆฌ์ž๋ฅผ ์‹ค์งˆ์  ๊ฒฐ๊ณผ๋กœ ์ „ํ™˜ํ•˜๋Š” ๋ฐ ํ•„์ˆ˜์ด๋‹ค. CEO๋‚˜ ์ด์‚ฌํšŒ ์ฐจ์›์˜ ํ›„์›๊ณผ ๊ฐ๋…์ด ์žˆ์„์ˆ˜๋ก ROI๊ฐ€ ๋†’๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์˜ˆ๋ฅผ ๋“ค์–ด IT ์ปจ์„คํŒ… ๊ธฐ์—… ์ œ๋น„์•„(Xebia)๋Š” ๊ธ€๋กœ๋ฒŒ CIO ์Šค๋ฏธํŠธ ์ƒน์ปค๊ฐ€ ์ฃผ์žฌํ•˜๋Š” โ€˜AI ์ถ”์ง„์œ„์›ํšŒโ€™๋ฅผ ์šด์˜ํ•˜๊ณ  ์žˆ๋‹ค. ์œ„์›ํšŒ์—๋Š” ๊ธ€๋กœ๋ฒŒ CFO, AI ๋ฐ ์ž๋™ํ™” ์ฑ…์ž„์ž, IT ์ธํ”„๋ผยท๋ณด์•ˆ ์ฑ…์ž„์ž, ๋น„์ฆˆ๋‹ˆ์Šค ์šด์˜ ์ฑ…์ž„์ž๊ฐ€ ํฌํ•จ๋ผ ์žˆ๋‹ค. ์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” โ€œAI ํ”„๋กœ์ ํŠธ๋งˆ๋‹ค ๋ชฉํ‘œ์— ๊ธฐ๋ฐ˜ํ•œ ์ฑ…์ž„ ๋ฆฌ๋”๋ฅผ ์ง€์ •ํ•ด์•ผ ํ•œ๋‹ค. ์กฐ์ง ์ „๋ฐ˜์˜ ํ”„๋กœ์ ํŠธ ๊ด€๋ฆฌ ์‹ค์„ ์„ธ์›Œ ์ฃผ์š” ์‚ฌ์šฉ๋ก€๋ฅผ ์ •์˜ํ•˜๊ณ  ์„ฑ๊ณต ์ง€ํ‘œ์™€ ๊ฐ€๋“œ๋ ˆ์ผ์„ ์„ค์ •ํ•˜๋ฉฐ ์ง„ํ–‰ ์ƒํ™ฉ์„ ์ •๊ธฐ์ ์œผ๋กœ ๊ณต์œ ํ•˜๋ผโ€๋ผ๊ณ  ์กฐ์–ธํ–ˆ๋‹ค.

ํ•˜์ง€๋งŒ ๋ฆฌ๋”์‹ญ์ด ํ™•๋ฆฝ๋ผ๋„ ์ง์› ๋Œ€๋ถ€๋ถ„์€ ์ผ์ƒ ์—…๋ฌด์—์„œ AI๋ฅผ ์–ด๋–ป๊ฒŒ ํ™œ์šฉํ•ด์•ผ ํ• ์ง€ ๋ชจ๋ฅธ๋‹ค. ์Šคํ‚ฌ์†Œํ”„ํŠธ์˜ CIO ์˜ฌ๋ผ ๋ฐ์ผ๋ฆฌ๋Š” โ€œ๋„๊ตฌ๋ฅผ ์ œ๊ณต๋ฐ›๋”๋ผ๋„ ๋Œ€๋ถ€๋ถ„ ์ง์›์€ ์–ด๋””์„œ๋ถ€ํ„ฐ ์‹œ์ž‘ํ•ด์•ผ ํ• ์ง€ ๋ชจ๋ฅธ๋‹ค. ์กฐ์ง ๋‚ด AI ํ™œ์šฉ์„ ์„ ๋„ํ•  โ€˜์ฑ”ํ”ผ์–ธโ€™์„ ์ง€์ •ํ•ด ์‹ค์งˆ์  ์‚ฌ์šฉ๋ก€์™€ ํŒ์„ ๊ณต์œ ํ•ด์•ผ ํ•œ๋‹ค. ํŠนํžˆ ์ฝ”ํŒŒ์ผ๋Ÿฟ ๊ฐ™์€ ๋„๊ตฌ์˜ ํšจ์œจ์ ์ธ ์‚ฌ์šฉ๋ฒ•์„ ์ „ํŒŒํ•˜๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•˜๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ผ๋งˆ์‚ฌ๋ฏธ๋Š” โ€œ๋ฆฌ๋”๋Š” ๋ฐ์ดํ„ฐ ์ค‘์‹ฌ ๋ฌธํ™”๋ฅผ ์กฐ์„ฑํ•˜๊ณ  AI๊ฐ€ ์‹ค์ œ ๋น„์ฆˆ๋‹ˆ์Šค ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๋Š” ๋น„์ „์„ ์ œ์‹œํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค. ๋˜, ์ด๋ฅผ ์œ„ํ•ด ๊ฒฝ์˜์ง„, ๋ฐ์ดํ„ฐ ๊ณผํ•™์ž, IT ๋ถ€์„œ ๊ฐ„ ๊ธด๋ฐ€ํ•œ ํ˜‘๋ ฅ์ด ํ•„์š”ํ•˜๋ฉฐ, ํŒŒ์ผ๋Ÿฟ ํ”„๋กœ์ ํŠธ์˜ ์‹คํ–‰๊ณผ ์„ฑ๊ณผ ์ธก์ •์„ ๊ฑฐ์ณ์•ผ ํ•œ๋‹ค๊ณ  ๋ง๋ถ™์˜€๋‹ค.

2. ์ธ์žฌ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์žฌํŽธํ•˜๊ณ  ์—…์Šคํ‚ฌ๋ง(์—ญ๋Ÿ‰ ๊ฐ•ํ™”)์— ํˆฌ์žํ•œ๋‹ค

AI ROI๋ฅผ ๋†’์ด๊ธฐ ์œ„ํ•ด CIO๋Š” ์ธ์žฌ ์ „๋žต์„ ์ƒˆ๋กญ๊ฒŒ ์„ค๊ณ„ํ•ด์•ผ ํ•œ๋‹ค. ๋ผ๋งˆ์‚ฌ๋ฏธ๋Š” โ€œCIO๋Š” ์ธ์žฌ ๋ฐ ๊ด€๋ฆฌ ์ „๋žต์„ ์กฐ์ •ํ•ด AI ์ฑ„ํƒ๊ณผ ROI๋ฅผ ๊ทน๋Œ€ํ™”ํ•ด์•ผ ํ•œ๋‹คโ€๋ฉฐ โ€œ๋ฐ์ดํ„ฐ ๊ณผํ•™์ž, ํ”„๋กฌํ”„ํŠธ ์—”์ง€๋‹ˆ์–ด ๊ฐ™์€ ์ƒˆ๋กœ์šด ์—ญํ• ์„ ๋งŒ๋“ค๊ณ  ๊ธฐ์กด ์ง์›์„ ์žฌ๊ต์œกํ•˜๋Š” ์ ‘๊ทผ์ด ํ•„์š”ํ•˜๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋กœํŽ˜์ฆˆ๋Š” โ€œ์ธ์žฌ๋Š” ๋ชจ๋“  AI ์ „๋žต์˜ ํ•ต์‹ฌ์ด๋‹ค. ๊ต์œก, ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜, ์ „๋ฌธ ์ธ๋ ฅ ํ™•๋ณด์— ํˆฌ์žํ•ด์•ผ ์ง์›์ด AI๋ฅผ ๋ฐ›์•„๋“ค์ด๊ณ  ์„ฑ๊ณผ๋ฅผ ๋‚ผ ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ—ธ๋‹ค. ๋˜, ํ•ด์ปคํ†ค์ด๋‚˜ ์‚ฌ๋‚ด ๊ต์œก์ด ์ง์›์˜ ๊ธฐ์ˆ ๊ณผ ์ž์‹ ๊ฐ์„ ๋†’์ด๋Š” ํšจ๊ณผ๊ฐ€ ํฌ๋‹ค๊ณ  ๋ง๋ถ™์˜€๋‹ค.

์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” โ€œ๋ชจ๋“  ์ง์›์—๊ฒŒ๋Š” ๊ธฐ๋ณธ์ ์ธ ํ”„๋กฌํ”„ํŠธ ์ดํ•ด๋ ฅ๊ณผ ์•ˆ์ „ ๊ต์œก์ด ํ•„์š”ํ•˜๊ณ , ํŒŒ์›Œ ์œ ์ €์—๊ฒŒ๋Š” ์›Œํฌํ”Œ๋กœ์šฐ ์„ค๊ณ„์™€ AI ์—์ด์ „ํŠธ ๊ตฌ์ถ• ๋Šฅ๋ ฅ์ด ํ•„์š”ํ•˜๋‹ค. ์šฐ๋ฆฌ๋Š” ์ „์‚ฌ ์„ค๋ฌธ์„ ํ†ตํ•ด ์—ญ๋Ÿ‰ ์ˆ˜์ค€์„ ํŒŒ์•…ํ•˜๊ณ  ๊ต์œก ๋ชฉํ‘œ๋ฅผ ์„ค์ •ํ•ด ์„ฑ์ˆ™๋„๊ฐ€ ์ œ๋Œ€๋กœ ํ–ฅ์ƒ๋๋Š”์ง€ ์žฌ์ธก์ •ํ–ˆ๋‹คโ€๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

์•„์ŠคํŽ˜๋ฆฌํƒ€์Šค ์ปจ์„คํŒ…(Asperitas Consulting)์˜ ํด๋ผ์šฐ๋“œ ์‚ฌ์—… ์ฑ…์ž„์ž ์Šค์ฝง ํœ ๋Ÿฌ๋Š” โ€œAI ๋„์ž…์€ ์ธ์  ์—ญ๋Ÿ‰๋ฟ ์•„๋‹ˆ๋ผ ์—…๋ฌด ํ”„๋กœ์„ธ์Šค ์ž์ฒด๋ฅผ ๋‹ค์‹œ ์ ๊ฒ€ํ•ด์•ผ ํ•œ๋‹คโ€๊ณ  ๋งํ–ˆ๋‹ค. ์ฆ‰, ์–ด๋–ค ์—…๋ฌด๋ฅผ ๋ˆ„๊ฐ€ ์ˆ˜ํ–‰ํ•ด์•ผ ํ•˜๋Š”์ง€๋ฅผ ์žฌ์ •์˜ํ•ด์•ผ ํ•œ๋‹ค๋Š” ๋œป์ด๋‹ค.

์Šคํ‚ฌ์†Œํ”„ํŠธ์˜ ๋ฐ์ผ๋ฆฌ๋Š” โ€œํ˜„๋Œ€์˜ ์ธ์žฌ ์ „๋žต์€ 4B(Build, Buy, Borrow, Bot) ์ „๋žต์œผ๋กœ ๊ท ํ˜•์„ ๋งž์ถฐ์•ผ ํ•œ๋‹คโ€๋ผ๋ฉฐ, โ€œ์กฐ์ง์„ ๊ณ ์ •๋œ ์ง๋ฌด๊ฐ€ ์•„๋‹ˆ๋ผ โ€˜์—ญ๋Ÿ‰์˜ ์ง‘ํ•ฉ์ฒดโ€™๋กœ ๋ณด๊ณ , ๋‚ด๋ถ€ ์ธ๋ ฅยท์†Œํ”„ํŠธ์›จ์–ดยทํŒŒํŠธ๋„ˆยท์ž๋™ํ™” ๊ธฐ์ˆ ์„ ์ƒํ™ฉ์— ๋งž๊ฒŒ ์กฐํ•ฉํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์Šคํ‚ฌ์†Œํ”„ํŠธ์˜ ํŒ€์€ ๊นƒํ—ˆ๋ธŒ ์ฝ”ํŒŒ์ผ๋Ÿฟ์„ ํ™œ์šฉํ•ด ๊ณ ๊ฐ์šฉ ํ•™์Šต ํฌํ„ธ์„ ๋น ๋ฅด๊ฒŒ ๊ตฌ์ถ•ํ–ˆ๋‹ค. ์ด ๊ฒฝํ—˜์„ ํ†ตํ•ด AI ๋„์šฐ๋ฏธ์™€ ์ธ๊ฐ„ ๊ฐœ๋ฐœ์ž๊ฐ€ ํ˜‘์—…ํ•  ๋•Œ ์ƒ์‚ฐ์„ฑ์ด ๋น„์•ฝ์ ์œผ๋กœ ๋†’์•„์ง„๋‹ค๋Š” ์ ์„ ํ™•์ธํ–ˆ๋‹ค.

๋ผ๋งˆ์‚ฌ๋ฏธ๋Š” โ€œ์ง์›์ด AI ๋•Œ๋ฌธ์— ์ผ์ž๋ฆฌ๋ฅผ ์žƒ์„ ๊ฒƒ์ด๋ผ๋Š” ๋ถˆ์•ˆ์„ ํ•ด์†Œํ•˜๋ ค๋ฉด, ์™œ AI๋ฅผ ๋„์ž…ํ•˜๋Š”์ง€ ๋ช…ํ™•ํžˆ ์„ค๋ช…ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค. ์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ” ์—ญ์‹œ โ€œํ•ต์‹ฌ์€ ์‹ ๋ขฐ๋‹ค. AI๊ฐ€ ๋ฐ˜๋ณต ์—…๋ฌด๋ฅผ ์ค„์ด๊ณ  ์ž„ํŒฉํŠธ๋ฅผ ๋†’์ธ๋‹ค๋Š” ์ ์„ ๋ณด์—ฌ์ฃผ๋ฉด ์ž์—ฐ์Šค๋Ÿฝ๊ฒŒ ์ฑ„ํƒ์ด ๋’ค๋”ฐ๋ฅธ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

3. AI์˜ ๊ฐ€์น˜๋ฅผ ์˜จ์ „ํžˆ ํ™•๋ณดํ•˜๊ธฐ ์œ„ํ•ด ์กฐ์ง ํ”„๋กœ์„ธ์Šค๋ฅผ ์žฌ์„ค๊ณ„ํ•œ๋‹ค

์ธ์žฌ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ๋ฐ”๊พธ๋Š” ๊ฒƒ๋งŒ์œผ๋กœ๋Š” ์ถฉ๋ถ„ํ•˜์ง€ ์•Š๋‹ค. ๋กœํŽ˜์ฆˆ๋Š” โ€œAI์˜ ์ž ์žฌ๋ ฅ์„ ์™„์ „ํžˆ ๋ฐœํœ˜ํ•˜๋ ค๋ฉด ์กฐ์ง์˜ ์šด์˜ ๋ฐฉ์‹ ์ž์ฒด๋ฅผ ์žฌ๊ตฌ์„ฑํ•ด์•ผ ํ•œ๋‹คโ€๊ณ  ์กฐ์–ธํ–ˆ๋‹ค. AI๋ฅผ ๋‹จ์ˆœํ•œ โ€˜๋ถ€๊ฐ€ ๊ธฐ๋Šฅโ€™์ด ์•„๋‹ˆ๋ผ ํ•ต์‹ฌ ์šด์˜ ํ”„๋กœ์„ธ์Šค์— ํ†ตํ•ฉํ•ด์•ผ ํ•œ๋‹ค๋Š” ์˜๋ฏธ๋‹ค.

์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” โ€œAI ๊ธฐ๋ฐ˜ ์›Œํฌํ”Œ๋กœ์šฐ๋ฅผ ์ œํ’ˆ์ฒ˜๋Ÿผ ๊ด€๋ฆฌํ•ด์•ผ ํ•œ๋‹ค. ์š”์ฒญ, ์šฐ์„ ์ˆœ์œ„, ๋กœ๋“œ๋งต์„ ์ฒด๊ณ„์ ์œผ๋กœ ์šด์˜ํ•˜๊ณ , ๋ฌธ์ œ ์ •์˜์™€ ๊ฐ€์น˜ ๊ฐ€์„ค์„ ๋ช…ํ™•ํžˆ ์„ค์ •ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค.

์ œ๋น„์•„๋Š” AI ํ”„๋กœ์ ํŠธ๋งˆ๋‹ค ์„ธ ๋‹จ๊ณ„ ๊ฒ€์ฆ ์ ˆ์ฐจ๋ฅผ ๊ฑฐ์นœ๋‹ค. โ€˜๊ฐ€์น˜ ํ‰๊ฐ€โ†’๋น„์ฆˆ๋‹ˆ์Šค ์Šน์ธโ†’IT ์ด๊ด€ ๋ฐ ๋ชจ๋‹ˆํ„ฐ๋งโ€™์˜ ๊ตฌ์กฐ๋‹ค. ์ƒน์ปค๋Š” โ€œ์ด ๊ณผ์ •์„ ํ†ตํ•ด ๋ถ€์„œ ๊ฐ„ ํ”„๋กœ์„ธ์Šค๊ฐ€ ๋‹จ์ˆœํ™”๋˜๊ณ  ์‚ฌ์ผ๋กœ๊ฐ€ ์ค„์–ด๋“ ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ผ๋งˆ์‚ฌ๋ฏธ๋Š” โ€œ๋Œ€๋ถ€๋ถ„ ๊ธฐ์—…์ด ํ•„์š”ํ•œ ๋ณ€ํ™” ๊ด€๋ฆฌ์˜ ๋ฒ”์œ„๋ฅผ ๊ณผ์†Œํ‰๊ฐ€ํ•œ๋‹ค. ์‚ฌ์ผ๋กœํ˜• ์˜์‚ฌ๊ฒฐ์ •์—์„œ ๋ฐ์ดํ„ฐ ์ค‘์‹ฌ ๋ฐฉ์‹์œผ๋กœ ์ „ํ™˜ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๋ฉฐ, โ€œAI๊ฐ€ ์ƒ์„ฑํ•œ ๊ฒฐ๊ณผ๋ฅผ ์—…๋ฌด ํ”„๋กœ์„ธ์Šค์— ์ž์—ฐ์Šค๋Ÿฝ๊ฒŒ ํ†ตํ•ฉํ•˜๊ณ , ์ง์›์ด ๋ฐ์ดํ„ฐ ๊ธฐ๋ฐ˜ ํ†ต์ฐฐ๋กœ ์˜์‚ฌ๊ฒฐ์ •์„ ๋‚ด๋ฆด ์ˆ˜ ์žˆ์–ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค.

๋ฐ์ผ๋ฆฌ๋Š” โ€œAI๊ฐ€ ํšจ์œจํ™”ํ•  ์ˆ˜ ์žˆ๋Š” ์—…๋ฌด๋ฅผ ์ฐพ๊ธฐ ์œ„ํ•ด์„œ๋Š” ํ˜„์žฌ์˜ ์›Œํฌํ”Œ๋กœ์šฐ๋ฅผ ์ •ํ™•ํžˆ ํŒŒ์•…ํ•ด์•ผ ํ•œ๋‹ค. ์—…๋ฌด ์ „๋ฌธ๊ฐ€๊ฐ€ ํ”„๋กœ์„ธ์Šค๋ฅผ ๊ฒ€ํ† ํ•ด ์ตœ์ ํ™”ํ•  ์˜์—ญ์„ ์ฐพ์•„์•ผ ํ•˜๋ฉฐ, ๊ฐ ๋ถ€์„œ์— AI๋ฅผ ์–ด๋–ป๊ฒŒ ๋…น์—ฌ๋‚ผ์ง€ ์งˆ๋ฌธ์„ ๋˜์ง€๋Š” ์ธ๋ฌผ์„ ์ง€์ •ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์Šคํ‚ฌ์†Œํ”„ํŠธ๋Š” AI ์‚ฌ์šฉ๋ก€๋ฅผ ์ฒด๊ณ„ํ™”ํ•˜๊ธฐ ์œ„ํ•ด โ€˜์—์ด์ „ํŠธ ๋ ˆ์ง€์ŠคํŠธ๋ฆฌโ€™๋ฅผ ๊ตฌ์ถ•ํ–ˆ๋‹ค. AI ์—์ด์ „ํŠธ์˜ ๊ธฐ๋Šฅ, ๊ฐ€๋“œ๋ ˆ์ผ, ๋ฐ์ดํ„ฐ ๊ด€๋ฆฌ ๋ฐฉ์‹์„ ๋ฌธ์„œํ™”ํ•ด ํ‘œ์ค€ํ™”ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์œค๋ฆฌ์™€ ๊ฑฐ๋ฒ„๋„Œ์Šค๋ฅผ ํฌํ•จํ•œ ๊ธฐ์—… AI ๊ด€๋ฆฌ ์ฒด๊ณ„๋ฅผ ์ •๋ฆฝ ์ค‘์ด๋‹ค.

์•„์ŠคํŽ˜๋ฆฌํƒ€์Šค์˜ ํœ ๋Ÿฌ๋Š” โ€œAI ์ฑ„ํƒ์„ ๊ฐ€์†ํ™”ํ•˜๋ ค๋ฉด โ€˜AI ์Šค์™“ํŒ€(SWAT team)โ€™์„ ์šด์˜ํ•ด ์ดˆ๊ธฐ ์žฅ์• ๋ฅผ ํ•ด๊ฒฐํ•˜๊ณ  ์‚ฌ์šฉ์ž ์ง€์›์„ ๊ฐ•ํ™”ํ•˜๋Š” ๊ฒƒ์ด ํšจ๊ณผ์ โ€์ด๋ผ๊ณ  ์กฐ์–ธํ–ˆ๋‹ค.

4. ์„ฑ๊ณผ๋ฅผ ์ธก์ •ํ•ด AI ํˆฌ์ž ์ˆ˜์ต์„ ๊ฒ€์ฆํ•œ๋‹ค

ROI๋ฅผ ํ‰๊ฐ€ํ•˜๋ ค๋ฉด CIO๋Š” AI ๋„์ž… ์ด์ „์˜ ๊ธฐ์ค€์„ ์„ ์„ค์ •ํ•˜๊ณ , ์ดˆ๊ธฐ์— ๋ช…ํ™•ํ•œ ๋ฒค์น˜๋งˆํฌ๋ฅผ ์„ธ์›Œ์•ผ ํ•œ๋‹ค. ๋งŽ์€ IT ๋ฆฌ๋”๊ฐ€ ๊ฐ€์น˜ ์‹คํ˜„๊นŒ์ง€ ๊ฑธ๋ฆฌ๋Š” ์‹œ๊ฐ„, ๋น„์šฉ ์ ˆ๊ฐ, ์‹œ๊ฐ„ ์ ˆ๊ฐ, ์‚ฌ๋žŒ ์ง์›์ด ์ฒ˜๋ฆฌํ•˜๋Š” ์—…๋ฌด๋Ÿ‰, AI๋กœ ์ƒˆ๋กœ ์ฐฝ์ถœ๋œ ๋งค์ถœ ๊ธฐํšŒ ๊ฐ™์€ ์ง€ํ‘œ์— ์ฑ…์ž„์ž๋ฅผ ์ง€์ •ํ•  ๊ฒƒ์„ ๊ถŒ๊ณ ํ•œ๋‹ค.

์•„์ŠคํŽ˜๋ฆฌํƒ€์Šค์˜ ํœ ๋Ÿฌ๋Š” โ€œAI ํ”„๋กœ์ ํŠธ๋ฅผ ์‹œ์ž‘ํ•˜๊ธฐ ์ „์— ๊ธฐ์ค€ ์ธก์ •๊ฐ’์„ ๋ฐ˜๋“œ์‹œ ์„ค์ •ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๋ฉฐ, ๊ฐ ์‚ฌ์—… ๋ถ€๋ฌธ์˜ ์˜ˆ์ธก ์ง€ํ‘œ๋ฅผ ๊ฒฝ์˜์ง„์˜ ์ •๊ธฐ ์„ฑ๊ณผ ๋ฆฌ๋ทฐ์— ํฌํ•จ์‹œํ‚ค๋ผ๊ณ  ์กฐ์–ธํ–ˆ๋‹ค. ํœ ๋Ÿฌ๋Š” ๋งŽ์€ ์กฐ์ง์ด ๋ชจ๋ธ ์ •ํ™•๋„, ์ง€์—ฐ ์‹œ๊ฐ„, ์ •๋ฐ€๋„ ๊ฐ™์€ ๊ธฐ์ˆ  ์ง€ํ‘œ๋งŒ ์ธก์ •ํ•˜๊ณ , ์ด ์ˆ˜์น˜๋ฅผ ๋น„์šฉ ์ ˆ๊ฐ, ๋งค์ถœ ์ฆ๊ฐ€, ๋ฆฌ์Šคํฌ ๊ฐ์†Œ ๊ฐ™์€ ๋น„์ฆˆ๋‹ˆ์Šค ์„ฑ๊ณผ์™€ ์—ฐ๊ฒฐํ•˜์ง€ ๋ชปํ•˜๋Š” ์‹ค์ˆ˜๋ฅผ ์ €์ง€๋ฅธ๋‹ค๊ณ  ์ง€์ ํ–ˆ๋‹ค.

๊ทธ๋ž˜์„œ ๋‹ค์Œ ๋‹จ๊ณ„๋Š” ์‹ค์งˆ์  ๊ฐ€์น˜๋ฅผ ์ž…์ฆํ•  ์ˆ˜ ์žˆ๋Š” ๋ช…ํ™•ํ•˜๊ณ  ์ธก์ • ๊ฐ€๋Šฅํ•œ ๋ชฉํ‘œ๋ฅผ ์„ธ์šฐ๋Š” ๊ฒƒ์ด๋‹ค. CMIT ์†”๋ฃจ์…˜์˜ ๋กœํŽ˜์ฆˆ๋Š” โ€œํ”„๋กœ์ ํŠธ ์ดˆ๊ธฐ ๋‹จ๊ณ„๋ถ€ํ„ฐ ์ธก์ • ํ•ญ๋ชฉ์„ ์„ค๊ณ„ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ๋กœํŽ˜์ฆˆ๋Š” CIO๊ฐ€ ๊ฐ AI ํ”„๋กœ์ ํŠธ๋งˆ๋‹ค โ€˜์ฒ˜๋ฆฌ ์†๋„ 20% ๊ฐœ์„ โ€™, โ€˜๊ณ ๊ฐ ๋งŒ์กฑ๋„ 15% ์ƒ์Šนโ€™์ฒ˜๋Ÿผ ๊ตฌ์ฒด์ ์ธ KPI๋ฅผ ์ •์˜ํ•ด์•ผ ํ•œ๋‹ค๊ณ  ์กฐ์–ธํ–ˆ๋‹ค. ๋˜, ๋น ๋ฅด๊ณ  ์ •๋Ÿ‰ํ™” ๊ฐ€๋Šฅํ•œ ๊ฒฐ๊ณผ๋ฅผ ๋‚ผ ์ˆ˜ ์žˆ๋Š” ์†Œ๊ทœ๋ชจ ํŒŒ์ผ๋Ÿฟ๋ถ€ํ„ฐ ์‹œ์ž‘ํ•˜๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

ํ•œ ๊ฐ€์ง€ ๋ช…ํ™•ํ•œ ์ธก์ • ์ง€ํ‘œ๋Š” ์‹œ๊ฐ„ ์ ˆ๊ฐ์ด๋‹ค. ์†Œํ”„ํŠธ์›จ์–ด ๊ธฐ๋ฐ˜ ์„œ๋น„์Šค ํšŒ์‚ฌ ๋ ˆ๋ชฌ๊ทธ๋ž˜์Šค(Lemongrass)์˜ CTO ์—์ด๋จผ ์˜ค๋‹์€ ๊ณ ๊ฐ์‚ฌ๊ฐ€ SAP ๊ฐœ๋ฐœ ๋ฌธ์„œ๋ฅผ ์ˆ˜์ž‘์—…์œผ๋กœ ์ž‘์„ฑํ•˜๋Š” ์žฅ๋ฉด์„ ์—ฌ๋Ÿฌ ๋ฒˆ ๋ดค๋Š”๋ฐ, ์ด ์ž‘์—…์€ ์—„์ฒญ๋‚œ ์‹œ๊ฐ„์ด ์†Œ์š”๋˜๋Š” ๊ณผ์ •์ด๋‹ค. ์˜ค๋‹์€ โ€œ๋ฌธ์„œ ์ž‘์„ฑ์„ ์ƒ์„ฑํ˜• AI๋กœ ์ฒ˜๋ฆฌํ•˜๋ฉด ์‚ฌ๋žŒ์˜ ํˆฌ์ž… ์‹œ๊ฐ„์„ ๋ถ„๋ช…ํ•˜๊ฒŒ ์ค„์ผ ์ˆ˜ ์žˆ๊ณ , ์ด ์ ˆ๊ฐ ํšจ๊ณผ๋ฅผ ๋งค์šฐ ๊ฐ„๋‹จํ•˜๊ฒŒ ๋‹ฌ๋Ÿฌ ๊ธฐ์ค€ ROI๋กœ ํ™˜์‚ฐํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์—…๋ฌด๋‹น ํˆฌ์ž…๋˜๋Š” ๋…ธ๋™๋ ฅ์˜ ๊ฐ์†Œ๋„ ์ค‘์š”ํ•œ ์‹ ํ˜ธ๋‹ค. ํ’€์Šคํƒ ๊ธฐ์ˆ  ์„œ๋น„์Šค ํšŒ์‚ฌ TEK์‹œ์Šคํ…œ์ฆˆ(TEKsystems)์˜ CTO ๋žจ ํŒ”๋ผ๋‹ˆ์•„ํŒ์€ โ€œ๋ชฉํ‘œ๊ฐ€ ์ƒ๋‹ด์›์ด ์ฒ˜๋ฆฌํ•˜๋Š” ์ฝœ์„ผํ„ฐ ๋ฌธ์˜ ๊ฑด์ˆ˜๋ฅผ ์ค„์ด๋Š” ๊ฒƒ์ด๋ผ๋ฉด, ์ด ์ˆ˜์น˜๋ฅผ ๋ช…ํ™•ํ•œ ์ง€ํ‘œ๋กœ ์ •ํ•˜๊ณ  ์‹ค์‹œ๊ฐ„์œผ๋กœ ์ถ”์ ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๋ฉฐ, AI ๋„์ž… ๊ณผ์ •์—์„œ ์ƒˆ๋กœ์šด ๋งค์ถœ ๊ธฐํšŒ๊ฐ€ ์ƒ๊ธธ ๊ฐ€๋Šฅ์„ฑ๋„ ํฌ๋‹ค๊ณ  ๋ง๋ถ™์˜€๋‹ค.

์ผ๋ถ€ CIO๋Š” ๊ฐœ๋ณ„ ์‚ฌ์šฉ๋ก€ ๋ณ„๋กœ ์„ธ๋ถ„ํ™”๋œ KPI๋ฅผ ๋ชจ๋‹ˆํ„ฐ๋งํ•˜๋ฉฐ, ๊ฒฐ๊ณผ์— ๋”ฐ๋ผ ์ „๋žต์„ ์กฐ์ •ํ•œ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด ์•„์‚ฌ๋‚˜์˜ ์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” ๊ฐœ๋ฐœ ํšจ์œจ์„ฑ์„ ๋ชจ๋‹ˆํ„ฐ๋งํ•  ๋•Œ ์‚ฌ์ดํด ํƒ€์ž„, ์ฒ˜๋ฆฌ๋Ÿ‰, ํ’ˆ์งˆ, ํŠธ๋žœ์žญ์…˜๋‹น ๋น„์šฉ, ๋ฆฌ์Šคํฌ ์ด๋ฒคํŠธ ๋ฐœ์ƒ ๊ฑด์ˆ˜๋ฅผ ์ถ”์ ํ•œ๋‹ค. ๋˜ ์—์ด์ „ํŠธ ์ง€์› ์‹คํ–‰ ๋น„์œจ, ํ™œ์„ฑ ์‚ฌ์šฉ์ž, ํœด๋จผ ์ธ ๋” ๋ฃจํ”„ ์Šน์ธ ๋น„์œจ, ์˜ˆ์™ธ ์ƒํ™ฉ ์—์Šค์ปฌ๋ ˆ์ด์…˜ ๋น„์œจ๋„ ํ•จ๊ป˜ ๋ณธ๋‹ค. ์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” ์ด๋Ÿฐ ๋ฐ์ดํ„ฐ๋ฅผ ๊ฒ€ํ† ํ•˜๋ฉด ํ”„๋กฌํ”„ํŠธ์™€ ๊ฐ€๋“œ๋ ˆ์ผ์„ ์‹ค์‹œ๊ฐ„์œผ๋กœ ์กฐ์ •ํ•˜๋Š” ๋ฐ ๋„์›€์ด ๋œ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

ํ•ต์‹ฌ์€ ์ดˆ๊ธฐ๋ถ€ํ„ฐ ์ธก์ • ์ง€ํ‘œ๋ฅผ ์„ค์ •ํ•˜๊ณ , ์‹ ํ˜ธ๋‚˜ ์„ฑ๊ณผ๋ฅผ ์ถ”์ ํ•˜์ง€ ์•Š๋Š” ์•ˆํ‹ฐ ํŒจํ„ด์— ๋น ์ง€์ง€ ์•Š๋Š” ๊ฒƒ์ด๋‹ค. ์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” โ€œ์ธก์ •์€ ์ข…์ข… ํ”„๋กœ์ ํŠธ ํ›„๋ฐ˜์— ๋’ค๋Šฆ๊ฒŒ ๋ถ™๋Š” ๋ฐ”๋žŒ์— ๋ฆฌ๋”๊ฐ€ ๊ฐ€์น˜๋ฅผ ์ž…์ฆํ•˜์ง€ ๋ชปํ•˜๊ณ , ์–ด๋–ค ๊ฒƒ์„ ํ™•์žฅํ•ด์•ผ ํ•˜๋Š”์ง€๋„ ๊ฒฐ์ •ํ•˜์ง€ ๋ชปํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๋˜, โ€œ์ฒ˜์Œ๋ถ€ํ„ฐ ๋ช…ํ™•ํ•œ ํ•ต์‹ฌ ๋ฏธ์…˜ ์ง€ํ‘œ๋ฅผ ์ •ํ•˜๊ณ  ๊ธฐ์ค€์„ ์„ ์„ธ์šด ๋‹ค์Œ, AI๋ฅผ ์—…๋ฌด ํ๋ฆ„ ์†์— ์ง์ ‘ ๋…น์—ฌ ๋„ฃ์–ด์•ผ ์ง์›์ด ๋” ์ค‘์š”ํ•œ ํŒ๋‹จ์ด ํ•„์š”ํ•œ ์ผ์— ์ง‘์ค‘ํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

5. AI ๋ฌธํ™”๋ฅผ ๊ฑฐ๋ฒ„๋„Œ์Šค๋กœ ๊ด€๋ฆฌํ•ด ๋ณด์•ˆ ์‚ฌ๊ณ ์™€ ๋ถˆ์•ˆ์ •์„ ๋ง‰๋Š”๋‹ค

์ƒ์„ฑํ˜• AI ๋„๊ตฌ๋Š” ์ด์ œ ์—…๋ฌด ํ˜„์žฅ์—์„œ ํ”ํ•˜๊ฒŒ ์“ฐ์ด์ง€๋งŒ, ์—ฌ์ „ํžˆ ์ƒ๋‹น์ˆ˜ ์ง์›์€ ์ด๋ฅผ ์•ˆ์ „ํ•˜๊ฒŒ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ๋ฒ•์„ ๋ชจ๋ฅธ๋‹ค. ์Šค๋ชฐPDF(SmallPDF)์˜ 2025๋…„ ์กฐ์‚ฌ์— ๋”ฐ๋ฅด๋ฉด, ๋ฏธ๊ตญ ๊ธฐ๋ฐ˜ ์ง์›์˜ ๊ฑฐ์˜ 1/5๋Š” AI ๋„๊ตฌ์— ๋กœ๊ทธ์ธ ์ž๊ฒฉ ์ฆ๋ช…์„ ์ž…๋ ฅํ•œ ๊ฒฝํ—˜์ด ์žˆ์—ˆ๋‹ค. ๋กœํŽ˜์ฆˆ๋Š” โ€œ์ข‹์€ ๋ฆฌ๋”์‹ญ์€ ๊ฑฐ๋ฒ„๋„Œ์Šค์™€ ๊ฐ€๋“œ๋ ˆ์ผ์„ ์„ธ์šฐ๋Š” ๊ฒƒ์—์„œ ์‹œ์ž‘๋œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ์—ฌ๊ธฐ์—๋Š” ์ฑ—GPT ๊ฐ™์€ ๋„๊ตฌ์— ๋ฏผ๊ฐํ•œ ๋น„๋ฐ€ ๋ ˆ์‹œํ”ผ ๋ฐ์ดํ„ฐ๊ฐ€ ์ž…๋ ฅ๋˜์ง€ ์•Š๋„๋ก ํ•˜๋Š” ์ •์ฑ… ์ˆ˜๋ฆฝ๋„ ํฌํ•จ๋œ๋‹ค.

AI๋ฅผ ๋งŽ์ด ์“ธ์ˆ˜๋ก ๊ธฐ์—…์˜ ๊ณต๊ฒฉ ํ‘œ๋ฉด๋„ ๋„“์–ด์ง„๋‹ค. ๊ฒฝ์˜์ง„์€ AI ๊ธฐ๋ฐ˜ ๋ธŒ๋ผ์šฐ์ €์˜ ๋ณด์•ˆ ์ทจ์•ฝ์ , ์„€๋„์šฐ AI ์‚ฌ์šฉ, LLM์˜ ํ™˜๊ฐ ๋ฌธ์ œ๋ฅผ ์ง„์ง€ํ•˜๊ฒŒ ๊ณ ๋ คํ•ด์•ผ ํ•œ๋‹ค. ์—์ด์ „ํŠธํ˜• AI๊ฐ€ ๋น„์ฆˆ๋‹ˆ์Šค ํ•ต์‹ฌ ํ”„๋กœ์„ธ์Šค์— ๊นŠ์ด ๊ด€์—ฌํ• ์ˆ˜๋ก, ์ œ๋Œ€๋กœ ๋œ ๊ถŒํ•œ ๊ด€๋ฆฌ์™€ ์ ‘๊ทผ ์ œ์–ด ์—†์ด๋Š” ๋ฏผ๊ฐ ๋ฐ์ดํ„ฐ ๋…ธ์ถœ์ด๋‚˜ IT ์‹œ์Šคํ…œ์— ๋Œ€ํ•œ ์•…์˜์  ์นจํˆฌ ์œ„ํ—˜์ด ์ปค์ง„๋‹ค.

์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ ๊ด€์ ์—์„œ ๋ณด๋ฉด, AI ์ฝ”๋”ฉ ์—์ด์ „ํŠธ๋ฅผ ํ†ตํ•ด ๋น„๋ฐ€๋ฒˆํ˜ธ๋‚˜ ํ‚ค, ํ† ํฐ์ด ์œ ์ถœ๋  ๊ฐ€๋Šฅ์„ฑ๋„ ๋งค์šฐ ํฌ๋‹ค. ๊ฐœ๋ฐœ์ž๋Š” ์™ธ๋ถ€ ๋ฐ์ดํ„ฐ๋‚˜ ๋„๊ตฌ, API์— ์ ‘๊ทผํ•˜๋„๋ก MCP ์„œ๋ฒ„๋ฅผ ์‚ฌ์šฉํ•ด AI ์ฝ”๋”ฉ ์—์ด์ „ํŠธ๋ฅผ ๊ฐ•ํ™”ํ•ด ์™”๋‹ค. ๊ทธ๋Ÿฌ๋‚˜ ์›”๋žŒ(Wallarm) ์กฐ์‚ฌ์— ๋”ฐ๋ฅด๋ฉด, 2025๋…„ 2~3๋ถ„๊ธฐ MCP ๊ด€๋ จ ์ทจ์•ฝ์ ์ด 270% ์ฆ๊ฐ€ํ–ˆ๊ณ , ๋™์‹œ์— API ์ทจ์•ฝ์ ๋„ ๊ธ‰์ฆํ–ˆ๋‹ค.

์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” ๊ธฐ์—…์ด AI๋ฅผ ๋„์ž…ํ•  ๋•Œ ์—์ด์ „ํŠธ ID, ๊ถŒํ•œ, ๊ฐ์‚ฌ ์ด๋ ฅ์„ ์†Œํ™€ํžˆ ํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค๊ณ  ์ง€์ ํ•˜๋ฉฐ, โ€œ์—์ด์ „ํŠธ ID ๋ฐ ์ ‘๊ทผ ๊ด€๋ฆฌ๋ฅผ ๋„์ž…ํ•ด, ์—์ด์ „ํŠธ๊ฐ€ ์‚ฌ๋žŒ๊ณผ ๋™์ผํ•œ ๊ถŒํ•œ๊ณผ ๊ฐ์‚ฌ ๊ฐ€๋Šฅ์„ฑ์„ ๊ฐ–๋„๋ก ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์กฐ์–ธํ–ˆ๋‹ค. ์—ฌ๊ธฐ์—๋Š” ๋กœ๊ทธ ๊ธฐ๋ก๊ณผ ์Šน์ธ ์ ˆ์ฐจ๋„ ํฌํ•จ๋œ๋‹ค.

์œ„ํ—˜์ด ์ด๋ ‡๊ฒŒ ํฐ๋ฐ๋„ ๊ด€๋ฆฌ ์ฒด๊ณ„๋Š” ์—ฌ์ „ํžˆ ํ—ˆ์ˆ ํ•œ ๊ณณ์ด ๋งŽ๋‹ค. ์˜ค๋”ง๋ณด๋“œ(AuditBoard)์˜ ๋ณด๊ณ ์„œ์— ๋”ฐ๋ฅด๋ฉด, AI๋ฅผ ๋„์ž… ์ค‘์ธ ์กฐ์ง ๋น„์ค‘์€ 82%์— ์ด๋ฅด์ง€๋งŒ, ๊ฑฐ๋ฒ„๋„Œ์Šค ํ”„๋กœ๊ทธ๋žจ์„ ์™„์ „ํžˆ ๊ตฌํ˜„ํ•œ ๊ณณ์€ 25%์— ๋ถˆ๊ณผํ•˜๋‹ค. IBM ๋ถ„์„์— ๋”ฐ๋ฅด๋ฉด, ๋ฐ์ดํ„ฐ ์œ ์ถœ 1๊ฑด๋‹น ํ‰๊ท  ํ”ผํ•ด์•ก์€ ๊ฑฐ์˜ 450๋งŒ ๋‹ฌ๋Ÿฌ์— ์ด๋ฅด๋ฉฐ, IDC๋Š” โ€˜์‹ ๋ขฐํ•  ์ˆ˜ ์žˆ๋Š” AIโ€™๋ฅผ ๊ตฌ์ถ•ํ•œ ์กฐ์ง์ด ๊ทธ๋ ‡์ง€ ์•Š์€ ์กฐ์ง๋ณด๋‹ค AI ํ”„๋กœ์ ํŠธ ROI๊ฐ€ 2๋ฐฐ ์ด์ƒ ๋†’์„ ๊ฐ€๋Šฅ์„ฑ์ด 60% ๋” ํฌ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. AI ๊ฑฐ๋ฒ„๋„Œ์Šค์— ํˆฌ์žํ•ด์•ผ ํ•˜๋Š” ๋น„์ฆˆ๋‹ˆ์Šค ๋…ผ๋ฆฌ๋Š” ๋”ํ•  ๋‚˜์œ„ ์—†์ด ๋ถ„๋ช…ํ•˜๋‹ค.

์Šค๋ฆฌ๋ฐ”์Šคํƒ€๋ฐ”๋Š” โ€œ๋†’์€ ๋ชฉํ‘œ ์˜์‹์— ๊ฐ•๋ ฅํ•œ ๊ฐ€๋“œ๋ ˆ์ผ์„ ์ง์ง€์–ด์•ผ ํ•œ๋‹คโ€๋ผ๋ฉฐ โ€œ๋ฐ์ดํ„ฐ ์ˆ˜๋ช… ์ฃผ๊ธฐ์™€ ์ ‘๊ทผ ์ œ์–ด๋ฅผ ๋ช…ํ™•ํžˆ ํ•˜๊ณ , ํ‰๊ฐ€์™€ ๋ ˆ๋“œํŒ€, ๊ทธ๋ฆฌ๊ณ  ์œ„ํ—˜์ด ํฐ ๊ตฌ๊ฐ„์—๋Š” ํœด๋จผ ์ธ ๋” ๋ฃจํ”„ ๊ฒ€์ฆ ์ ˆ์ฐจ๋ฅผ ๋‘์–ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ๋˜ โ€œ๋ณด์•ˆ๊ณผ ํ”„๋ผ์ด๋ฒ„์‹œ, ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค๋ฅผ ์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ ๋ผ์ดํ”„์‚ฌ์ดํด์— ๋…น์—ฌ ๋ฐฐํฌ์™€ ๋ณด์•ˆ์„ ๋™์‹œ์— ์ถ”์ง„ํ•ด์•ผ ํ•œ๋‹ค. ๋ฐ์ดํ„ฐ ๊ณ„๋ณด๋‚˜ ๋ชจ๋ธ ๋™์ž‘์„ ์•Œ ์ˆ˜ ์—†๋Š” ๋ธ”๋ž™๋ฐ•์Šค๋ฅผ ํ—ˆ์šฉํ•ด์„œ๋Š” ์•ˆ ๋œ๋‹คโ€๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

AI๋Š” ๋งˆ๋ฒ•์ด ์•„๋‹ˆ๋‹ค

BCG์— ๋”ฐ๋ฅด๋ฉด, ๊ธฐ์—… ๊ฐ€์šด๋ฐ 22%๋งŒ์ด AI๋ฅผ ๊ฐœ๋… ์ฆ๋ช… ๋‹จ๊ณ„ ์ด์ƒ์œผ๋กœ ์ง„์ฒ™์‹œ์ผฐ๊ณ , 4%๋งŒ์ด ์ƒ๋‹นํ•œ ๊ฐ€์น˜๋ฅผ ์ฐฝ์ถœํ•˜๊ณ  ์žˆ๋‹ค. ์ด๋Ÿฐ ๋ƒ‰์ •ํ•œ ํ†ต๊ณ„๋ฅผ ๊ฐ์•ˆํ•˜๋ฉด, CIO๋Š” AI ํˆฌ์ž ์ˆ˜์ต์— ๋Œ€ํ•ด ๋น„ํ˜„์‹ค์ ์ธ ๊ธฐ๋Œ€๋ฅผ ์„ธ์›Œ์„œ๋Š” ์•ˆ ๋œ๋‹ค.

AI์—์„œ ์˜๋ฏธ ์žˆ๋Š” ROI๋ฅผ ์–ป์œผ๋ ค๋ฉด ์ƒ๋‹นํ•œ ์ดˆ๊ธฐ ๋…ธ๋ ฅ์ด ํ•„์š”ํ•˜๋ฉฐ, ์กฐ์ง ํ”„๋กœ์„ธ์Šค๋ฅผ ๊ทผ๋ณธ์ ์œผ๋กœ ๋ฐ”๊พธ๋Š” ์ž‘์—…์ด ๋’ค๋”ฐ๋ผ์•ผ ํ•œ๋‹ค. ๋งˆ์Šคํ„ฐ์นด๋“œ์˜ ์šด์˜ CTO ์กฐ์ง€ ๋งˆ๋‹ฌ๋กœ๋‹ˆ๋Š” ๋Ÿฐํƒ€์ž„(Runtime)๊ณผ์˜ ์ตœ๊ทผ ์ธํ„ฐ๋ทฐ์—์„œ ์ƒ์„ฑํ˜• AI ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๋„์ž…์€ ๋ณธ์งˆ์ ์œผ๋กœ ๋ณ€ํ™” ๊ด€๋ฆฌ์™€ ์ฑ„ํƒ์˜ ๋ฌธ์ œ๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

AI์—๋Š” ํ•จ์ •์ด ๋์—†์ด ๋งŽ๊ณ , ์กฐ์ง์ด ๊ฐ€์น˜๋ฅผ ๋”ฐ์ง€๊ธฐ๋ณด๋‹ค ์œ ํ–‰์„ ์ข‡๋Š” ๊ฒฝ์šฐ๋„ ํ”ํ•˜๋‹ค. ๋ฐ์ดํ„ฐ ์ „๋žต ์—†์ด ์„ฑ๊ธ‰ํžˆ ํ”„๋กœ์ ํŠธ๋ฅผ ์‹œ์ž‘ํ•˜๊ฑฐ๋‚˜ ๋„ˆ๋ฌด ๋นจ๋ฆฌ ํ™•์žฅํ•˜๊ฑฐ๋‚˜ ๋ณด์•ˆ์„ ์‚ฌํ›„์— ๋ถ™์ด๋Š” ๊ฒฝ์šฐ๋„ ๋งŽ๋‹ค. ๋งŽ์€ AI ํ”„๋กœ๊ทธ๋žจ์ด ๋ชฉํ‘œํ•œ ์ˆ˜์ค€์— ๋„๋‹ฌํ•˜์ง€ ๋ชปํ•˜๋Š” ์ด์œ ๋Š” ์ตœ๊ณ  ๊ฒฝ์˜์ง„์˜ ํ›„์›์ด๋‚˜ ๊ฑฐ๋ฒ„๋„Œ์Šค๊ฐ€ ๋ถ€์กฑํ•˜๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค. ๋ฐ˜๋Œ€๋กœ, ์†”๋ฃจ์…˜ ์—…์ฒด์˜ ํ™๋ณด๋ฅผ ๊ณง์ด๊ณง๋Œ€๋กœ ๋ฏฟ๊ณ  ๊ณผ๋„ํ•˜๊ฒŒ ์ง€์ถœํ•˜๊ฑฐ๋‚˜ AI ํ”Œ๋žซํผ์„ ๊ธฐ์กด ๋ ˆ๊ฑฐ์‹œ ์ธํ”„๋ผ์™€ ํ†ตํ•ฉํ•˜๋Š” ๋‚œ์ด๋„๋ฅผ ๊ณผ์†Œํ‰๊ฐ€ํ•˜๋Š” ๊ฒฝ์šฐ๋„ ๋งŽ๋‹ค.

์•ž์œผ๋กœ AI์˜ ๋น„์ฆˆ๋‹ˆ์Šค ์˜ํ–ฅ์„ ๊ทน๋Œ€ํ™”ํ•˜๋ ค๋ฉด, ํ™•์žฅ์„ ๋’ท๋ฐ›์นจํ•  ๋ฐ์ดํ„ฐ ์ธํ”„๋ผ์™€ ํ”Œ๋žซํผ ์—ญ๋Ÿ‰์— ํˆฌ์žํ•˜๊ณ , ์‚ฌ๋žŒ์˜ ๋ฐ˜๋ณต ์ž‘์—…์„ ์ค„์ด๊ณ  ๋งค์ถœ์ด๋‚˜ ํšจ์œจ์„ ๋ถ„๋ช…ํ•˜๊ฒŒ ๋Œ์–ด์˜ฌ๋ฆด ์ˆ˜ ์žˆ๋Š” ์†Œ์ˆ˜์˜ ํ•ต์‹ฌ ์‚ฌ์šฉ๋ก€์— ์ง‘์ค‘ํ•ด์•ผ ํ•œ๋‹ค.

AI ์—ด๊ธฐ๋ฅผ ํ•ต์‹ฌ ์›์น™์— ๋‹ค์‹œ ์—ฐ๊ฒฐํ•˜๊ณ , ์กฐ์ง์ด ์ถ”๊ตฌํ•˜๋Š” ๋น„์ฆˆ๋‹ˆ์Šค ์ „๋žต์„ ๋ช…ํ™•ํžˆ ์ดํ•ดํ•˜๋Š” ์ž‘์—…์ด ์žˆ์–ด์•ผ๋งŒ ROI์— ํ•œ ๊ฑธ์Œ์”ฉ ๋‹ค๊ฐ€๊ฐˆ ์ˆ˜ ์žˆ๋‹ค. ํƒ„ํƒ„ํ•œ ๋ฆฌ๋”์‹ญ๊ณผ ๋ถ„๋ช…ํ•œ ๋ชฉํ‘œ ์—†์ด AI์—๋งŒ ๊ธฐ๋Œ€๋ฉด, AI๋Š” ์†์— ์žกํž ๋“ฏ ์žกํžˆ์ง€ ์•Š๋Š” ๋ณด์ƒ์„ ์•ฝ์†ํ•˜๋Š” ํฅ๋ฏธ๋กœ์šด ๊ธฐ์ˆ ์— ๊ทธ์น  ๋ฟ์ด๋‹ค.
dl-ciokorea@foundryco.com

A CIOโ€™s 5-point checklist to drive positive AI ROI

26 November 2025 at 05:00

Earlier this year, MIT made headlines with a report that found 95% of organizations are getting no return from AI โ€” and this despite a groundbreaking $30 billion investment, or more, into US-based internal gen AI initiatives. So why do so many AI initiatives fail to deliver positive ROI? Because they often lack a clear connection to business value, says Neal Ramasamy, global CIO at Cognizant, an IT consulting firm. โ€œThis leads to projects that are technically impressive but donโ€™t solve a real need or create a tangible benefit,โ€ he says.

Technologists often follow the hype, diving headfirst into AI tests without considering business results. โ€œMany start with models and pilots rather than business outcomes,โ€ says Saket Srivastava, CIO of Asana, the project management application. โ€œTeams run demos in isolation, without redesigning the underlying workflow or assigning a profit and loss owner.โ€

A combination of a lack of upfront product thinking, poor underlying data practices, nonexistent governance, and minimal cultural incentives to adopt AI can produce negative results. So to avoid poor outcomes, many of the techniques boil down to better change management. โ€œWithout process change, AI speeds todayโ€™s inefficiencies,โ€ adds Srivastava.

Here, we review five tips to manage change within an organization that CIOs can put into practice today. By following this checklist, enterprises should start to turn the tide on negative AI ROI, learn from anti-patterns, and discover which sort of metrics validate successful company-wide AI ventures.

1. Align leadership upfront by communicating business goals and stewarding the AI initiative

AI initiatives require executive sponsorship and a clear vision for how they improve the business. โ€œStrong leadership is essential to translate AI investments into results,โ€ says Adam Lopez, president and leadโ€ฏvCIO at managed IT support provider CMIT Solutions. โ€œExecutive sponsorship and oversight of AI programs, ideally at the CEO or board level, correlates with higher ROI.โ€

For example, at IT services and consulting company Xebia, a subgroup of executives steers its internal AI efforts. Chaired by global CIO Smit Shanker, the team includes the global CFO, head of AI and automation, head of IT infrastructure and security, and head of business operations.

Once upper leadership is assembled, accountability becomes critical. โ€œStart by assigning business ownership,โ€ advises Srivastava. โ€œEvery AI use case needs an accountable leader with a target tied to objectives and key results.โ€ He recommends standing up a cross-functional PMO to define lighthouse use cases, set success targets, enforce guardrails, and regularly communicate progress.

Still, even with leadership in place, many employees will need hands-on guidance to apply AI in their daily work. โ€œFor most individuals, even if you give them the tools in the morning, they donโ€™t know where to start,โ€ says Orla Daly, CIO of Skillsoft, a learning management system. She recommends identifying champions across the organization who can surface meaningful use cases and share practical tips, such as how to get more out of tools like Copilot. Those with a curiosity and a willingness to learn will make the most headway, she says.

Finally, executives must invest in infrastructure, talent, and training. โ€œLeaders must champion a data-driven culture and promote a clear vision for how AI will solve business problems,โ€ says Cognizantโ€™s Ramasamy. This requires close collaboration between business leaders, data scientists, and IT to execute and measure pilot projects before scaling.

2. Evolve by shifting the talent framework and investing in upskilling

Organizations must be open to shift their talent framework and redesign roles. โ€œCIOs should adapt their talent and management strategies to ensure successful AI adoption and ROI for the organization,โ€ says Ramasamy. โ€œThis could involve creating new roles and career paths for AI-focused professionals, such as data scientists and prompt engineers, while upskilling existing employees.โ€

CIOs should also view talent as a cornerstone of any AI strategy, adds CMITโ€™s Lopez. โ€œBy investing in people through training, communication, and new specialist roles, CIOs can be assured that employees will embrace AI tools and drive success.โ€ He adds that internal hackathons and training sessions often yield noticeable boosts in skills and confidence.

Upskilling, for instance, should meet employees where they are, so Asanaโ€™s Srivastava recommends tiered paths: all staff need basic prompt literacy and safety training, while power users require deeper workflow design and agent-building knowledge. โ€œWe took the approach of surveying the workforce, targeting enablement, and remeasuring to confirm that maturity moved in the right direction,โ€ he says.

But assessing todayโ€™s talent framework goes beyond human skillsets. It also means reassessing your work to be done, and who or what performs what tasks. โ€œItโ€™s essential to review business processes for opportunities to refactor them, given the new capabilities that AI brings,โ€ says Scott Wheeler, cloud practice lead at cloud consulting firm Asperitas Consulting.

For Skillsoftโ€™s Daly, todayโ€™s AI age necessitates a modern talent management framework that artfully balances the four Bs: build, buy, borrow, and bots. In other words, leaders should view their organization as a collection of skills rather than fixed roles, and apply the right mix of in-house staff, software, partners, or automation as needed. โ€œItโ€™s requiring us to break things down into jobs or tasks to be done, and looking at your work in a more fragmented way,โ€ says Daly.

For instance, her team used GitHub Copilot to quickly code a learning portal for a certain customer. The project highlighted how pairing human developers with AI assistants can dramatically accelerate delivery, raising new questions about what skills other developers need to be equally productive and efficient.

But as AI agents take over more routine work, leaders must dispel fears that AI will replace jobs outright. โ€œCommunicating the why behind AI initiatives can alleviate fears and demonstrate how these tools can augment human roles,โ€ says Ramasamy. Srivastava agrees. โ€œThe throughline is trust,โ€ he says, โ€œShow people how AI removes toil and increases impact; keep humans in the decision loop and adoption will follow.โ€

3. Adapt organizational processes to fully capture AI benefitsย 

Shifting the talent framework is only the beginning. Organizations must also reengineer core processes. โ€œFully unlocking AIโ€™s value often requires reengineering how the organization works,โ€ says CMITโ€™s Lopez, who urges embedding AI into day-to-day operations and supporting it with continual experimentation rather than treating it as a static add-on.

To this end, one necessary adaptation is toward treating internal AI-driven workflows like products and codifying patterns across the organization, says Srivastava. โ€œEstablish productโ€‘management rigor for intake, prioritization, and roadmapping of AI use cases, with clear owners, problem statements, and value hypotheses,โ€ he says.

At Xebia, a governance board oversees this rigor through a three-stage tollgate process of identifying and assessing value, securing business acceptance, and then handing off to IT for monitoring and support. โ€œA core group is responsible for organizational and functional simplification with each use case,โ€ says Shanker. โ€œThat encourages cross-functional processes and helps break down silos.โ€

Similarly for Ramasamy, the biggest hurdle is organizational resistance. โ€œMany companies underestimate the change management required for successful adoption,โ€ he says. โ€œThe most critical shift is moving from siloed decision-making to a data-centric approach. Business processes should integrate AI outputs seamlessly, automating tasks and empowering employees with data-driven insights.โ€

Identifying the right areas to automate also depends on visibility. โ€œThis is where most companies fall down because they donโ€™t have good, documented processes,โ€ says Skillsoftโ€™s Daly. She recommends enlisting subject-matter experts across business lines to examine workflows for optimization. โ€œItโ€™s important to nominate individuals within the business to ask how to drive AI into your flow of work,โ€ she says.

Once you identify units of work common across functions that AI can streamline, the next step is to make them visible and standardize their application. Skillsoft is doing this through an agent registry that documents agentic capabilities, guardrails, and data management processes. โ€œWeโ€™re formalizing an enterprise AI framework in which ethics and governance are part of how we manage the portfolio of use cases,โ€ she adds.

Organizations should then anticipate roadblocks and create support structures to help users. โ€œOne strategy to achieve this is to have AI SWAT teams whose purpose is to facilitate adoption and remove obstacles,โ€ says Asperitasโ€™ Wheeler.

4. Measure progress to validate your returnย ย ย 

To evaluate ROI, CIOs must establish a pre-AI baseline and set benchmarks upfront. Leaders recommend assigning ownership around metrics such as time to value, cost savings, time savings, work handled by human agents, and new revenue opportunities generated.

โ€œBaseline measurements should be established before initiating AI projects,โ€ says Wheeler, who advises integrating predictive indicators from individual business units into leadershipโ€™s regular performance reviews. A common fault, he says, is only measuring technical KPIs like model accuracy, latency, or precision, and failing to link these to business outcomes, such as savings, revenue, or risk reduction.

Therefore, the next step is to define clear, measurable goals that demonstrate tangible value. โ€œBuild measurement into projects from day one,โ€ says CMITโ€™s Lopez. โ€œCIOs should define a set of relevant KPIs for each AI initiative. For example, 20% faster processing time or a 15% boost in customer satisfaction.โ€ Start with small pilots that yield quick, quantifiable results, he adds.

One clear measurement is time savings. For instance, Eamonn Oโ€™Neill, CTO at Lemongrass, a software-enabled services provider, shares how heโ€™s witnessed clients documenting SAP development manually, which can be an extremely time-intensive process. โ€œLeveraging generative AI to create this documentation provides a clear reduction in human effort, which can be measured and translated to a dollar ROI quite simply,โ€ he says.

Reduction of human labor per task is another key signal. โ€œIf the goal is to reduce the number of support desk calls handled by human agents, leaders should establish a clear metric and track it in real time,โ€ says Ram Palaniappan, CTO at full-stack tech services provider TEKsystems. He adds that new revenue opportunities may also surface through AI adoption.

Some CIOs are monitoring multiple granular KPIs across individual use cases and adjusting strategies based on results. Asanaโ€™s Srivastava, for instance, tracks engineering efficiency by monitoring cycle time, throughput, quality, cost per transaction, and risk events. He also measures the percentage of agent-assisted runs, active users, human-in-the-loop acceptance, and exception escalations. Reviewing this data, he says, helps tune prompts and guardrails in real time.

The resounding point is to set metrics early on, and not fall into the anti-patterns of not tracking signals or value gained. โ€œMeasurement is often bolted on late, so leaders canโ€™t prove value or decide what to scale,โ€ says Srivastava. โ€œThe remedy is to begin with a specific mission metric, baseline it, and embed AI directly in the flow of work so people can focus on higher-value judgment.โ€

5. Govern your AI culture to avoid breaches and instability

Gen AI tools are now commonplace, yet many employees still lack training to use them safely. For instance, nearly one in five US-based employees has entered login credentials into AI tools, according to a 2025 study from SmallPDF. โ€œGood leadership involves establishing governance and guardrails,โ€ says Lopez. That includes setting policies to prevent sensitive secret sauce data from being fed into tools like ChatGPT.

Heavy AI use also widens the enterprise attack surface. Leadership must now seriously consider things like security vulnerabilities in AI-driven browsers, shadow AI use, and LLM hallucinations. As agentic AI gets more involved in business-critical processes, proper authorization and access controls are essential to prevent exposure of sensitive data or malicious entry into IT systems.

From a software development standpoint, the potential for leaking passwords, keys, and tokens through AI coding agents is very real. Engineers have jumped at MCP servers to empower AI coding agents with access to external data, tools, and APIs, yet research from Wallarm found a 270% rise in MCP-related vulnerabilities from Q2 to Q3 2025, alongside surging API vulnerabilities.

Neglecting agent identity, permissions, and audit trails is a common trap that CIOs often stumble into with enterprise AI, says Srivastava. โ€œIntroduce agent identity and access management so agents inherit the same permissions and auditability as humans, including logging and approvals,โ€ he says.

Despite the risks, oversight remains weak. An AuditBoard report found that while 82% of organizations are deploying AI, only 25% have fully implemented governance programs. With data breaches now averaging nearly $4.5 million each, according to IBM, and IDC reporting organizations that build trustworthy AI are 60% more likely to double the ROI of AI projects, the business case for AI governance is crystal clear.

โ€œPair ambition with strong guardrails: clear data lifecycle and access controls, evaluation and redโ€‘teaming, and humanโ€‘inโ€‘theโ€‘loop checkpoints where stakes are high,โ€ says Srivastava. โ€œBake security, privacy, and data governance into the SDLC so ship and secure move together โ€” no black boxes for data lineage or model behavior.โ€

Itโ€™s not magic

According to BCG, only 22% of companies have advanced their AI beyond the POC stage, and just 4% are creating substantial value. With these sobering statistics in mind, CIOs shouldnโ€™t set unrealistic expectations for getting a return.

Finding ROI from AI will require significant upfront effort, and necessitate fundamental changes to organizational processes. As Mastercardโ€™s CTO for operations George Maddaloni said in a recent interview with Runtime, he thinks gen AI app adoption is largely about change management and adoption.

The pitfalls with AI are nearly endless and itโ€™s common for organizations to chase hype rather than value, launch without a clear data strategy, scale too quickly, and implement security as an afterthought. Many AI programs simply donโ€™t have the executive sponsorship or governance to get where they need to be, either. Alternatively, itโ€™s easy to buy into vendor hype on productivity gains and overspend, or underestimate the difficulty of integrating AI platforms with legacy IT infrastructure.


Looking ahead, to better maximize AIโ€™s business impact, leaders recommend investing in the data infrastructure and platform capabilities needed to scale, and hone on one or two high-impact use cases that can remove human toil and clearly drive revenue or efficiency.

Grounding AI fervor in core tenets and understanding the business strategy youโ€™re aiming for is necessary to inch toward ROI. Because, without sound leadership and clear objectives, AI is only a fascinating technology with a reward thatโ€™s just always out of reach.

โ€œ์šฐ์„ ์ˆœ์œ„๋ฅผ ์กฐ์ •ํ•˜๋ผโ€ AI ํ™•์‚ฐ๊ธฐ์— ํ•„์š”ํ•œ ์˜ˆ์‚ฐ ์šด์˜ ๋ฐฉ์‹

23 November 2025 at 21:36

๊ฐ€ํŠธ๋„ˆ๋Š” ์ „ ์„ธ๊ณ„ AI ์ง€์ถœ์ด ์˜ฌํ•ด 1์กฐ 5,000์–ต ๋‹ฌ๋Ÿฌ์—์„œ 2026๋…„์— 2์กฐ ๋‹ฌ๋Ÿฌ์— ์ด๋ฅผ ๊ฒƒ์œผ๋กœ ์ „๋งํ–ˆ๋‹ค. ๊ฒฝ์˜ ์ปจ์„คํŒ… ํšŒ์‚ฌ ์›จ์ŠคํŠธ ๋จผ๋กœ๊ฐ€ ๋Œ€๊ธฐ์—… ์ž„์› 300๋ช… ์ด์ƒ์„ ๋Œ€์ƒ์œผ๋กœ ์ง„ํ–‰ํ•œ ์„ค๋ฌธ์—์„œ๋„ ์‘๋‹ต์ž์˜ 85%๊ฐ€ ๋‚ด๋…„์— IT ์˜ˆ์‚ฐ์„ ๋Š˜๋ฆด ๊ณ„ํš์ด๋ผ๊ณ  ๋‹ตํ–ˆ์œผ๋ฉฐ, ์ด ๊ฐ€์šด๋ฐ ์ƒ๋‹น ๋ถ€๋ถ„์ด AI์— ํˆฌ์ž…๋  ์˜ˆ์ •์ด๋ผ๊ณ  ๋ฐํ˜”๋‹ค. ์‘๋‹ต ์ž„์›์˜ 42%๋Š” AI์™€ ๋ฐ์ดํ„ฐ ์—ญ๋Ÿ‰ ํ™•์žฅ์„ ๊ธฐ์ˆ  ํˆฌ์ž ์ตœ์šฐ์„  ๊ณผ์ œ๋กœ ๊ผฝ์•˜๊ณ , 91%๋Š” AI๊ฐ€ ๊ธฐ์ˆ  ์ง€์ถœ ์ฆ๊ฐ€๋ฅผ ์œ ๋ฐœํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ๋‹ตํ–ˆ๋‹ค. ์•ฝ 3/4์€ AI ๋„์ž…์œผ๋กœ ์™ธ๋ถ€ ๊ณ„์•ฝ ๋น„์šฉ๋„ ๋Š˜๋ฆด ๊ณ„ํš์ด๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

์›จ์ŠคํŠธ ๋จผ๋กœ์˜ ์ตœ๊ณ  AI ์ฑ…์ž„์ž ๋ธŒ๋ › ๊ทธ๋ฆฐ์Šคํƒ€์ธ์€ ์ง€๋‚œ ๋ช‡ ๋…„ ๋™์•ˆ ๋งŽ์€ ๊ธฐ์—…์ด AI๊ฐ€ ๋ฌด์—‡์„ ํ•  ์ˆ˜ ์žˆ๋Š”์ง€ ํŒŒ์•…ํ•˜๊ธฐ ์œ„ํ•œ ๊ฐœ๋…๊ฒ€์ฆ ์ž‘์—…์„ ์ง„ํ–‰ํ•˜๊ณ  ์žˆ์—ˆ์ง€๋งŒ, ์ตœ๊ทผ ๋ถ„์œ„๊ธฐ๋Š” ๋‹ฌ๋ผ์กŒ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋ฆฐ์Šคํƒ€์ธ์€ โ€œ์ด์ œ๋Š” ์‚ฌ์šฉ๋ก€๋‚˜ ๊ฐœ๋…๊ฒ€์ฆ์„ ์ด์•ผ๊ธฐํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ํฌ๊ฒŒ ์ค„๊ณ , 1๋‹จ๊ณ„ ๋˜๋Š” 2๋‹จ๊ณ„ ํ”„๋กœ์ ํŠธ ๋…ผ์˜๊ฐ€ ๋งŽ์•„์กŒ๋‹คโ€๊ณ  ๋งํ–ˆ๋‹ค.

๊ทธ๋ฆฐ์Šคํƒ€์ธ์€ AI๊ฐ€ ๋ฌด์—‡์„ ํ•  ์ˆ˜ ์žˆ๋Š”์ง€ ํŒ๋‹จํ•˜๋Š” ์ผ์ด ์ด์ œ๋Š” ์–ด๋ ต์ง€ ์•Š๋‹ค๋ฉฐ, โ€œ์–ด๋–ค ๊ณผ์ œ๋ฅผ ๋ณด๋ฉด AI๋กœ ์ถฉ๋ถ„ํžˆ ํ•ด๊ฒฐํ•  ์ˆ˜ ์žˆ๋Š”์ง€ ๋ฐ”๋กœ ํŒ๋‹จํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ํ•˜์ง€๋งŒ CIO๊ฐ€ ์›ํ•˜๋Š” ๋งŒํผ ๋งˆ์Œ๋Œ€๋กœ ์˜ˆ์‚ฐ์„ ์“ธ ์ˆ˜ ์žˆ๋‹ค๋Š” ์˜๋ฏธ๋Š” ์•„๋‹ˆ๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

๊ธ€๋กœ๋ฒŒ ํˆฌ์žยท๋ณดํ—˜์‚ฌ ํ”„๋ฆฐ์‹œํŽ„ ํŒŒ์ด๋‚ธ์…œ ๊ทธ๋ฃน์˜ ์ตœ๊ณ  ๋ฐ์ดํ„ฐ ๋ฐ ์• ๋„๋ฆฌํ‹ฑ์Šค ์ฑ…์ž„์ž ๋ผ์ œ์‹œ ์•„๋กœ๋ผ๋Š” ํ˜„์žฌ ํšŒ์‚ฌ์˜ ์ดˆ์ ์€ ์ธก์ • ๊ฐ€๋Šฅํ•œ ๋น„์ฆˆ๋‹ˆ์Šค ๊ฐ€์น˜๋ฅผ ์ œ๊ณตํ•˜๋Š” ๋ฐ ๋งž์ถฐ์ ธ ์žˆ๋‹ค๊ณ  ๋งํ–ˆ๋‹ค.

ํ”„๋ฆฐ์‹œํŽ„ ํŒŒ์ด๋‚ธ์…œ ๊ทธ๋ฃน์€ ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ํ”Œ๋žซํผ๊ณผ ๊ณ ๊ฐ€์น˜ ์‚ฌ์šฉ๋ก€๋ฅผ ์ค‘์‹ฌ์œผ๋กœ ์˜ˆ์‚ฐ์„ ์žฌ๋ฐฐ๋ถ„ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์—„๊ฒฉํ•œ ROI ์ถ”์ ๊ณผ ๋น„์šฉ ๊ฑฐ๋ฒ„๋„Œ์Šค๋„ ๋„์ž…ํ•˜๊ณ  ์žˆ๋‹ค. ์•„๋กœ๋ผ๋Š” ์ด์ œ ์‹คํ—˜์  ํŒŒ์ผ๋Ÿฟ ๋‹จ๊ณ„๋ฅผ ๋„˜์–ด์„ฐ์œผ๋ฉฐ, ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ํ”Œ๋žซํผ๋ฟ ์•„๋‹ˆ๋ผ ๋ผ์ดํ”„์‚ฌ์ดํด ๊ด€๋ฆฌ ๋„๊ตฌ, ๋ฐ์ดํ„ฐ ๊ธฐ๋ฐ˜, ์šด์˜ AI ์—ญ๋Ÿ‰๋„ ํ•จ๊ป˜ ๊ฒ€ํ† ํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์•„๋กœ๋ผ๋Š” โ€œํ”„๋กœ์„ธ์Šค๋ฅผ ์ž๋™ํ™”ํ•˜๊ณ  ๊ณ ๊ฐ ๊ฒฝํ—˜์„ ๊ฐœ์„ ํ•˜๊ณ , ์ƒˆ๋กœ์šด ์—ญ๋Ÿ‰์„ ๊ตฌ์ถ•ํ•˜๊ณ  ๋ฆฌ์Šคํฌ ๊ด€๋ฆฌ ์ฒด๊ณ„๋ฅผ ๊ฐ•ํ™”ํ•˜๋Š” ์†”๋ฃจ์…˜์„ ์ฐพ๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ, โ€œ๋ชจ๋“  ๋น„์šฉ์„ ์ตœ๋Œ€ํ•œ ํšจ์œจ์ ์œผ๋กœ ์‚ฌ์šฉํ•˜๋Š” ๊ฒƒ์ด ๋ชฉํ‘œ๋‹ค. ์˜ˆ์‚ฐ์ด ์œ ํ•œํ•œ ๋งŒํผ ์šฐ์„ ์ˆœ์œ„ ์กฐ์ •๋„ ๋ถˆ๊ฐ€ํ”ผํ•˜๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค.

์˜ˆ๋ฅผ ๋“ค์–ด ํ”„๋ฆฐ์‹œํŽ„ ํŒŒ์ด๋‚ธ์…œ ๊ทธ๋ฃน์€ ๊ฐ€์น˜๊ฐ€ ๋‚ฎ์€ ํˆฌ์ž ํ•ญ๋ชฉ์„ ์ค‘๋‹จํ•˜๊ณ , ๊ณ ๊ฐ€์น˜ ์‚ฌ์šฉ๋ก€์— ์ง‘์ค‘ํ•˜๊ณ  ์žˆ๋‹ค. ๊ณ„์•ฝ ๊ด€๋ฆฌ๋„ ๊ฐ•ํ™”ํ•˜๊ณ  ์กฐ๊ฑด ์žฌํ˜‘์ƒ๋„ ์ง„ํ–‰ ์ค‘์ด๋‹ค. ๋˜ ์ž๋™ํ™” ์ธก๋ฉด์—์„œ๋Š” LLM ์šด์˜ ๋น„์šฉ ๊ฒฝ๋ณด ๊ธฐ๋Šฅ๊ณผ ์ฃผ์š” ์‚ฌํ•ญ ๋ฒ„์ „ ๊ด€๋ฆฌ ๊ธฐ๋Šฅ์„ ๊ตฌ์ถ•ํ•ด ์ด์ƒ ์ง•ํ›„๋ฅผ ์‹๋ณ„ํ•˜๊ณ  ์ดˆ๊ณผ ๋น„์šฉ์„ ๋ฐฉ์ง€ํ•˜๊ณ  ์žˆ๋‹ค.

LLM์€ ๋™์ผํ•œ ์ž…๋ ฅ์— ๋‹ค๋ฅธ ๊ฒฐ๊ณผ๋ฅผ ์ƒ์„ฑํ•  ์ˆ˜ ์žˆ๊ณ  ๋ชจ๋ธ ๋ฒ„์ „์— ๋”ฐ๋ผ์„œ๋„ ์„ฑ๋Šฅ๊ณผ ๋น„์šฉ์ด ํฌ๊ฒŒ ๋‹ฌ๋ผ์งˆ ์ˆ˜ ์žˆ๋‹ค. ๋ฒ„์ „ ๊ด€๋ฆฌ ๊ธฐ๋Šฅ์€ ์†Œํ”„ํŠธ์›จ์–ด ๋ณ€๊ฒฝ๋ฟ ์•„๋‹ˆ๋ผ ๋ชจ๋ธ, ๋ฐ์ดํ„ฐ, ํ”„๋กฌํ”„ํŠธ ์‚ฌ์šฉ ์ด๋ ฅ์„ ์ถ”์ ํ•œ๋‹ค. ์•„๋กœ๋‚˜๋Š” โ€œ๊ฒฐ๊ตญ ๋น„์šฉ ํ†ต์ œ๋ฅผ ์œ„ํ•ด AI๋ฅผ ์ „๋žต์  ๊ด€๋ฆฌ ๋Œ€์ƒ์œผ๋กœ ์‚ผ๋Š”๋‹ค๋Š” ์˜๋ฏธโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์•„๋กœ๋ผ์˜ ๊ฒฝํ—˜์€ ๋“œ๋ฌธ ์‚ฌ๋ก€๊ฐ€ ์•„๋‹ˆ๋‹ค. ๋‹ค์–‘ํ•œ ๊ทœ๋ชจ์™€ ์—…์ข…์˜ ๊ธฐ์—…์ด ๊ฐœ๋…๊ฒ€์ฆ ๋‹จ๊ณ„๋ฅผ ๋„˜์–ด ๋ณธ๊ฒฉ์ ์ธ ๋Œ€๊ทœ๋ชจ ๋„์ž… ๋‹จ๊ณ„๋กœ ์ด๋™ํ•˜๋ฉด์„œ AI ์ง€์ถœ ๊ด€๋ฆฌ๋ผ๋Š” ์ƒˆ๋กœ์šด ๊ณผ์ œ์— ์ง๋ฉดํ•˜๊ณ  ์žˆ๋‹ค. ROI ์š”๊ตฌ ์ฆ๊ฐ€, ๋ ˆ๊ฑฐ์‹œ ์˜ˆ์‚ฐ์˜ AI ์ „ํ™˜, ๊ธฐ์ˆ  ๋ถ€์ฑ„ ๊ด€๋ฆฌ ๋“ฑ์ด ๋Œ€ํ‘œ์ ์ธ ๊ณผ์ œ๋‹ค.

์„ฑ๊ณผ ์ฆ๋ช…์— ๋Œ€ํ•œ ์••๋ฐ•

ํ”„๋ฆฐ์‹œํŽ„ ํŒŒ์ด๋‚ธ์…œ ๊ทธ๋ฃน์€ AI ํˆฌ์ž๊ฐ€ ๊ฐ€์น˜๊ฐ€ ์žˆ์Œ์„ ์ž…์ฆํ•˜๊ธฐ ์œ„ํ•ด ํšจ์œจ์„ฑ ํ–ฅ์ƒ, ๋ฆฌ์Šคํฌ ๊ฐ์†Œ, ๊ณ ๊ฐ ๋งŒ์กฑ๋„ ๊ฐœ์„ , ์ง์› ๊ฒฝํ—˜ ๊ฐœ์„ ์„ ์ถ”์ ํ•˜๊ณ  ์žˆ๋‹ค. ์ด๋Š” AI๊ฐ€ ์ฐฝ์ถœํ•˜๋Š” ๊ฐ€์น˜๋ฅผ ์ „์ฒด์ ์œผ๋กœ ํŒŒ์•…ํ•˜๋Š” ๋ฐ๋„ ๋„์›€์ด ๋œ๋‹ค.

์•„๋กœ๋ผ๋Š” โ€œ๊ท ํ˜• ์žกํžŒ ํฌํŠธํด๋ฆฌ์˜ค๋ฅผ ์œ ์ง€ํ•˜๋Š” ์ ‘๊ทผ ๋ฐฉ์‹์„ ์ทจํ•˜๊ณ  ์žˆ๋‹ค. ์ด๋Š” ๋‹จ๊ธฐ ์„ฑ๊ณผ๋กœ ๋™๋ ฅ์„ ์–ป๋Š” ๋™์‹œ์— ์ „๋žต์  ์šฐ์œ„์™€ ์„ฑ์žฅ์„ ์œ„ํ•œ ์žฅ๊ธฐ ํ˜์‹ ์— ํˆฌ์žํ•œ๋‹ค๋Š” ์˜๋ฏธ์ด๋‹คโ€๋ผ๋ฉฐ, โ€œAI ์—ญ๋Ÿ‰์ด ์„ฑ์ˆ™ํ•ด์ง€๋ฉด์„œ ์„ฑ๊ณต ๊ธฐ์ค€์„ ๋”์šฑ ๋ช…ํ™•ํ•˜๊ฒŒ ์„ค์ •ํ•˜๊ณ  ์žฅ๊ธฐ์  ์ง€์† ๊ฐ€๋Šฅ์„ฑ์„ ๋ณด์žฅํ•˜๋Š” ์ผ์ด ์ค‘์š”ํ•ด์กŒ๋‹คโ€๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

๊ทœ๋ชจ๊ฐ€ ์ž‘์€ ๊ธฐ์—…์˜ ๊ธฐ์ˆ  ์ฑ…์ž„์ž๋„ AI ํ”„๋กœ์ ํŠธ์˜ ๊ฒฐ๊ณผ๋ฅผ ์ž…์ฆํ•ด์•ผ ํ•˜๋Š” ์ƒํ™ฉ์ด๋‹ค. ๋ฏธ๊ตญ ์˜ค์Šคํ‹ด๊ณผ ์ƒŒ์•ˆํ† ๋‹ˆ์˜ค์— 4๊ฐœ ์‚ฌ๋ฌด์†Œ๋ฅผ ๋ณด์œ ํ•œ JB๊ตฟ์œˆ ๋ฆฌ์–ผํ„ฐ์Šค๋Š” 800๋ช… ๊ทœ๋ชจ์˜ ์—์ด์ „ํŠธ, ํŒŒํŠธ๋„ˆ, ์ง์›์ด ๊ทผ๋ฌดํ•˜๋Š”๋ฐ, ๋ชจ๋‘๊ฐ€ AI๋ฅผ ์ ๊ทน์ ์œผ๋กœ ํ™œ์šฉํ•˜๊ณ  ์žˆ๋‹ค.

JB๊ตฟ์œˆ ๋ฆฌ์–ผํ„ฐ์Šค์˜ ๋ถ€์‚ฌ์žฅ ์—๋“œ์›Œ๋“œ ํˆด์€ โ€œCEO๋„ ๋งค์ผ AI๋ฅผ ์‚ฌ์šฉํ•œ๋‹ค. ๋ชจ๋“  ์—์ด์ „ํŠธ๋„ AI๋ฅผ ํ™œ์šฉํ•˜๊ณ  ์žˆ๊ณ , ์ถ”๊ฐ€ ์ง€์ถœ ์Šน์ธ๋„ ๋ฐ›์•˜๋‹ค. ํ•˜์ง€๋งŒ ROI๋Š” ์ง์ ‘ ์ฆ๋ช…ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค. ํˆด์€ โ€œ์†Œ๊ทœ๋ชจ๋กœ ์ง€์ถœํ•˜๊ณ  ์‚ฌ์šฉ๋ก€๋ฅผ ์ž…์ฆํ•˜๋ฉด, ์˜ˆ์‚ฐ์„ ์กฐ๊ธˆ ๋” ๋ฐ›๊ณ  ๋‹ค์‹œ ์กฐ๊ธˆ ๋” ์ง€์ถœํ•˜๋Š” ๋ฐฉ์‹์ด๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด ํšจ์œจ์„ฑ์ด ๊ฐœ์„ ๋˜์—ˆ๋Š”์ง€ ๋ณด์—ฌ์ฃผ๊ธฐ ์œ„ํ•ด ๊ธฐ์กด ๋ฐฉ์‹๊ณผ AI ๋ฐฉ์‹ ๋‘ ๊ฐ€์ง€ ํ”„๋กœ์„ธ์Šค๋ฅผ ๋ณ‘ํ–‰ํ•ด ๋น„๊ตํ•˜๊ธฐ๋„ ํ•œ๋‹ค.

๊ฐ€ํŠธ๋„ˆ ์• ๋„๋ฆฌ์ŠคํŠธ ๋ฉœ๋Ÿฌ๋‹ˆ ํ”„๋ฆฌ์ฆˆ๋Š” ๋น„์šฉ ์ ˆ๊ฐ ํšจ๊ณผ๊ฐ€ ์žˆ๋Š” AI ํ”„๋กœ์ ํŠธ์— ์ง‘์ค‘ํ•˜๋Š” ๊ฒƒ์ด ์„ฑ๊ณผ๋ฅผ ๋ณด์—ฌ์ฃผ๊ณ  ๋™๋ ฅ์„ ํ™•๋ณดํ•˜๋Š” ์ข‹์€ ๋ฐฉ์‹์ด๋ผ๊ณ  ํ‰๊ฐ€ํ–ˆ๋‹ค. ํ”„๋ฆฌ์ฆˆ๋Š” โ€œ๋น„์šฉ ์™ธ์—๋„ ์žฅ๊ธฐ์  ๊ฐ€์น˜๋กœ ์ด์–ด์ง€๋Š” ๋‹ค์–‘ํ•œ ํšจ๊ณผ๋ฅผ ์–ป์„ ์ˆ˜ ์žˆ๋‹คโ€๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด ์ธํ”„๋ผ์™€ ์šด์˜์—์„œ๋Š” ํด๋ผ์šฐ๋“œ ๋น„์šฉ ๊ด€๋ฆฌ, IT ์„œ๋น„์Šค ์ง€์›, ์ „๋ฐ˜์  ์ง์› ์ƒ์‚ฐ์„ฑ์ด ๋Œ€ํ‘œ์  ์„ฑ๊ณผ์ด๋‹ค.

ํ”„๋ฆฌ์ฆˆ๋Š” โ€œ๋น„์šฉ ์ตœ์ ํ™”๋ฟ ์•„๋‹ˆ๋ผ ํ˜์‹ , ํšจ์œจ์„ฑ, ์ธ์žฌ ๊ด€๋ฆฌ ์ตœ์ ํ™” ๊ฐ™์€ ๋‹ค์–‘ํ•œ ๊ฐ€์น˜๋ฅผ ํ•จ๊ป˜ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

ํˆฌ์ž ์šฐ์„ ์ˆœ์œ„์˜ ๋ณ€ํ™”

์•„์ง ๋ช…ํ™•ํ•œ ROI๊ฐ€ ์—†๋Š” ์‹คํ—˜์  AI ํ”„๋กœ์ ํŠธ ๋น„์šฉ์„ ์ถฉ๋‹นํ•˜๋Š” ๋˜ ํ•˜๋‚˜์˜ ๋ฐฉ๋ฒ•์€ ๋‹ค๋ฅธ ์˜ˆ์‚ฐ ํ•ญ๋ชฉ์—์„œ ์ž๊ธˆ์„ ์ด์ „ํ•˜๋Š” ๊ฒƒ์ด๋‹ค. ํˆด์€ ๊ทธ๋Ÿฐ ๋ฐฉ์‹์„ ์‹ค์ œ๋กœ ์‚ฌ์šฉํ•œ๋‹ค๋ฉฐ, โ€œAI ์ง€์ถœ์„ ์ƒ์‡„ํ•˜๊ธฐ ์œ„ํ•ด ๋‹ค๋ฅธ ๋น„์šฉ์„ ์ค„์ธ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์›จ์ŠคํŠธ ๋จผ๋กœ์˜ ๋ธŒ๋ › ๊ทธ๋ฆฐ์Šคํƒ€์ธ์€ ๋ชจ๋“  ๊ธฐ์—…์ด AI ์ค‘์‹ฌ ๋˜๋Š” AI ๋„ค์ดํ‹ฐ๋ธŒ ์กฐ์ง์ด ๋˜๊ณ  ์‹ถ์–ด ํ•˜์ง€๋งŒ, โ€œํšŒ์‚ฌ ์ƒ์กด์— ์ง๊ฒฐ๋˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ์•„๋‹ˆ๋ผ๋ฉด ์ถ”๊ฐ€ ์˜ˆ์‚ฐ์ด ๋”ฐ๋กœ ์žˆ๋Š” ๊ณณ์€ ์—†๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค. ๊ทธ๋ž˜์„œ ๋ ˆ๊ฑฐ์‹œ ํ”„๋กœ์ ํŠธ ์˜ˆ์‚ฐ์„ AI๋กœ ์ „ํ™˜ํ•˜๋Š” ์ „๋žต์ด ๋„๋ฆฌ ์“ฐ์ธ๋‹ค๋Š” ๊ฒƒ.

๊ทธ๋ฆฐ์Šคํƒ€์ธ์€ โ€œ๊ธฐ์—… ๋‚ด๋ถ€ ์šฐ์„ ์ˆœ์œ„๊ฐ€ ์ด๋™ํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค. ๊ธฐ์—…์€ ๊ธฐ์กด ํˆฌ์ž๊ฐ€ AI ๋•Œ๋ฌธ์— ๋ถˆํ•„์š”ํ•ด์กŒ๋Š”์ง€, ๋˜๋Š” AI๋กœ ๋Œ€์ฒด ๊ฐ€๋Šฅํ•œ์ง€๋ฅผ ๊ฒ€ํ† ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์†”๋ฃจ์…˜ ์—…์ฒด์—๋„ ๋น„์šฉ ์ ˆ๊ฐ์„ ๊ฐ•ํ•˜๊ฒŒ ์š”๊ตฌํ•˜๊ณ  ์žˆ๋‹ค.

๊ธฐ์ˆ  ์„ ๋„ ๊ธฐ์—…์กฐ์ฐจ ์ด๋Ÿฐ ์กฐ์ •์ด ํ•„์š”ํ•  ์ˆ˜ ์žˆ๋‹ค. ํฌ์ถ˜ 500๋Œ€ ๋ณดํ—˜ํšŒ์‚ฌ ํ•œ ๊ณณ์˜ ์ˆ˜์„ ๊ธฐ์ˆ  ์ž„์›์€ โ€œAI๋ฅผ ์œ„ํ•ด ์™„์ „ํžˆ ์ƒˆ๋กœ์šด ์˜ˆ์‚ฐ ํ•ญ๋ชฉ์„ ๋งŒ๋“  ๊ฒƒ์€ ์•„๋‹ˆ๋‹ค. AI ์˜ˆ์‚ฐ ์ฑ…์ • ๋ฐฉ์‹์€ ์•„์ง๋„ ์กฐ์œจ ์ค‘์ด๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

ํ•ด๋‹น ๊ธฐ์—…์€ ๋Œ€์‹  ๋‹ค๋ฅธ ์˜์—ญ์—์„œ ์ž๊ธˆ์„ ๋–ผ์–ด AI์— ํˆฌ์ž…ํ•˜๊ณ  ์žˆ๋‹ค. ์ด ๊ธฐ์ˆ  ์ž„์›์€ โ€œํ˜„์žฌ AI๋Š” ์ž์ฒด ์กฐ๋‹ฌ ๋ชจ๋ธ๋กœ ์šด์˜๋˜๊ณ  ์žˆ๋‹ค. ๋ ˆ๊ฑฐ์‹œ ๊ธฐ์ˆ ์— ํˆฌ์ž…ํ•˜๋˜ ํˆฌ์ž๋ฅผ AI๋กœ ์ด๋™ํ•˜๋Š” ๋ฐฉ์‹์ด๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด ์ด์ „์— ํŠน์ • ๊ธฐ์ˆ ์— ์—ฐ๊ฐ„ 100๋งŒ ๋‹ฌ๋Ÿฌ๋ฅผ ์‚ฌ์šฉํ•˜๋˜ ๋น„์šฉ์„ ์ž๋™ํ™”๋ฅผ ํ†ตํ•ด 90๋งŒ ๋‹ฌ๋Ÿฌ๋กœ ์ค„์˜€๋‹ค๋ฉด, ์ ˆ๊ฐํ•œ 10๋งŒ ๋‹ฌ๋Ÿฌ๋ฅผ AI์— ํˆฌ์ž…ํ•˜๋Š” ์‹์ด๋‹ค.

์ผ๋ถ€ ๊ธฐ์—…์€ ์†”๋ฃจ์…˜ ์—…์ฒด๊ฐ€ ๊ธฐ์กด ์ œํ’ˆ์— AI ๊ธฐ๋Šฅ์ด๋‚˜ ์—์ด์ „ํŠธ ๊ธฐ๋Šฅ์„ ์ถ”๊ฐ€ํ•˜๋ฉด์„œ ํšŒ์‚ฌ๊ฐ€ ์ƒˆ๋กœ์šด AI ๊ธฐ๋Šฅ์„ ๋ฌด๋ฃŒ๋กœ ์–ป๊ธฐ๋„ ํ•œ๋‹ค. ํ•˜์ง€๋งŒ ์–ด๋–ค ํ”Œ๋žซํผ์€ ์ƒˆ ๊ธฐ๋Šฅ์— ๋Œ€ํ•ด ์ถ”๊ฐ€ ๋น„์šฉ์„ ์ฒญ๊ตฌํ•œ๋‹ค. ์ด๋Ÿฐ ๊ตฌ์กฐ๋Š” 2026~2027๋…„์— ์ƒˆ๋กœ์šด ์ž๊ธˆ ๋ชจ๋ธ๋กœ ๋ฐœ์ „ํ•  ๊ฐ€๋Šฅ์„ฑ์ด ์žˆ์œผ๋ฉฐ, ๊ธฐ์—…์˜ AI ํ™œ์šฉ์ด ์„ฑ์ˆ™ํ•ด์งˆ์ˆ˜๋ก ์ž๊ธˆ ๋ชจ๋ธ๋„ ํ•จ๊ป˜ ์ง„ํ™”ํ•  ๊ฒƒ์œผ๋กœ ๋ณด์ธ๋‹ค.

์ด ๊ธฐ์ˆ  ์ž„์›์€ โ€œ๋†’์€ ๋น„์ฆˆ๋‹ˆ์Šค ๊ฐ€์น˜๋ฅผ ์ฐฝ์ถœํ•˜๊ฑฐ๋‚˜ ํšจ์œจ์„ฑ์„ ๋†’์ด๋Š” ์—ญ๋Ÿ‰์„ ์ž…์ฆํ•˜๋ฉด ๋ณ€ํ™”๊ฐ€ ๋‚˜ํƒ€๋‚  ๊ฒƒ์ด๋‹ค. ๊ทธ๋Ÿฌ๋ฉด ์†๋„๋ฅผ ๋†’์ด๊ธฐ ์œ„ํ•œ ์ถ”๊ฐ€ ํˆฌ์ž๋กœ ์ „ํ™˜ํ•  ๊ฒƒโ€์ด๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์˜ˆ์ƒ์น˜ ๋ชปํ•œ ์ƒํ™ฉ์„ ๋Œ€๋น„ํ•œ ๊ณ„ํš

IT ํ”„๋กœ์ ํŠธ ์˜ˆ์‚ฐ ์ˆ˜๋ฆฝ์€ ์›๋ž˜๋„ ์‰ฝ์ง€ ์•Š์•˜์ง€๋งŒ, AI๋Š” ์—ฌ๊ธฐ์— ์ƒˆ๋กœ์šด ๋‚œ์ œ๋ฅผ ๋”ํ•˜๊ณ  ์žˆ๋‹ค. ์ „๋ก€ ์—†๋Š” ๋ณ€ํ™” ์†๋„๊ฐ€ ๋Œ€ํ‘œ์ ์ธ ๊ณผ์ œ๋‹ค.

ํผ๋ธ”๋ฆฌ์‹œ์Šค ์‚ฌํ”ผ์—ฅํŠธ์˜ ์ตœ๊ณ  ์ œํ’ˆ ์ฑ…์ž„์ž ์…ธ๋˜ ๋ชฌํ…Œ์ด๋กœ๋Š” โ€œ์ง€๊ธˆ ์„ธ์šฐ๋Š” ์–ด๋–ค ๋ชจ๋ธ๋„ 6๊ฐœ์›” ๋’ค์—๋Š” ์œ ํšจํ•˜์ง€ ์•Š๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค. ๋ฌผ๋ก  ์ด๋Ÿฐ ๋ณ€ํ™”๊ฐ€ ํ•ญ์ƒ ๋‚˜์œ ๊ฒƒ๋งŒ์€ ์•„๋‹ˆ๋‹ค. ๋ชฌํ…Œ์ด๋กœ๋Š” ์ง€๋‚œ 2๋…„๊ฐ„ ์ผ๋ถ€ ๋ชจ๋ธ์˜ ํ† ํฐ๋‹น ๋น„์šฉ์ด ๊ธ‰๊ฒฉํžˆ ๋–จ์–ด์ง„ ๊ฒƒ์„ ์˜ˆ๋กœ ๋“ค์—ˆ๋‹ค. ํ•˜์ง€๋งŒ ๋ฐ˜๋Œ€๋กœ ๋” ์ƒˆ๋กญ๊ณ  ๋›ฐ์–ด๋‚œ ๋ชจ๋ธ์ด ๊ณ„์† ๋“ฑ์žฅํ•˜๊ณ , ์‚ฌ์šฉ๋Ÿ‰์ด ์ฆ๊ฐ€ํ•˜๋ฉด์„œ ์„ฑ๋Šฅ๋„ ์˜ˆ์ธกํ•˜๊ธฐ ์–ด๋ ต๋‹ค.

๋ชฌํ…Œ์ด๋กœ๋Š” โ€œ์ „ํ†ต ์†Œํ”„ํŠธ์›จ์–ด ๊ฒฝ์ œํ•™์—์„œ๋Š” ๊ฐœ๋ฐœ, ์—”์ง€๋‹ˆ์–ด๋ง, ์ธํ”„๋ผ ๊ฐ™์€ ์„ ํ–‰ ๋น„์šฉ์ด ๋“ค์–ด๊ฐ€์ง€๋งŒ, ์ผ๋‹จ ๊ณ ์ • ๋น„์šฉ์ด ํ™•์ •๋˜๋ฉด ์šด์˜ ๋น„์šฉ์€ ๋น„๊ต์  ์˜ˆ์ธก ๊ฐ€๋Šฅํ•˜๊ณ  ๊ด€๋ฆฌ ๊ฐ€๋Šฅํ•˜๋‹ค. ํ•˜์ง€๋งŒ AI๋Š” ์ถ”๋ก  ๋น„์šฉ์ด ๋ณ€๋™์ ์ด๊ณ , ์•ˆ์ „์„ฑ ๊ฒ€์ฆ๊ณผ ์ปดํ”Œ๋ผ์ด์–ธ์Šค ์ ๊ฒ€์—๋„ ์ถ”๊ฐ€ ๋น„์šฉ์ด ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค. ํ™•์žฅ๋„ ์„ ํ˜•์ ์ด์ง€ ์•Š๊ณ  ๊ธฐ์ˆ  ์ž์ฒด๊ฐ€ ๊ณ„์† ๋ณ€ํ•œ๋‹ค๋Š” ์ ๋„ ๋ฌธ์ œ๋‹ค.

๋ชฌํ…Œ์ด๋กœ๋Š” โ€œํƒ„๋ ฅ์ ์œผ๋กœ ๋Œ€์‘ํ•  ์ˆ˜ ์žˆ์–ด์•ผ ํ•œ๋‹ค. ์ง€๊ธˆ์€ ์–ด๋–ค ๊ธฐ์ˆ ์ด ์Šน์ž๋‚˜ ํŒจ์ž๊ฐ€ ๋ ์ง€ ํŒ๋‹จํ•˜๊ธฐ์กฐ์ฐจ ์–ด๋ ต๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค.

AI๊ฐ€ ์‚ฌ๋žŒ, ์‹œ์Šคํ…œ, ๋ฐ์ดํ„ฐ์— ์š”๊ตฌํ•˜๋Š” ๋ถ€๋‹ด๋„ ์˜ˆ์‚ฐ ์ˆ˜๋ฆฝ์˜ ๋˜ ๋‹ค๋ฅธ ๋‚œ์ œ๋‹ค. ํ”„๋ฆฐ์‹œํŽ„ ํŒŒ์ด๋‚ธ์…œ ๊ทธ๋ฃน์˜ ๋ผ์ œ์‹œ ์•„๋กœ๋ผ๋Š” AI ๋น„์šฉ ๊ด€๋ฆฌ์—์„œ ๊ฐ€์žฅ ํฐ ๊ณผ์ œ๊ฐ€ ์ธ์žฌ ํ™•๋ณด๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ์•„๋กœ๋ผ๋Š” โ€œ๊ธฐ์ˆ  ์—ญ๋Ÿ‰ ๊ฒฉ์ฐจ์™€ ๋ถ€์„œ ๊ฐ„ ์˜์กด์„ฑ ๋•Œ๋ฌธ์— ์ž‘์—…์ด ์ง€์—ฐ๋˜๊ณ  ๋น„์šฉ์ด ์ฆ๊ฐ€ํ•œ๋‹คโ€๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทœ์ œ ๋ณ€ํ™”๋„ ๋ฌธ์ œ๋‹ค. ๊ทœ์ œ๊ฐ€ ๊ณ„์† ๋ฐ”๋€Œ๋Š” ๋งŒํผ ๊ฑฐ๋ฒ„๋„Œ์Šค ์ฒด๊ณ„๋ฅผ ์ง€์†์ ์œผ๋กœ ์กฐ์ •ํ•ด์•ผ ํ•œ๋‹ค. ๋˜ ๊ธฐ์—…์€ AI์— ํ•„์š”ํ•œ ๋ฐ์ดํ„ฐ์™€ ๊ธฐ๋ฐ˜ ์‹œ์Šคํ…œ์„ ์ •๋น„ํ•˜๋Š” ๋ฐ ์‹ค์ œ๋กœ ์–ผ๋งˆ๋งŒํผ์˜ ๋น„์šฉ์ด ๋“œ๋Š”์ง€ ๊ณผ์†Œํ‰๊ฐ€ํ•˜๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค. ์ง์› ๊ต์œก ๋น„์šฉ๋„ ๋งˆ์ฐฌ๊ฐ€์ง€๋‹ค.

์•„๋กœ๋ผ๋Š” โ€œ๋ ˆ๊ฑฐ์‹œ ํ™˜๊ฒฝ์€ ๋ณต์žก์„ฑ๊ณผ ๋น„์šฉ์„ ์ฆ๊ฐ€์‹œํ‚จ๋‹ค. ์ด๋Ÿฐ ์ดˆ๊ธฐ ๋น„์šฉ์€ ๋ถ€๋‹ด์ด ํฌ์ง€๋งŒ ์žฅ๊ธฐ์  ๋น„ํšจ์œจ์„ ๋ง‰๊ธฐ ์œ„ํ•ด ํ•„์ˆ˜์ ์ด๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋˜ํ•œ AI ๊ธฐ์ˆ ์ด ๊ฐœ๋…๊ฒ€์ฆ ๋‹จ๊ณ„๋ฅผ ์ง€๋‚˜ ์‹ค์ œ ์šด์˜ ํ™˜๊ฒฝ์œผ๋กœ ์ด์ „๋˜๋ฉด, ๊ธฐ์—…์ด ๊ธฐ๋Œ€ํ•œ ๊ฒƒ๊ณผ ์ „ํ˜€ ๋‹ค๋ฅธ ๊ฒฐ๊ณผ๊ฐ€ ๋‚˜์˜ค๋Š” ๊ฒฝ์šฐ๋„ ๋งŽ๋‹ค.

ํ„ฐํ”„์ธ ๋Œ€ํ•™๊ต ๊ณต๋Œ€ ๋Œ€ํ•™์› ํ•™์žฅ ๊ฒธ IEEE ํŽ ๋กœ์šฐ์ธ ์นด๋ Œ ํŒจ๋„คํƒ€๋Š” โ€œ์ง€๊ธˆ์€ ๋ถˆํ™•์‹คํ•œ ์š”์†Œ๊ฐ€ ๋„ˆ๋ฌด ๋งŽ๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค. ํŒจ๋„คํƒ€๋Š” ๋งŽ์€ ์กฐ์ง์ด AI๋ฅผ ์‚ฌ๋žŒ์˜ ๋Œ€์ฒด์žฌ๋กœ ์ƒ๊ฐํ•˜์ง€๋งŒ ํ˜„์‹ค์€ ๊ทธ๋ ‡์ง€ ์•Š์œผ๋ฉฐ, ์ด์ „์—๋Š” ์‹ ๊ฒฝ ์“ฐ์ง€ ์•Š์•„๋„ ๋๋˜ ์ƒˆ๋กœ์šด ๋ฌธ์ œ๊ฐ€ ์ƒ๊ธฐ๊ธฐ ์‹œ์ž‘ํ–ˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ฒ‰์œผ๋กœ ๋ณด๋ฉด ๋งค์šฐ ๋งค๋ ฅ์ ์ธ ์„ ํƒ์ง€์ฒ˜๋Ÿผ ๋ณด์ธ๋‹ค. ์ง€๊ธˆ์€ ์ง์› 10๋ช…์ด ๊ณ ๊ฐ ์ „ํ™”๋ฅผ ๋ฐ›๊ณ  ์žˆ๊ณ , AI๊ฐ€ ๊ทธ 10๋ช…์˜ ์ผ์„ ๋Œ€์‹ ํ•  ์ˆ˜ ์žˆ์„ ๊ฒƒ์ฒ˜๋Ÿผ ๋А๊ปด์ง„๋‹ค. ํ•˜์ง€๋งŒ ํŒจ๋„คํƒ€๋Š” ๋ชจ๋ธ์„ ์ •์ƒ ํ๋ฆ„ ๊ธฐ์ค€์œผ๋กœ ์„ค๊ณ„ํ•œ ๊ฒฝ์šฐ๊ฐ€ ๋งŽ์•„ ์˜ˆ์™ธ ์ƒํ™ฉ์„ ์ฒ˜๋ฆฌํ•˜์ง€ ๋ชปํ•œ๋‹ค๊ณ  ์ง€์ ํ–ˆ๋‹ค. ์ด ๋•Œ๋ฌธ์— ๊ณ ๊ฐ ๋ถˆ๋งŒ์ด ์ƒ๊ธฐ๊ฑฐ๋‚˜ ์‹œ์Šคํ…œ์ด ์ค‘๋‹จ๋˜๋Š” ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ•  ์ˆ˜ ์žˆ๊ณ , ๋ณด์•ˆ ์ธก๋ฉด์—์„œ๋„ ์ด์ „์—๋Š” ์‚ฌ๋žŒ์ด ๊ฐ์ง€ํ•˜๋˜ ์œ„ํ—˜์„ ๋†“์น  ์ˆ˜๋„ ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

ํŒจ๋„คํƒ€๋Š” CIO๊ฐ€ AI๋กœ ๋ฌด์—‡์„, ์™œ ํ•˜๋ ค๋Š”์ง€ ๋ฉด๋ฐ€ํ•˜๊ฒŒ ๊ณ ๋ฏผํ•ด์•ผ ํ•œ๋‹ค๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค.

๋งŽ์€ CIO๋Š” ์ด๋ฏธ ๋น„์šฉ๊ณผ ๋ฆฌ์Šคํฌ ๊ด€๋ฆฌ์—์„œ ํƒˆํ”ผํ•ด ๋ฐ์ดํ„ฐ ๊ด€๋ฆฌ์™€ ์ธ์‚ฌ์ดํŠธ ์ฐฝ์ถœ ์—ญํ• ๋กœ ์ด๋™ํ–ˆ๊ณ , ๋น„์ฆˆ๋‹ˆ์Šค ์กฐ์ง๊ณผ ๋” ๊ฐ€๊นŒ์šด ์œ„์น˜์— ์„œ๊ณ  ์žˆ๋‹ค. ์ด์ œ CIO๋Š” ์•ˆ์ „ํ•˜๊ณ  ๋น„์šฉ ํšจ์œจ์ ์ธ ๋ฐฉ์‹์œผ๋กœ AI๋ฅผ ์ง€์›ํ•˜๋Š” ์—ญํ• ๋กœ ๋ฐœ์ „ํ•˜๊ณ  ์žˆ๋‹ค.

์›จ์ŠคํŠธ ๋จผ๋กœ์˜ ๋ธŒ๋ › ๊ทธ๋ฆฐ์Šคํƒ€์ธ์€ โ€œAI ๋„๊ตฌ๊ฐ€ ์ถœ์‹œ๋˜๋Š” ์ฆ‰์‹œ ๋ชจ๋“  ์ ‘๊ทผ์„ ์ฐจ๋‹จํ•˜๊ณ  ๋ฐฉํ™”๋ฒฝ์„ ๊ฑธ์–ด๋ฒ„๋ฆฐ CIO๋„ ์žˆ์—ˆ๋‹ค. ์ด๋Š” ์กฐ์ง์˜ AI ๋„์ž… ์ž์ฒด๋ฅผ ๋ง‰๋Š” ๊ฒฐ๊ณผ๋ฅผ ๋‚ณ์•˜๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๋ฐ˜๋ฉด ์ง„๋ณด์  CIO๋Š” ์‹ ์ค‘ํ•˜๊ณ  ์ฒด๊ณ„์ ์œผ๋กœ ์ ‘๊ทผํ•ด ๊ฑฐ๋ฒ„๋„Œ์Šค๋ฅผ ๊ตฌ์ถ•ํ•˜๊ณ , ์ƒˆ๋กœ์šด AI ์ค‘์‹ฌ ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ์•„ํ‚คํ…์ฒ˜๋ฅผ ๋งŒ๋“ค๊ณ  ์žˆ๋‹ค. ๊ทธ๋ฆฐ์Šคํƒ€์ธ์€ ์ด๋Ÿฐ CIO๊ฐ€ ๋ฏธ๋ž˜ ๊ธฐ์—…์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ๋งŒ๋“œ๋Š” ์—ญํ• ์„ ํ•˜๊ฒŒ ๋  ๊ฒƒ์ด๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

How CIOs can get a better handle on budgets as AI spend soars

21 November 2025 at 05:00

Gartner predicts global AI spending will hit $2 trillion in 2026, up from $1.5 trillion this year. And In a survey of over 300 executives at large companies by management consulting firm West Monroe Partners, 85% said they plan to increase IT budgets next year, with a big chunk going to AI. For 42% of executives, scaling AI and data capabilities is the top priority for technology investment, and 91% said AI is causing their tech spend to increase while nearly three quarters plan to spend more on contractors as a result of AI.

In the previous couple of years, many companies were doing POCs, just figuring out what AI can do, says Bret Greenstein, CAIO at West Monroe. But thatโ€™s all changing now. โ€œI see a lot less discussion of use cases and POCs, and more about phase one, phase two projects,โ€ he says.

Itโ€™s not as hard to assess whether or not AI can do something anymore, he adds. โ€œI can look at something and say this is highly addressable by AI.โ€ But that doesnโ€™t mean CIOs get carte blanche to spend all they want.

At the Principal Financial Group, a global investment and insurance company, the focus is now on delivering measurable business value, says Rajesh Arora, the companyโ€™s chief data and analytics officer.

โ€œWeโ€™re reallocating budgets toward scalable platforms and high-impact use cases,โ€ he says. Plus, the firm is implementing rigorous ROI tracking and cost governance. Thatโ€™s because the firm is moving past experimental pilots, he says. In addition to looking for platforms that can scale, the company is also looking at lifecycle management tools, data foundations, and operational AI capabilities.

โ€œThese are solutions thatโ€™ll automate processes, enhance customer experience, build new capabilities, and strengthen risk management,โ€ he says. โ€œOur goal is to make every dollar work harder.โ€ And that means some things have to go.

The company is pausing low-impact investments, for example, in favor of high-value use cases. And theyโ€™re tightening up their contract governance and renegotiating terms. Thereโ€™s also automation. โ€œWeโ€™re deploying cost-alerting for LLM ops and feature story versioning to flag anomalies and prevent overruns,โ€ he says.

LLMs can produce different results for the same output, and different versions of the model can have very different performance metrics and costs. And feature story versioning tracks software changes, as well as the model, data, and prompts used. So it all comes down to AI becoming a strategic focus to manage costs.

Aroraโ€™s experience isnโ€™t unique. Enterprises of all sizes and in all verticals are grappling with their AI spending as they move on from POCs to actual deployment at scale, which often means facing new demands for ROI, shifting money from legacy to AI projects, and struggling to get a handle on technical debt.

The push for proof

To prove its AI investments are worth the dough, Principal tracks efficiency gains, reductions in risk, improved customer satisfaction, and better employee experience. This creates a holistic view of the value that AI creates, says Arora.

โ€œOur approach is to maintain a balanced portfolio,โ€ he says. That means both short-term wins that build momentum, and long-term innovation to drive strategic advantage and growth. โ€œAs AI capabilities mature, we must be more intentional about how we define success and ensure long-term sustainability,โ€ he adds.

Tech executives at smaller companies are also having to show results from their AI projects. JBGoodwin Realtors, with four offices in Austin and San Antonio, has 800 agents, partners, and employees, and everyone is all-in on AI, says Edward Tull, the companyโ€™s VP of technology and operations.

โ€œThe CEO uses it every day,โ€ he says. โ€œAll the agents use it, too, and we have approval to spend more.โ€ But he has to show ROI. โ€œI have to prove it,โ€ he adds. โ€œI spend a little, prove the use case, and then I get a little more and spend a little more.โ€ So for example, AI might result in better efficiency so to demonstrate this, he might run two processes in parallel, one the old-fashioned way, and the other with AI.

Focusing on AI projects that result in cost savings is a good way to show results and build momentum, agrees Gartner analyst Melanie Freeze. โ€œWe know that can lead to other non-cost considerations and long-term value.โ€ For example, in infrastructure and operations, likely wins include cloud cost management, IT service support, and general employee productivity, she says.

โ€œYou can get cost optimization, but also all that other value like innovation, efficiency, optimizing talent management,โ€ she says.

A shift in priorities

Another way to pay for AI projects, especially experimental ones that donโ€™t yet have clear ROI, is to take money from other areas. JBGoodwinโ€™s Tull says he does that. โ€œIโ€™ll get rid of other things we spend on, to offset what I spend on AI,โ€ he says.

Everyone wants to become AI-centric or AI-native, says West Monroeโ€™s Greenstein. โ€œBut nobody has extra buckets of money to do this unless itโ€™s existential to their company,โ€ he says. So moving money from legacy projects to AI is a popular strategy.

โ€œItโ€™s a shift of priorities within companies,โ€ he says. โ€œThey look at their investments and ask how many are no longer needed because of AI, or how many can be done with AI. Plus, theyโ€™re putting pressure on vendors to drive down costs. Theyโ€™re definitely squeezing existing suppliers.โ€

Even large, tech-forward companies might have to do this kind of juggling.

โ€œWe didnโ€™t create a whole new allocation for AI,โ€ says one senior tech executive at a Fortune 500 insurance company. โ€œWeโ€™re still working through the mechanics of budgeting for AI.โ€

Instead, the firm is carving out funds from other areas.

โ€œAI is in a self-funding model at the moment,โ€ he says. โ€œWeโ€™re shifting investment from legacy technologies to AI.โ€ For example, he says, if the company was spending a million dollars on a particular technology and used automation to get it down to $900,000 a year, the $100,000 savings could go toward AI.

And sometimes the company can get new AI for free, he says, as vendors add AI functionality or agentic capabilities to existing products. But other platforms charge extra for the new features. โ€œSome of it is inherent in the solution, though, and doesnโ€™t really change the cost,โ€ he says. That might evolve to new funding in 2026 to 2027, he adds. But as the companyโ€™s use of AI continues to mature, the funding model will evolve as well, he says.

โ€œWeโ€™ll see that change as we demonstrate capabilities that either deliver high business value or efficiency gains,โ€ he says. โ€œThen weโ€™ll shift to additional infusions of investment to accelerate things.โ€

Planning for the unexpected

Budgeting for IT projects has never been simple, but AI adds its own challenges. The unprecedented pace of change is one of them.

โ€œWhatever modeling I do now is not going to be valid in six months,โ€ says Sheldon Monteiro, chief product officer at Publicis Sapient. This isnโ€™t always a bad thing. For example, the per token prices of some models have dropped dramatically over the past two years, he says. But on the flipside, there are always newer and better models, growing usage, and unpredictable performance.

โ€œWith traditional software economics, you have upfront costs like development, engineering, or infrastructure, but once you have those fixed costs, operating costs are relatively predictable and manageable,โ€ he says. With AI, though, the inference costs are variable, and the guardrail and compliance checks might have additional costs, he says. Scaling is also non-linear and the tech itself is in constant flux.

โ€œYou need to be able to flex,โ€ says Monteiro, โ€œAnd to recognize that now, winners and losers are hard to call.โ€

Another challenge to budgeting is the demands that AI places on people, systems, and data. One of the most significant challenges to managing AI costs is talent, says Principalโ€™s Arora. โ€œSkill gaps and cross-team dependencies can slow deliveries and drive up costs,โ€ he says.

Then thereโ€™s the problem of evolving regulations, and the need to continuously adapt governance frameworks to stay resilient in the face of these changes. Organizations also often underestimate how much money will be needed to train employees, and to bring data and other foundational systems in line with whatโ€™s needed for AI.

โ€œLegacy environments add complexity and expense,โ€ he adds. โ€œThese one-time costs are heavy but essential to avoid long-term inefficiencies.โ€

Finally, when AI technology is actually moved out of POCs into production, it often turns out very different to what companies expected.

โ€œThere are so many unknowns right now,โ€ says Karen Panetta, IEEE fellow and dean of graduate engineering at Tufts University. โ€œPeople think of it as a replacement for a human, and itโ€™s not. And you get new areas you havenโ€™t had to worry about before.โ€ For example, many companies look to use AI agents to replace customer service or support teams.

โ€œItโ€™s really appealing,โ€ she says. โ€œYouโ€™ve got 10 people answering phone calls now, and it feels like AI is going to do the job of those ten people. But Iโ€™ve designed it for normal process flows, so what about all the exceptions? Now you have angry customers, or it breaks and is unavailable. And what about security? Before, we had humans to detect these things.โ€

CIOs have to be thoughtful about what theyโ€™re doing with the AI and why, she says.

Many CIOs have already transitioned from managing costs and risks, to managing data and becoming enablers of insight, and getting closer to the business units. Now theyโ€™re in a position to become enablers of AI, while doing it safely and at cost.

โ€œThere are some CIOs that blocked and firewalled every AI tool the day it came out,โ€ says West Monroeโ€™s Greenstein. โ€œThat blocked companies from adoption. The ones who are progressive are being thoughtful, deliberate, are building governance models, and creating a new enterprise architecture around AI. The CIOs who are embracing that are enabling the enterprises of tomorrow.โ€

Salesforce unveils observability tools to manage and optimize AI agents

20 November 2025 at 08:00

Salesforce today unveiled new Agentforce 360 observability tools to give teams visibility into why AI agents behave the way they do, and which reasoning paths they follow to reach decisions.

Salesforce is providing the new tools as its agentic AI customers increasingly shift focus from building agents to maintaining them in production.

โ€œWeโ€™ve had thousands of customers use the platform,โ€ says Madhav Thattai, SVP and COO of Agentforce at Salesforce. โ€œIn the first 12 months, the focus has been, โ€˜How do we build good agents? How do we connect the right data? How do we control the agent behavior?โ€™ Itโ€™s really been about build and design of the agents.โ€

But Thattai adds that an internal Salesforce survey now shows a three-fold increase in implementations that customers are now moving to production.

โ€œThousands of customers are now live with agents in production,โ€ he says. โ€œAs customers move up the maturity curve, the problems shift from how to build a good agent to how to manage an agent at scale. Thatโ€™s really what the observability tool stack is about.โ€

The new tools, he says, help customers understand whether agents deliver value, how they perform in interactions, and how to improve their performance.

Triple threat

The tools span three areas: Agent Analytics, Agent Optimization, and Agent Health Monitoring.

Agent Analytics is intended to help organizations continuously refine their AI agents by providing visibility into agent performance. Teams can use Agent Analytics to track agent usage and effectiveness metrics to understand how agents perform in real customer interactions. The tool surfaces KPI trends over time and highlights ineffective topics, actions, or flows.

Agent Optimization traces session flows to reveal how agents make decisions and help teams diagnose issues that arise. Teams can use the tool to see how agents respond, step-by-step, across complex reasoning chains. The tool groups similar requests into clusters to uncover patterns, friction points and quality trends. It scores agent responses using intent, topic, and quality metrics, and teams can also use the tool to identify configuration issues that affect performance and pinpoint the need for tuning, retraining, or guardrails.

Finally, Agent Health Monitoring focuses on uptime, reliability, and responsiveness, and tracks key health metrics in near real-time. It also provides alerts on critical errors, latency spikes, and escalations so teams can detect, investigate, and resolve problems while minimizing downtime.

Case in point

1-800Accountant, a virtual accounting firm for small businesses, has been a beta customer of the new observability tools. Because of the nature of its business, the small firm is busiest February through April. Ryan Teeples, the companyโ€™s chief strategy officer, explains that automation and agentic AI have been essential to help it service customers in that period. It already has more than 20 agents in production. These agents work together to address some of 1-800Accountantโ€™s biggest pain points.

For instance, when one of the companyโ€™s salespeople meets with a new lead for the first time, they take the lead through an assessment to understand specific needs. That process is supported by a slate of agents that records and transcribes the conversation, analyzes it, summarizes it, and then dynamically generates an agenda, based on data from the conversation, for a potential follow-up onboarding appointment if the lead opts to become a client.

โ€œObservability allows us to have much faster throughput,โ€ Teeples says, adding that the tools allow his team to iterate and improve agents at a much faster pace. โ€œNow you have tools that allow you to monitor the interactions, get feedback, and get better prompting and better data to the agent. It accelerates the process of the agent and gives us more time on the development side to focus on better client experience, rather than on monitoring and ensuring the AI agents arenโ€™t going off the rails.โ€

Agent Analytics and Agent Optimization are both part of Agentforce Studio and are available today; regional rollout for APAC customers will happen on Friday. Agent Health Monitoring will be generally available in spring 2026.

Inside the product mindset that runs 7-Eleven

20 November 2025 at 05:00

In 2016, 7-Eleven began a digital transformation aimed at redefining convenience. The starting point was loyalty. โ€œStep one was to build a product discipline, bring the technology in house, and reduce reliance on third parties,โ€ says Scott Albert, VP and head of store and enterprise products.

Two years later, the Texas-based retailer reapplied the product playbook, now powering store systems across more than 13,000 US and Canadian locations. โ€œWe moved from projects โ€” start date, end date โ€” to product: continuous improvement and iteration,โ€ Albert says. โ€œFrom outputs to outcomes, co-owned with design and engineering.โ€

Albert knows the terrain. A company veteran who cut his teeth in operations, he led product for loyalty and now oversees digital product for store systems, fuel, restaurant concepts, and merchandising, evidence of how far the model has scaled.

Setting the foundation

The idea was straightforward but the shift wasnโ€™t. โ€œIt was tough early on because it meant change,โ€ Albert says. โ€œThe business was used to saying, โ€˜I need X.โ€™ Often that wasnโ€™t the real problem. Our job was to get underneath, understand the problem, design a solution for now and the future, and then iterate.โ€

It takes several ingredients to solve big problems, like customer research, business process knowledge, data, and technology, so itโ€™s natural that product teams are cross-functional. But that structure can also create competing priorities if not managed correctly. While the setting is convenience retail, the lesson applies to any CIO shifting from project-based delivery to product-driven transformation. โ€œSuccess depends not on org charts, but on cross-functional trust, buy-in, and commitment,โ€ he says.

That structure set the foundation, and the real breakthroughs came from applying product thinking to their daily work.

Product thinking in action

โ€œFor me and my team, the customer is the store associate,โ€ Albert says. That focus shaped priorities to remove low-value tasks, surface just-in-time insights, and let systems work for people, not the other way around.

The team learned this firsthand on midnight store walks. In one New York City visit, they noticed a new associate glued to her phone. โ€œWe thought she was distracted,โ€ Albert says. โ€œTurns out sheโ€™d recorded her trainer so she could remember.โ€ That single observation sparked a redesign of training to move job aids and how-to videos from a back-room PCs to mobile devices on the floor, embedded in the flow of work.

The same product instinct of watching users, identifying friction, and iterating has carried into 7-Elevenโ€™s AI initiatives. AI-assisted ordering, for example, reduced what was once up to 30 hours a week of manual work to under an hour a day, freeing up associates to focus on customers. At scale, those savings add up to more than 13 million hours reclaimed annually, and test-and-learn pilots tying the changes to about $340 million in incremental sales.

The back office has been transformed as well. After migrating store systems to the cloud with its 7-BOSS platform, 7-Eleven layered in โ€œquick cardsโ€ that surface AI-generated insights and let associates act in three clicks or less. A clustering model identifies lookalike stores by sales mix, location type, even seasonality, and pushes tailored assortment recommendations. โ€œWith three clicks, you can add an item, forecasting kicks in, and delivery happens in days,โ€ Albert says.

Together, these stories trace a clear pattern of observing the customer (in this case the store personnel), solving for their pain points, then amplifying the solution with data and AI. Itโ€™s product thinking at work.

Operating like a product company

Behind the scenes, the mechanics mirror digital natives. Teams run in pods with product, engineering, and design as a three-legged stool. Quarterly planning sets direction, but roadmaps flex. โ€œTell me everything youโ€™ll do next year โ€” that was the old model,โ€ Albert says. โ€œNow we focus on quarters, but sometimes thatโ€™s too long. We plan, then adapt.โ€

Release cadence has accelerated as well, from two or three big bangs a year to monthly releases.

The cultural shift is ongoing funding for work that never ends. โ€œThereโ€™s no such thing as done in product,โ€ he says. โ€œWeโ€™re on the fifth iteration of our forecasting model. Weโ€™ll keep improving.โ€

Start small, measure hard

Albertโ€™s advice to other tech executives: start small. โ€œFind a problem that matters, build a cross-functional team, measure success, and validate results,โ€ he says. โ€œThen add a second team, a third, and youโ€™re off.โ€

And above all, measure. โ€œPick metrics backed by data so no one can debate the results,โ€ he adds.

Nearly 10 years after its first loyalty decision, 7-Elevenโ€™s product mindset now extends far beyond consumer apps. The store itself has become a living product, updated monthly, informed by data, and built around the associate.

For Albert, the real measure of success is to make the system work for the associate, so they can delight customers. โ€œItโ€™s the same product discipline, now applied to every corner of the store, and itโ€™s redefining what convenience looks like at scale,โ€ he says.

โŒ
โŒ