Circle’s chief executive painted a brisk picture at Davos this week: autonomous software agents that act for people could be using stablecoins to pay for everyday things within three to five years.
He said these agents will need a money system that is stable, fast, and programmable. That, he argued, points to stablecoins as the likely choice.
AI Agents And Money
According to reports, Jeremy Allaire of Circle said “literally billions” of AI agents may be transacting on behalf of users in the near term.
“Three years, five years from now, one can expect that there will be billions, literally billions of AI agents conducting economic activity in the world on a continuous basis,” Allaire said during the World Economic Forum in Davos, Switzerland.
He described work on new networks and tools aimed at letting software act like small businesses or helpers that buy services, settle bills, and tip content creators.
This idea is simple on the surface: software needs a reliable unit of account when it spends, and tokenized dollars can fit that role.
Building The Tools
Reports say companies across the crypto and tech world are racing to build the plumbing for this future. Circle is pitching USDC as a neutral payments layer that software can plug into.
Other firms are testing protocols that let a machine sign off on a payment when certain conditions are met. Some large tech groups are also exploring ways for their platforms to let software pay for services automatically. Progress is visible, but the path is not yet clear.
What Regulators Might Ask
Regulators will have questions. Reports note concerns about money flow, consumer protections, and where bank deposits sit if stablecoins grow rapidly.
At Davos, the CEO pushed back on the idea that stablecoins would drain bank deposits the way some fear, saying comparisons to other financial instruments are more fitting.
Still, lawmakers in the US and elsewhere are watching closely. Rules could move faster if policy makers see real volume coming from so-called agentic commerce.
New Networks, New Risks
Based on reports, the technical choices will shape both convenience and danger. If agents can move value at scale, fraud and theft risks may rise too.
Systems will need clear identity checks, fault handling, and ways to stop runaway payments. Some safety work is already under way, but much remains to be designed and tested.
Featured image from Pexels, chart from TradingView
On Tuesday, eBay updated its User Agreement to explicitly ban third-party "buy for me" agents and AI chatbots from interacting with its platform without permission, first spotted by Value Added Resource. On its face, a one-line terms of service update doesn't seem like major news, but what it implies is more significant: The change reflects the rapid emergence of what some are calling "agentic commerce," a new category of AI tools designed to browse, compare, and purchase products on behalf of users.
eBay's updated terms, which go into effect on February 20, 2026, specifically prohibit users from employing "buy-for-me agents, LLM-driven bots, or any end-to-end flow that attempts to place orders without human review" to access eBay's services without the site's permission. The previous version of the agreement contained a general prohibition on robots, spiders, scrapers, and automated data gathering tools but did not mention AI agents or LLMs by name.
At first glance, the phrase "agentic commerce" may sound like aspirational marketing jargon, but the tools are already here, and people are apparently using them. While fitting loosely under one label, these tools come in many forms.
ZEST Security introduces AI Sweeper Agents that identify which vulnerabilities are truly exploitable, helping security teams cut patch backlogs and focus on real risk.
Microsoft CEO Satya Nadella and former UK Prime Minister Rishi Sunak at the World Economic Forum in Davos. (Screenshot via LinkedIn)
Bicycles for the mind. … Information at your fingertips. … Managers of infinite minds?
Microsoft CEO Satya Nadella riffed on some famous lines from tech leaders past this week in an appearance at the World Economic Forum in Davos, Switzerland, and offered up his own trippy candidate to join the canon of computing metaphors.
Nadella traced the lineage in a conversation with former UK Prime Minister Rishi Sunak.
“Computers are like a bicycle for the mind” was the famous line from Apple’s Steve Jobs.
“Information at your fingertips” was Bill Gates’ classic Microsoft refrain back in the day.
And now? “All of us are going to be managers of infinite minds,” Nadella said. “And so if we have that as the theory, then the question is, what can we do with it?”
He was referring to AI agents — the autonomous software that can take on tasks, work through problems, and keep going while you sleep. Microsoft and others have been talking for the better part of a year now about people starting to oversee large fleets of them.
Nadella said it’s already reshaping how teams are structured. At Microsoft-owned LinkedIn, the company has merged design, program management, product management, and front-end engineering into a single new role: full-stack builders. Overall, he called it the biggest structural change to software teams he’s seen in a career that started at Microsoft in the 1990s.
“The jobs of the future are here,” Nadella said, putting his own spin on a famous line often attributed to sci-fi writer William Gibson. “They’re just not evenly distributed.”
Nadella’s comments came during a live stream for LinkedIn Premium members, hosted from Davos by LinkedIn VP and Editor in Chief Daniel Roth, after Sunak mentioned his two teenage daughters, and the world they’ll enter. Young people may not manage lots of people at age 20 or 21, he said, “but they will be managing a team of agents.”
Sunak was referencing an essay by Goldman Sachs CIO Marco Argenti in Time.
The agentic shift, Argenti wrote, requires “moving from being a sole performer to an orchestra conductor” — your team now includes AI agents that “must be guided and supervised with the same approach you would apply to a new, junior colleague.”
Nadella agreed, saying “we do need a new theory of the mind” to navigate what’s coming, before he offered up his new metaphor about managing infinite minds.
In other remarks at Davos, Nadella made headlines with his warning that AI’s massive energy demands risk eroding its “social permission” unless it delivers tangible benefits in health, education, and productivity. Energy costs, he added, will decide the AI race’s winners, with GDP growth tied to cheap power for processing AI tokens.
Whether “infinite minds” catches on like “bicycles” and “fingertips” remains to be seen. But it’s definitely more psychedelic. And if this shift is stranger than what came before, maybe we do need a mind-expanding metaphor to make sense of it all.
Misaligned agents are just one layer of the AI security challenge that startup Witness AI is trying to solve. It detects employee use of unapproved tools, blocks attacks, and ensures compliance.
PromptArmor threat researchers uncovered a vulnerability in Anthropic's new Cowork that already was detected in the AI company's Claude Code developer tool, and which allows a threat actor to trick the agent into uploading a victim's sensitive files to their own Anthropic account.
As companies deploy AI-powered chatbots, agents, and copilots across their operations, they’re facing a new risk: how do you let employees and AI agents use powerful AI tools without accidentally leaking sensitive data, violating compliance rules, or opening the door to prompt-based injections? Witness AI just raised $58 million to find a solution, building what they call “the […]
We’re excited to announce a new GeekWire event for 2026: “Agents of Transformation: Inside the AI Shift.” This half-day summit will be held the afternoon of Tuesday, March 24, in Seattle, exploring how agentic AI is reshaping work, creativity, and leadership.
The event, presented by Accenture, features fireside chats, expert panels, and real-world stories from technology leaders, business execs, and others navigating how AI is changing the way we work and lead, from copilots and automation to the rise of intelligent agents.
Tickets are available now, with discounted early bird rates set to end Feb. 24. Speakers will be announced in the coming weeks.
AI agents is the tech industry’s obsession right now, but there can be a big gap between the pitch and the reality. We’re bringing together people who are in the thick of it to talk candidly about what they’re seeing: breakthroughs, challenges, and what comes next.
The event is part of GeekWire’s longstanding tradition of convening tech, business, and policy leaders for insights and new connections. Hosted at one of our favorite Seattle venues, Block 41, the afternoon will include networking opportunities before, during, and after the program, bringing together founders, executives, and technologists from across the region.
It builds on an ongoing GeekWire editorial series, underwritten by Accenture, spotlighting how startups, developers and tech giants are using intelligent agents to innovate.
Microsoft’s Copilot Checkout lets users browse and buy products without leaving the chat. (Microsoft Image, click for larger version)
[Editor’s Note: Agents of Transformation is an independent GeekWire series and March 24, 2026 event, underwritten by Accenture, exploring the people, companies, and ideas behind AI agents.]
Microsoft is making its own bid to turn AI conversations into agentic commerce, announcing a new feature called Copilot Checkout that lets users complete purchases directly within its AI chatbot, without being redirected to an external website.
The company is betting that its existing enterprise technology footprint and established relationships with large retailers will give it an edge over OpenAI, Google, and Amazon in winning over merchants wary of giving up control to retail rivals or AI intermediaries.
Kathleen Mitford, Microsoft corporate vice president of global industry marketing. (Microsoft Photo)
“We’ve designed it in such a way that retailers own those relationships with the customers,” said Kathleen Mitford, corporate vice president of global industry marketing at Microsoft. “It is their data, it is their relationship, and that’s something that’s really important to us.”
It’s part of a broader AI rollout by Microsoft at NRF 2026, the retail industry’s annual conference in New York. Microsoft is also launching Brand Agents, pitched as a complete solution for Shopify merchants to add AI assistants to their websites, along with new AI tools to assist store employees and help retailers enhance their online product listings and metadata.
Copilot Checkout works by surfacing products from partner retailers within Copilot search results. Purchases can be completed without leaving the conversation. Microsoft says the retailer remains the merchant of record, handling fulfillment and customer service.
But will people buy in chat?
The bigger question for the tech industry is whether chat-based commerce is actually the next big thing. Forrester analyst Sucharita Kodali, for example, previously told GeekWire that “e-commerce isn’t a problem that needs to be fixed.” She added that it’s unclear what value chat-based commerce is bringing to retailers, “other than disintermediating Google.”
Microsoft’s Mitford offered a different take in an interview this week, saying that consumer behavior is shifting faster than it may seem. She drew a parallel to how quickly businesses moved from experimenting with AI to putting it into operation over the past year.
“I see the same thing happening with consumers … it just takes a little bit of time,” Mitford said, predicting that the speed of consumer adoption will eventually match the rapid uptake seen in the business world.
Copilot Checkout is rolling out now in the U.S. on Copilot.com, with PayPal, Shopify, and Stripe handling payment processing. Etsy sellers will be among the first available on the platform. Shopify merchants are set to be automatically enrolled following an opt-out window.
That last detail is notable given the backlash Amazon has faced over its “Buy for Me” feature, where brands complained about being included without consent and seeing inaccurate listings.
Microsoft’s approach is more tightly connected to its partners — the company said Shopify will management the opt-out process for its merchants — but automatic enrollment seems to raise the potential for some of the same concerns. (We’ve contacted Shopify for more information.)
The competitive landscape
More broadly, Microsoft is playing catch-up on the consumer side.
OpenAI launched Instant Checkout in ChatGPT last September, partnering with Shopify and Stripe to let users buy from more than a million merchants. Google followed in November with its own “Buy for Me” feature which lets its Gemini assistant purchase products on a user’s behalf.
Despite its inroads with businesses, Copilot has a fraction of ChatGPT’s market share with consumers. Recent data from Similarweb’s Global AI Tracker showed ChatGPT with about 68% of AI chatbot web traffic, with Google Gemini at 18% and Copilot in the single digits.
But Microsoft has its advantages: Unlike Amazon and Google, which compete directly with retailers through their own marketplaces, it isn’t a retailer. And retail has long been a major vertical for its enterprise cloud and software business, with large chains running on Azure and Microsoft 365.
Mitford said Microsoft is leaning on its existing trust and long-standing relationships with retailers, along with a commitment to responsible AI, to help differentiate itself from rivals.
Microsoft is making the broader case for AI to retailers based on return on investment. A Microsoft-commissioned study from IDC, released in November, found that retail and consumer packaged goods companies are seeing a 2.7x return on every dollar spent on generative AI.
Mitford, a former fashion designer who has been in the technology industry for most of her career, said she sees the retail sector among the leaders in AI uptake across the business world.
The technology, she said, is being “adopted at a pace that I’ve never seen.”
The cover of Microsoft’s 1990 annual report, showing Microsoft Word for Windows 3.0, reflected the company’s confidence as Windows was emerging as a true platform.
[Editor’s Note: Agents of Transformation is an independent GeekWire series and March 24, 2026 event, underwritten by Accenture, exploring the people, companies, and ideas behind AI agents.]
It was “like bringing a Porsche into a world of Model Ts.”
That’s what Microsoft said in its 1990 annual report about the shift from MS-DOS to Windows. But the bigger breakthrough for the company wasn’t the graphical interface. It was Windows’ ability to serve as a platform for applications made by others.
Windows 3.0, released that year, made third-party software easier to find and launch, and offered developers a clear bargain: build to Microsoft’s specs, and your software would become a first-class citizen on the computers that were arriving “on every desk and in every home,” as the company’s original mission statement put it.
Thirty-five years later, AI feels less like a car and more like a rocket ship. But Microsoft is hoping that Windows can once again serve as the platform where it all takes off.
A new framework called Agent Launchers, introduced earlier this month as a preview in the latest Windows Insider build, lets developers register agents directly with the operating system. They can describe an agent through what’s known as a manifest, which then lets the agent show up in the Windows taskbar, inside Microsoft Copilot, and across other apps.
The long-term promise for Windows users is autonomous assistants that operate on their behalf, directly on their machines. Beyond routine tasks like assembling a PDF or organizing files, agents could monitor email and calendars to resolve scheduling conflicts, or scan documents across multiple apps to pull together a briefing for an upcoming meeting.
Achieving that level of autonomy requires more than just a clever interface. It will take deep, persistent memory that operates more like the human brain.
Microsoft CEO Satya Nadella this week framed AI agents as a new layer of computing infrastructure that requires greater engineering sophistication. Windows is one of the places where Microsoft is attempting to implement that vision. (GeekWire File Photo / Kevin Lisota)
“We are now entering a phase where we build rich scaffolds that orchestrate multiple models and agents; account for memory and entitlements; enable rich and safe tools use,” Microsoft CEO Satya Nadella wrote in a blog post this week looking ahead to 2026. “This is the engineering sophistication we must continue to build to get value out of AI in the real world.”
Elements of this are already emerging elsewhere.
Google’s Gemini and Anthropic’s Claude offer desktop-style agents through browsers and native apps, with extensions that can read pages, fill forms, and take limited actions on a user’s behalf.
Amazon is developing “frontier agents” aimed at automating business processes in the cloud.
Startups like Seattle-based Vercept are building standalone agentic apps that coordinate work across tools.
But Microsoft’s Windows team is betting that agents tightly linked to the operating system will win out over ones that merely run on top of it, just as a new class of Windows apps replaced a patchwork of DOS programs in the early days of the graphical operating system.
Microsoft 365 Copilot is using the Agent Launchers framework for first-party agents like Analyst, which helps users dig into data, and Researcher, which builds detailed reports. Software developers will be able to register their own agents when an app is installed, or on the fly based on things like whether a user is signed in or paying for a subscription.
The risks posed by PC agents
The parallels to the past only go so far. Traditional PC applications ran in their own windows, worked with their own files, and didn’t touch the rest of the system for the most part.
“Agents are going to need to be able to scratchpad their work,” Microsoft CTO Kevin Scott said recently on the South Park Commons Minus 1 podcast, explaining that agents will need to retain a history of user interactions and tap into the necessary context to solve problems.
Agents are meant to maintain this context across apps, ask follow-up questions, and take actions on a user’s behalf. That requires a different level of trust than Windows has ever had to manage, which is already raising difficult questions for the company.
Microsoft acknowledges that agents introduce unique security risks. In a support document, the company warned that malicious content embedded in files or interface elements could override an agent’s instructions — potentially leading to stolen data or malware installation.
To address this, Microsoft says it has built a security framework that runs agents in their own contained workspace, with a dedicated user account that has limited access to user folders. The idea is to create a boundary between the agent and what the rest of the system can access.
The agentic features are off by default, and Microsoft is advising users to “understand the security implications of enabling an agent on your computer” before turning them on.
A different competitive landscape
Even if Microsoft executes perfectly, the landscape is different now. In the early 1990s, Windows became dominant because developers flocked to the platform, which attracted more users, which attracted more developers. It was a virtuous cycle, and Microsoft was at the heart of it.
But Windows isn’t the center of the computing world anymore. Smartphones, browsers, and cloud platforms have fragmented the landscape in ways that didn’t exist back then. Microsoft missed the mobile era almost entirely, and the PC is now one screen among many.
In the enterprise, Microsoft has better footing. Azure, Microsoft 365 Copilot, and a growing ecosystem of business-focused agents give the company a strong position, competing against Google, Amazon, OpenAI and others for cloud-based AI agents and services.
Agent Launchers is a different bet — an attempt to make Windows the home for agents that serve individual users on their own machines. That’s a harder sell when the PC is competing with phones, browsers, and cloud apps for people’s attention. Microsoft can build the platform, but it can’t guarantee that developers will show up the way they did 35 years ago.
And unlike in the 1990s, Microsoft can’t count on users to embrace what it’s building. There’s a growing sentiment that these AI capabilities are being pushed into Windows not because users want them, but because Microsoft needs to justify its massive AI investments.
In October, for example, Microsoft announced new features including “Hey Copilot” voice activation, a redesigned taskbar with Copilot built in, and the expansion of “Copilot Actions” agentic capabilities beyond the browser to the PC itself.
“They’re thinking about revenue first and foremost,” longtime tech journalist and Microsoft observer Ed Bott said on the GeekWire Podcast at the time. The more users rely on these AI features, he explained, the easier it becomes for the company to upsell them on premium services.
There is a business reality driving all of this. In Microsoft’s most recent fiscal year, Windows and Devices generated $17.3 billion in revenue — essentially flat for the past three years.
That’s less than Gaming ($23.5 billion) and LinkedIn ($17.8 billion), and a fraction of the $98 billion in revenue from Azure and cloud services or the nearly $88 billion from Microsoft 365 commercial.
By comparison, in fiscal 1995, five years after the launch of Windows 3.0, Microsoft’s platforms group (which included MS-DOS and Windows) represented about 40% of its total revenue of $5.9 billion. Windows was the growth engine for the company.
Windows is unlikely to play that kind of outsized role again. But AI integration is the company’s best bet to return the OS to growth. Whether that ultimately looks like a restored Porsche or a rocket ship on the launchpad probably doesn’t matter as much as keeping it out of the junkyard.
Microsoft CTO Kevin Scott has some advice for AI startups waiting for the next breakthrough model: the technology can already do far more than most people are getting out of it, so stop waiting and start building.
Also: real customer traction still matters more than online buzz.
Speaking at a recent South Park Commons event with the organization’s general partner, former Dropbox CTO and Facebook engineer Aditya Agarwal, Scott said founders are sitting on a “gigantic capability overhang” — meaning that current AI systems can do far more than most apps built on top of them.
He cited ChatGPT itself as a past example: the underlying model was “pretty old” when it launched, as he put it, and nobody (including Scott and his peers) predicted at the time it would become a potential trillion-dollar product.
“The cost of doing the experiments has never been cheaper,” Scott said. “So do the damned experiments. Try things.”
The barrier isn’t model capability, he said, but the unglamorous integration work needed to put it to practical use.
“Some of the things that you need to do to squeeze the capability out of these systems is just ugly-looking plumbing stuff, or grungy product building,” he said. “But you’re in a startup, that’s kind of your life. It’s more about the grind.”
Scott also cautioned founders against mistaking online attention for real traction. The current environment, he said, is flooded with “false signal” — from media coverage to investor interest — that doesn’t really correlate with whether you’ve built something useful.
“You’ve got a bunch of people whose business model is getting clicks on articles online or getting people to subscribe to their Substack,” he said. “If you believe the things that particular part of the ecosystem is sending to you in terms of feedback, it could be that you’re steering yourself in exactly the wrong direction.”
The real signal, he said, comes from building something customers actually love.
Other topics included:
Open-source vs. closed-source models (he effectively framed this as a toolbox, not a battle, and said Microsoft uses both).
The importance of expert feedback in AI training, which he views as a potential startup advantage.
The infrastructure challenge of building memory systems for AI agents, a problem he said won’t be solved by simply training bigger models.
See the full talk above or on the South Park Commons Minus One Podcast.
The 2025 Uncommon Thinkers on stage at the GeekWire Gala. From left: Anindya Roy (Lila Biologics), Kiana Ehsani (Vercept), Max Blumen (Tin Can, accepting for co-founder Chet Kittleson), Jay Graber (Bluesky), Brian Pinkard (Aquagga), and Jeff Thornburg (Portal Space Systems). (GeekWire Photo / Kevin Lisota)
At the GeekWire Gala this week, we spent time talking backstage with five of this year’s Uncommon Thinkers — the inventors, scientists, and entrepreneurs who were selected in partnership with Greater Seattle Partners for their work transforming industries and the world.
You can hear the full conversations on this week’s episode of the GeekWire Podcast. As I mentioned at the end, I came away with an unexpected sense of optimism.
Jeff Thornburg of Portal Space Systems spent years building rocket engines for Elon Musk at SpaceX and Paul Allen at Stratolaunch. Now he and his team are reviving a NASA concept from decades ago: spacecraft propelled by focused sunlight.
Jeff Thornburg, CEO of Portal Space Systems, addresses the audience while being recognized as a 2025 Uncommon Thinker at the GeekWire Gala. (GeekWire Photo / Kevin Lisota)
When I asked what the world will look like “if Portal succeeds,” he made a classic entrepreneurial pivot: “When we’re successful,” he said, “we become the backbone of Earth-Moon logistics.”
From there, he said, it’s about protecting orbits for commerce, supporting human presence on the moon, and eventually pushing out to Jupiter’s moons.
Anindya Roy of Lila Biologics is using AI to design proteins from scratch — molecules that have never existed in nature — to fight cancer. He trained in David Baker’s Nobel Prize-winning lab at UW, so he saw the before and after of machine learning’s impact on the field.
Anindya Roy of Lila Biologics on stage at the GeekWire Gala, where he was honored as a 2025 Uncommon Thinker. (GeekWire Photo / Kevin Lisota)
Before: success rates below 1%, ordering hundreds of thousands of designs to find one that worked. Now: 5-20% success rates, ordering a few hundred designs to find a drug candidate.
“If you told me a couple of years ago that we can design an antibody from a computer, I would not believe you,” he said.
Jay Graber of Bluesky runs the decentralized social network that has become a leading alternative to X. But while most tech CEOs build moats, she and her team are building a protocol designed to help users leave.
Jay Graber, CEO of Bluesky, is recognized as a 2025 Uncommon Thinker during the GeekWire Gala. (GeekWire Photo / Kevin Lisota)
She talks about Bluesky and the underlying AT Protocol as a “collective organism,” and describes her role as guiding and stewarding the ecosystem rather than controlling it.
The industry and the world would be better off, she says, if leaders would think about their role “more as guides and stewards, rather than just dictators or emperors as they like to style themselves.”
Kiana Ehsani of Vercept came to Seattle from Iran for her PhD, spent four years at the Allen Institute for AI, and is now competing with OpenAI and Google in the AI agent space with a fraction of their resources.
Kiana Ehsani, CEO of Vercept, accepts her 2025 Uncommon Thinker award on stage at the GeekWire Gala. (GeekWire Photo / Kevin Lisota)
The ultimate vision is to help people move beyond mouse, keyboard, and touchscreen, letting them interact with computers the way they’d talk to a coworker.
AI agents are still early, she cautions. “Think of ChatGPT three years ago. Don’t think of it today.” Her advice for getting started with AI agents: “Start small, start with simple tasks that you don’t want to do, and then slowly build on top of it to see the magic.”
Brian Pinkard of Aquagga is tackling forever chemicals, the PFAS compounds that have spread through our water, food chain, and bloodstreams. The industry standard is to filter them out and then landfill or incinerate the waste, approaches that don’t truly solve the problem and can simply move it elsewhere.
Brian Pinkard, CTO of Aquagga, speaks on stage at the GeekWire Gala after being named a 2025 Uncommon Thinker. (GeekWire Photo / Kevin Lisota)
Aquagga uses technology originally designed to destroy chemical weapons to break PFAS down into inert salts under extreme heat and pressure. Pinkard didn’t believe it was possible until he saw the data. “I’m a skeptic, I’m cynical, I’m a scientist,” he said. “I wanted to see proof.”
His bigger vision is to transform hazardous waste processing entirely. Today, huge volumes of wastewater are trucked to incinerators and burned — which he calls “thermodynamic insanity.”
We’ll speak on a future episode with our sixth honoree, Chet Kittleson, co-founder and CEO of Tin Can, the startup making WiFi-enabled landline phones to help kids connect without screens.
Colleen Aubrey, AWS senior vice president of Applied AI Solutions, speaks during the AWS re:Invent keynote about the company’s push toward AI “teammates” and agentic development. (Amazon Photo)
LAS VEGAS — Speaking this week on the Amazon Web Services re:Invent stage, AWS executive Colleen Aubrey delivered a prediction that doubled as a wake-up call for companies still thinking of AI as just another tool.
“I believe that over the next few years, agentic teammates can be essential to every team — as essential as the people sitting right next to you,” Aubrey said during the Wednesday keynote. “They will fundamentally transform how companies build and deliver for their customers.”
But what does that look like in practice? On her own team, for example, Aubrey says she challenged groups that once had 50 people taking nine months to deliver a new product to do the same with 10 people working for three months.
Meanwhile, non-engineers such as finance analysts are building working prototypes using AI tools, contributing code in Amazon’s Kiro agentic development tool alongside engineers, and feeding those prototypes into Amazon’s famous PR/FAQ planning process on weekly cycles.
Those are some of the details that Aubrey shared when we sat down with her after the keynote at the GeekWire Studios booth in the re:Invent expo hall to dig into the themes from her talk. Aubrey is senior vice president of Applied AI Solutions at AWS, overseeing the company’s push into business applications for call centers, supply chains, and other sectors.
Continue reading for takeaways from the conversation, watch the video below, and listen to the conversation starting in the second segment of this week’s GeekWire Podcast.
The ‘teammate’ mental model changes everything. Aubrey draws a clear line between single-purpose AI tools that do one thing well and the agentic teammates she sees emerging — systems that take responsibility for whole objectives, and require a different kind of management.
“I think people will increasingly be managers of AI,” she said. “The days of having to do the individual keystrokes ourselves, I think, are fast fading. And in fact, everyone is going to be a manager now. You have to think about prioritization, delegation, and auditing. What’s the quality of our feedback, providing coaching. What are the guardrails?”
Amazon Connect crosses $1 billion. AWS’s call center platform reached $1 billion in annual revenue on a run rate basis, with Aubrey noting it has accelerated year-over-year growth for two consecutive years.
This week at re:Invent, the team announced 29 new capabilities across four areas: Nova Sonic voice interaction that Aubrey says is “very close to being indistinguishable” from human conversation; agents that complete tasks on behalf of customers; clickstream intelligence for product recommendations; and observability tools for inspecting AI reasoning.
One interesting detail: Aubrey said she’s often surprised by Nova Sonic’s sophistication and empathy in complex conversations — and equally surprised when it fails at basic tasks like spelling an address correctly.
“There’s still work to do to really polish that,” she said.
The ROI question gets a “yes and no.” Asked whether companies are seeing the business value to justify AI agent investments, Aubrey offered a nuanced response. “I observe companies to struggle to realize the business impact,” she said. But she said the value often shows up as eliminating bottlenecks — clearing backlogs, erasing technical debt, accelerating security patching — rather than immediate revenue gains.
“I’m not going to see the impact on my P&L today,” she said, “but if I fast forward a year, I’m going to have a product in market where real customers are using and getting real value, and we’re learning and iterating where I might not have even been halfway there in the past.”
Her advice for companies still hesitating: “If you don’t start today, that’s a one way door decision… I think you have to start the journey today. I would suggest people get focused, they get moving, because if you don’t, I think that becomes existential.”
Trust requires observability. Aubrey says companies won’t get full value from AI teammates if they can’t see how they’re reasoning.
“If you don’t trust an AI teammate, then you’re never going to realize the full benefit,” she said. “You’re not going to give them the hard tasks, you’re not going to invest in their development.”
The solution is treating AI inspection the same way you’d manage a human colleague: understand why it took an action, audit the quality, and iterate.
“You can refine your knowledge bases. You can refine your workflows. You can refine your guardrails, and then confidently keep iterating… the same way we do with each other. We keep iterating, we keep learning, and we keep getting better,” she said.
Product updates: Beyond Connect, Aubrey offered updates on other parts of her portfolio of Amazon’s applied AI solutions.
Just Walk Out, Amazon’s cashierless checkout technology, deployed more than 150 new stores in 2025 and should accelerate next year.
AWS Supply Chain, meanwhile, is getting a reset. “I’m going to declare that a pivot,” she said, with a Q1 announcement coming around agentic decision-making for supply and demand planning.
Also coming in Q1: a life sciences product focused on antibody discovery, currently in beta.
She teased “a few other new investment areas” expected to come in early 2026.
Amazon is experimenting again. This week on the GeekWire Podcast, we dig into our scoop on Amazon Now, the company’s new ultrafast delivery service. Plus, we recap the GeekWire team’s ride in a Zoox robotaxi on the Las Vegas Strip during Amazon Web Services re:Invent.
In our featured interview from the expo hall, AWS Senior Vice President Colleen Aubrey discusses Amazon’s push into applied AI, why the company sees AI agents as “teammates,” and how her team is rethinking product development in the age of agentic coding.
Login
What Regulators Might Ask
