Read AI, which made its mark analyzing online meetings and messages, is expanding its focus beyond the video call and the email inbox to the physical world, in a sign of the growing industry trend of applying artificial intelligence to offline and spontaneous work data.

The Seattle-based startup on Wednesday introduced a new system called Operator that captures and analyzes interactions throughout the workday, including impromptu hallway conversations and in-person meetings in addition to virtual calls and emails, working across a wide range of popular apps and platforms. 

With the launch, Read AI is releasing new desktop clients for Windows and macOS, and a new Android app to join its existing iOS app and browser-based features.

For offline conversations — like a coffee chat or a conference room huddle — users can open the Read AI app and manually hit record. The system then transcribes that audio and incorporates it into the company’s AI system for broader insights into each user’s meetings and workday.

It comes as more companies bring workers back to the office for at least part of the week. According to new Read AI research, 53% of meetings now happen in-person or without a calendar invite — up from 47% in 2023 — while a large number of workday interactions occur outside of meetings entirely.

In a break from others in the industry, Operator works via smartphone in these situations and does not require a pendant or clip-on recording device. 

“I don’t think we’d ever build a device, because I think the phones themselves are good enough,” said Read AI CEO David Shim in a recent interview, as featured on this week’s GeekWire Podcast.

This differs from hardware-first competitors like Limitless and Plaud, which require users to purchase and wear dedicated devices to capture “real-world” audio throughout the day.

While these companies argue that a wearable provides a frictionless, “always-on” experience without draining your phone’s battery, Read AI is betting that the friction of charging and wearing a separate gadget is a bigger hurdle than simply using the device you already have.

To address the privacy concerns of recording in-person chats, Read AI relies on user compliance rather than an automated audible warning. When a user hits record on the desktop or mobile app, a pop-up prompts them to declare that the conversation is being captured, via voice or text. On mobile, a persistent reminder remains visible on the screen for the duration of the recording.

Founded in 2021 by David Shim, Robert Williams, and Elliott Waldron, Read AI has raised more than $80 million and landed major enterprise customers for its cross-platform AI meeting assistant and productivity tools. It now reports 5 million monthly active users, with 24 million connected calendars to date.

Operator is included in all of Read AI’s existing plans at no additional cost.

[Editor’s Note: Agents of Transformation is an independent GeekWire series and 2026 event, underwritten by Accenture, exploring the people, companies, and ideas behind the rise of AI agents.]

What separates the dot-com bubble from today’s AI boom? For serial entrepreneur David Shim, it’s two things the early internet never had at scale: real business models and customers willing to pay.

People used the early internet because it was free and subsidized by incentives like gift certificates and free shipping. Today, he said, companies and consumers are paying real money and finding actual value in AI tools that are scaling to tens of millions in revenue within months.

But the Read AI co-founder and CEO, who has built and led companies through multiple tech cycles over the past 25 years, doesn’t dismiss the notion of an AI bubble entirely. Shim pointed to the speculative “edges” of the industry, where some companies are securing massive valuations despite having no product and no revenue — a phenomenon he described as “100% bubbly.”

He also cited AMD’s deal with OpenAI — in which the chipmaker offered stock incentives tied to a large chip purchase — as another example of froth at the margins. The arrangement had “a little bit” of a 2000-era feel of trading, bartering and unusual financial engineering that briefly boosted AMD’s stock.

But even that, in his view, is more of an outlier than a systemic warning sign.

“I think it’s a bubble, but I don’t think it’s going to burst anytime soon,” Shim said. “And so I think it’s going to be more of a slow release at the end of the day.”

Shim, who was named CEO of the Year at this year’s GeekWire Awards, previously led Foursquare and sold the startup Placed to Snap. He now leads Read AI, which has raised more than $80 million and landed major enterprise customers for its cross-platform AI meeting assistant and productivity tools.

He made the comments during a wide-ranging interview with GeekWire co-founder John Cook. They spoke about AI, productivity, and the future of work at a recent dinner event hosted in partnership with Accenture, in conjunction with GeekWire’s new “Agents of Transformation” editorial series.

We’re featuring the discussion on this episode of the GeekWire Podcast. Listen above, and subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen. Continue reading for more takeaways.

Successful AI agents solve specific problems: The most effective AI implementations will be invisible infrastructure focused on particular tasks, not broad all-purpose assistants. The term “agents” itself will fade into the background as the technology matures and becomes more integrated.

Human psychology is shaping AI deployment: Internally, ReadAI is testing an AI assistant named “Ada” that schedules meetings by learning users’ communication patterns and priorities. It works so quickly, he said, that Read AI is building delays into its responses, after finding that quick replies “freak people out,” making them think their messages didn’t get a careful read.

Global adoption is happening without traditional localization: Read AI captured 1% of Colombia’s population without local staff or employees, demonstrating AI’s ability to scale internationally in ways previous technologies couldn’t.

“Multiplayer AI” will unlock more value: Shim says an AI’s value is limited when it only knows one person’s data. He believes one key is connecting AI across entire teams, to answer questions by pulling information from a colleague’s work, including meetings you didn’t attend and files you’ve never seen.

“Digital Twins” are the next, controversial frontier: Shim predicts a future in which a departed employee can be “resurrected” from their work data, allowing companies to query that person’s institutional knowledge. The idea sounds controversial and “a little bit scary,” he said, but it could be invaluable for answering questions that only the former employee would have known.

Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you listen.

Every Friday at 5 p.m., Diego Oppenheimer gets an email that remembers his week better than he does. It pulls from his calendar, meeting transcripts, and inbox to figure out what really mattered: decisions made, promises to keep, and priorities for the week ahead.

“It gives me a superpower,” said Oppenheimer, a machine-learning entrepreneur best known as the co-founder of Algorithmia, who’s now working with startups as an investor in Seattle. 

What’s notable is that Oppenheimer didn’t buy this tool off the shelf — he built it. What started as a personal experiment turned into a challenge: could he still code after years away from writing production software?

With the rise of AI-powered coding assistants, he realized he could pick up where he left off. His personal project, with the unglamorous name “Actionary,” has grown to somewhere around 40,000 lines of what he jokingly calls vibe-coded “spaghetti.” It’s messy but functional.

Oppenheimer’s do-it-yourself AI assistant is more than a novelty. It’s a window into a broader shift. Individuals and companies are starting to hand off pieces of judgment and workflow to autonomous systems — software that analyzes data, makes recommendations, and acts independently.

Exploring the agentic frontier

This emerging frontier is the subject of Agents of Transformation, a new GeekWire editorial series exploring the people, companies, and ideas behind the rise of AI agents. A related event is planned for Seattle in early 2026. This independent project is underwritten by Accenture.

For this first installment, we spoke with startup founders and DIY builders working to replicate different aspects of the work of great executive assistants — coordinating calendars, managing travel, and anticipating needs — to see how close AI agents are getting to the human standard.

The consensus: today’s agents excel at narrow, well-defined tasks — but struggle with broader human judgment. Attempts to create all-purpose digital assistants often run up against the limits of current AI models. 

“I might have my travel agent and my finance agent and my stock trading agent and my personal health coach agent and my home chef agent, etc.,” said T.A. McCann, a Seattle-based serial entrepreneur and managing director at Pioneer Square Labs, on a recent GeekWire Podcast episode. 

McCann foresees these narrow agents handling discrete tasks, potentially coordinated by higher-level AI acting like a personal chief operating officer.

But even the term “AI agent” is up for debate. Oppenheimer defines a true agent as one with both autonomy and independent decision-making. By that standard, his system doesn’t quite qualify. It’s more a network of models completing tasks on command than a self-directed entity.

“If you asked a marketing department, they would say, absolutely, this is fully agentic,” he said. “But if I stick to my AI nerd cred, is there autonomous decision-making? Not really.”

It’s part of a much larger trend. The market for AI workplace assistants is projected to grow from $3.3 billion this year to more than $21 billion by 2030. according to MarketsandMarkets. Growth is being driven both by enterprise giants such as Microsoft and Salesforce embedding agents into workplace software, and by startups building specialized agents. 

A report by the newsletter “CCing My EA,” citing an ASAP survey, notes that 26% of EAs now use AI tools. Some fear job loss due to AI, but most top EAs see AI as an augmentation tool that frees time for strategic work. 

From summaries to scheduling

One company exploring this emerging frontier is Read AI, a Seattle-based startup known for its cross-platform AI meeting summarization and analysis technologies, which has raised more than $80 million in funding.

Co-founder and CEO David Shim revealed that Read AI has been internally developing and piloting an AI executive assistant called “Ada” for tasks including scheduling meetings and responding to emails.

Ada replies so quickly that Read AI has been working on building in a delay into the email response time so that it seems more natural to the recipients.

Shim has been personally testing the limits of the technology — giving Ada access to a range of workplace data (from Outlook, Teams, Slack, JIRA, and other cloud services) and letting the assistant autonomously answer questions about Read AI’s business that come in from the company’s investors in response to his periodic updates.

“It answers questions that I would not have the answer to right off the bat, because it’s not just pulling from my data set, but it’s pulling in from my team’s data set,” Shim said during a fireside chat with GeekWire co-founder John Cook at a recent Accenture reception.

Shim laughed, “I’m willing to take that risk. We’re doing well, so I don’t mind giving out the data.” 

However, there are limitations. Ada can struggle with complex multi-person scheduling or tasks requiring data it can’t access, and can still occasionally hallucinate. To manage this, ReadAI incorporates human oversight mechanisms like “sidebars” where Ada asks for confirmation before sending replies to messages deemed more sensitive or difficult. 

Shim argues against the idea of building a single, all-encompassing agent. 

“The approach of agents doing everything is not the right approach,” he said. “If you try to do everything, you’re not going to do anything well.”

Instead, he believes successful AI assistants will focus on solving very specific problems, much like Google Maps gives driving directions without trying to be a general travel agent. 

The “book-me-a-hotel” challenge

Travel is a use case that’s close to the heart of Brad Gerstner, founder and CEO of Altimeter Capital. Gerstner is known for backing some of the biggest names in tech — from Snowflake to Expedia — and for distilling big tech shifts into simple tests, such as his hotel booking challenge.

The specific example he gave at the 2024 Madrona IA Summit in Seattle was telling an AI agent to book the Mercer Hotel in New York on a specific day at the lowest price — a common challenge for business travelers.  

“Until we can do that, we have not built a personal assistant,” he said. 

That’s part of the larger problem Michael Gulmann, a former Expedia product executive, set out to solve with the startup Otto, which is developing an AI agent specifically for business travelers.

As shown publicly for the first time at this year’s Madrona conference, Otto tackled Gerstner’s specific challenge. After receiving the request to book the Mercer Hotel on a specific day, it found the cheapest available room, confirmed the price and details, and completed the booking, with minimal prompting, within about two minutes.

“Who would have thought that Brad Gerstner wanted the cheapest room?” Gullman joked.

Otto handles various aspects of travel. It understands and learns detailed user preferences — from specific amenities like rooftop bars to preferred airline seats, hotel room types, and loyalty programs — using this knowledge to refine searches and make personalized recommendations. 

As Gulmann explained in an interview, Otto doesn’t use a single monolithic model. It coordinates a bunch of narrow agents: one to interpret messages, another to manage loyalty programs, another to handle payments. Together they simulate a small operations team working behind the scenes.

Otto confirms details with the user before completing purchases, even though it could do that autonomously. Gulmann described that precaution as psychological, not technical — knowing that most people aren’t yet comfortable with AI buying things without their involvement.

After learning about Otto’s capabilities, Gerstner was impressed and wanted to see how it performs as it moves into public beta, said Mike Fridgen, a venture partner at Madrona, which incubated the company.

The grand challenge of scheduling

If hotel booking is the acid test for autonomous assistants, scheduling meetings is the everyday nightmare.

That’s the problem Howie is trying to solve. The Seattle startup’s AI assistant lives in the email inbox. CC Howie on a thread, and it proposes times, confirms with all parties, creates invites, and adds meeting links.

Howie works from a detailed “preferences document,” inspired by how experienced executives train their human EAs — which cafés are acceptable for meetings, how late is too late on Fridays, etc.

The company recently launched publicly with $6 million in funding and a growing number of paying customers. It uses a hybrid model: AI supported by human reviewers. That helps avoid the tiny errors that destroy trust — mixing up time zones, dropping a name from a thread, or misreading social cues.

The system simulates decisions internally, flags potential errors for review, and escalates anything ambiguous to a human before hitting send. 

“If you think about the things that a great human EA does, software is not replacing that anytime soon,” said Howie co-founder Austin Petersmith.

In fact, Petersmith said, many of Howie’s users are human EAs themselves, using it to offload logistics. “Nobody wants to do scheduling,” he said. “Everybody wants the machines to take this particular task on.”

As models improve, Petersmith hopes Howie can expand into other “meta-work” — the administrative overhead that keeps knowledge workers from the higher-value activities that are still the realm of humans.

More time in the day

For Diego Oppenheimer, this isn’t a hypothetical issue. “I’m extremely calendar dyslexic,” he explained. “I’ll triple-book myself. I’ll agree to go to places I shouldn’t be. I’ll travel to the wrong city. Really bad.”

Over the years, he relied on human EAs and a chief of staff to keep him on track. But when he stepped back from running a company full-time, hiring someone just to manage his complex, multi-role calendar no longer made sense. So he built Actionary to help. It sends the Friday recap to catch him up on the week, flagging issues right before his weekend “reboot.”

Oppenheimer’s project won the People’s Choice Award at an AI Tinkerers event in New York last month. But he is very clear: Actionary is a personal project, not a product in the making. He developed it for himself, and can’t imagine taking on the headache of feature requests and technical support from others.

He’s bullish on the larger trend, and a user and investor in tools like Howie. But he also recognizes that AI agents can’t match the comprehensive skills and judgment of a human EA, let alone a chief of staff in a higher-level strategic role.

Oppenheimer’s ultimate goal is more straightforward, but still ambitious. “I’m trying to make time in the day,” he said. “That’s what I’m trying to do.” 

GeekWire’s Todd Bishop reported and wrote this article with editing assistance from AI tools including Gemini and a custom OpenAI GPT trained in GeekWire’s editorial approach. All facts, quotes, and conclusions were reviewed and verified prior to publication.