[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 as we speak’s AI increase? For serial entrepreneur David Shim, it’s two issues the early web by no means had at scale: actual enterprise fashions and prospects prepared to pay.
Individuals used the early web as a result of it was free and backed by incentives like reward certificates and free transport. At present, he mentioned, firms and shoppers are paying actual cash and discovering precise worth in AI instruments which might be scaling to tens of thousands and thousands in income inside months.
However the Learn AI co-founder and CEO, who has constructed and led firms by means of a number of tech cycles over the previous 25 years, doesn’t dismiss the notion of an AI bubble fully. Shim pointed to the speculative “edges” of the trade, the place some firms are securing huge valuations regardless of having no product and no income — a phenomenon he described as “100% bubbly.”
He additionally cited AMD’s cope with OpenAI — during which the chipmaker supplied inventory incentives tied to a big chip buy — as one other instance of froth on the margins. The association had “somewhat bit” of a 2000-era really feel of buying and selling, bartering and strange monetary engineering that briefly boosted AMD’s inventory.
However even that, in his view, is extra of an outlier than a systemic warning signal.
“I believe it’s a bubble, however I don’t assume it’s going to burst anytime quickly,” Shim mentioned. “And so I believe it’s going to be extra of a gradual launch on the finish of the day.”
Shim, who was named CEO of the 12 months at this yr’s GeekWire Awards, beforehand led Foursquare and offered the startup Positioned to Snap. He now leads Learn AI, which has raised greater than $80 million and landed main enterprise prospects for its cross-platform AI assembly assistant and productiveness instruments.
He made the feedback throughout a wide-ranging interview with GeekWire co-founder John Cook dinner. They spoke about AI, productiveness, and the way forward for work at a current dinner occasion hosted in partnership with Accenture, along side GeekWire’s new “Brokers of Transformation” editorial collection.
We’re that includes the dialogue on this episode of the GeekWire Podcast. Pay attention above, and subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you hear. Proceed studying for extra takeaways.
Profitable AI brokers remedy particular issues: The simplest AI implementations might be invisible infrastructure centered on specific duties, not broad all-purpose assistants. The time period “brokers” itself will fade into the background because the expertise matures and turns into extra built-in.
Human psychology is shaping AI deployment: Internally, ReadAI is testing an AI assistant named “Ada” that schedules conferences by studying customers’ communication patterns and priorities. It really works so shortly, he mentioned, that Learn AI is constructing delays into its responses, after discovering that fast replies “freak individuals out,” making them assume their messages didn’t get a cautious learn.
World adoption is going on with out conventional localization: Learn AI captured 1% of Colombia’s inhabitants with out native workers or workers, demonstrating AI’s capability to scale internationally in methods earlier applied sciences couldn’t.
“Multiplayer AI” will unlock extra worth: Shim says an AI’s worth is proscribed when it solely is aware of one particular person’s information. He believes one secret’s connecting AI throughout whole groups, to reply questions by pulling data from a colleague’s work, together with conferences you didn’t attend and recordsdata you’ve by no means seen.
“Digital Twins” are the subsequent, controversial frontier: Shim predicts a future during which a departed worker will be “resurrected” from their work information, permitting firms to question that particular person’s institutional data. The thought sounds controversial and “somewhat bit scary,” he mentioned, but it surely may very well be invaluable for answering questions that solely the previous worker would have recognized.
Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you hear.
