In a GeekWire podcast conversation, Read AI CEO David Shim provided a nuanced take on the current AI craze. Unlike dot-com-era hype, Shim argues, today’s AI boom is more grounded: “real business models and customers willing to pay” are driving growth, rather than free-for-all speculation.
Still, he doesn’t deny that bubbles exist in some corners. Shim called out parts of the market where companies have huge valuations without meaningful products or revenue — what he described as “100% bubbly.” That said, he thinks any correction will be gradual, not a sudden crash: “I think it’s a bubble, but I don’t think it’s going to burst anytime soon … more of a slow release.”
Shim also shared his vision for how AI agents will really add value. He expects the most powerful AI helpers to be domain-specific and embedded deeply into workflows — what he calls “invisible infrastructure” — rather than broad assistants that try to do everything. He gave the example of Read AI’s own assistant, “Ada,” which learns how users communicate, but even adds deliberate delays in replies because “quick replies freak people out” by seeming too robotic.
Looking further ahead, Shim talked about some bold possibilities. He sees “multiplayer AI” — where agents share data across teams — as a way to unlock far more insight than when they only know one person’s context. Even more provocatively, he raised the idea of “digital twins”: preserving a departed employee’s institutional knowledge through the data they left behind so AI could continue to answer questions as if they were still around.
