Institutions are splitting into two camps with AI: those encoding discipline directly into the agent layer, and those blindly buying capability. The most durable AI transformations are coming from the first camp. And they got there by slowing down.
Linear, a project management platform used by companies like OpenAI and Coinbase to manage their own AI agents, made a choice most organizations aren’t making. Instead of shipping AI features because they could, they stopped to ask whether each one was actually useful. Often the answer was no. The result is a platform where agents are first-class users: not bolted on, not a chatbot interface grafted onto existing workflows, but genuinely redesigned to work the way agents and humans do. Their CEO puts it plainly: when AI moves fast, decisions matter more, not less, because the cost of a wrong one compounds faster.
I’ve watched the opposite play out across large enterprises. The organizations struggling most aren’t moving too slowly. They moved fast in the wrong direction and are now managing the consequences: fragmented tooling, conflicting ownership, adoption fatigue. UiPath’s CMO recently put a number on it. 70 to 80 percent of agentic AI initiatives never make it out of pilot. The diagnosis isn’t capability. It’s coordination.
The discipline to not do something is harder to build than the capability to do anything. What’s on your AI roadmap right now that probably shouldn’t be?
References: Every, “If SaaS Is Dead, Linear Didn’t Get the Memo” (April 2026); The Rundown AI, “Exclusive: UiPath CMO Michael Atalla on AI at Work” (May 2026).
