Cognition Acquires Poke to Make AI Personality a Core Product Feature

What happened

Cognition, the AI company behind the autonomous coding agent Devin, has acquired Poke, a startup known for its distinctive conversational style and interaction design. The deal signals a deliberate strategic shift: Cognition intends to bring Poke's personality-driven interaction model directly into Devin, betting that the way an AI assistant communicates is now as commercially significant as the underlying model powering it.

The acquisition reflects a broader industry reckoning. As foundation models converge in raw capability, differentiation is migrating upstream — into tone, pacing, feedback loops and the felt quality of the exchange. Cognition's move suggests the company believes that developers choose and stick with AI tools not purely on benchmark performance, but on whether working with the agent feels productive, legible and even enjoyable.

Why it matters

For customer-experience and service-design practitioners, this acquisition is a signal worth taking seriously. The logic Cognition is acting on — that interaction personality drives adoption and retention — is the same logic that governs every human-facing service. Behavioral economics has long established that the process utility of an experience (how it feels to go through it) shapes satisfaction and loyalty independently of the outcome. Cognition is, in effect, applying that principle to AI product design: a coding agent that communicates with clarity, appropriate warmth and a consistent voice will be trusted and used more than one that is merely accurate.

For operators designing AI-assisted customer journeys — from chatbots to voice agents to in-app assistants — the implication is direct. Personality and conversational coherence are not cosmetic finishing touches; they are load-bearing elements of the service architecture. Organisations that treat interaction design as an afterthought, bolting a generic LLM onto a workflow and calling it done, are likely to see the same attrition patterns that plagued early IVR systems: technically functional, humanly alienating.

The Renascence take

Most commentary on this deal will focus on the competitive dynamics between AI coding tools. The more important story is what Cognition's willingness to acquire for personality reveals about where value is actually accruing in the AI stack — and what that means for anyone designing services around AI agents.

The race to the best model is, for most enterprise use cases, already over — the differences are marginal and narrowing fast. What Cognition understands, and what most operators have yet to internalise, is that trust in an AI system is built through thousands of micro-interactions: the phrasing of a clarifying question, the cadence of a status update, the tone of an error message. These are service-design decisions, not engineering ones. Customer-obsessed operators should stop asking "which model?" and start asking "what does our agent sound like when things go wrong?" — because that moment is where loyalty is won or lost.