The AI race just changed categories. Cognition, the company behind Devin and a $2 billion valuation according to TechCrunch, acquired Poke, a startup that gives AI agents real character. This isn’t a feature add. It’s a bet that bland AI is about to lose, and that personality is the next competitive moat worth more than any benchmark score.
What Just Happened
Devin launched in 2024 as one of the first true AI software engineers. It could take a GitHub issue and ship a working fix without hand-holding. Impressive then. Table stakes now. According to TechCrunch, the number of enterprise AI coding tools doubled in the twelve months following Devin’s debut. Every serious player can write solid code in 2026. The capability gap has closed.
Cognition’s answer was Poke. Poke builds what they call a character engine. This isn’t a system prompt. It’s a behavioral profile that keeps an AI agent’s personality consistent across long sessions, complex workflows, and context switches. The AI remembers how it communicates, what it prioritizes, and how it handles disagreement. It feels like a colleague, not a vending machine.
According to The Information, enterprise pilots using Poke’s character layer saw 40 percent higher 90-day retention compared to standard AI assistants. People didn’t just use these tools more. They advocated for them internally. That kind of word-of-mouth is worth more than any ad spend in enterprise sales.
Why This Is Bigger Than One Deal
The crypto world figured this out before enterprise tech did. ai16z, the AI agent project built on Solana, attracted over $100 million in community capital in under 90 days according to CoinDesk. The technology wasn’t the draw. The character was. Eliza had opinions. She had a voice. She had a point of view people wanted to follow. Investors poured money in because the agent felt worth trusting.
Cognition just made the same bet for enterprise software. And I think the enterprise version scales to ten times the value.
Here’s the mindset split most people will miss. Average operators are still shopping AI tools by benchmark scores. HumanEval performance, tokens per second, price per million tokens. That’s fine if you want the fastest horse in a sprint. It’s the wrong frame for a marathon.
Smart operators ask a different question: which AI will my team want to work with six months from now? Which one will my developers trust enough to hand off real work? That’s not a capability question. It’s a personality question.
According to Stanford HAI’s 2025 AI Index Report, user trust in AI tools correlates more with perceived behavioral consistency than with task accuracy. People forgive a mistake. They don’t forgive unpredictability. Poke sold Cognition on consistency, and Cognition is now selling that to enterprises at a premium.
If your business is building an AI stack right now, you need to track this spending separately from your general software budget. Wallester business cards let you assign virtual cards per vendor so you see exactly what each AI tool costs your team as your stack grows. When the personality race heats up and vendors start charging more for character, that visibility is the difference between a strategic decision and a surprise invoice.
What This Means for You
I’d make three moves right now.
First, audit every AI tool your team uses for personality fit, not just performance. Run each one for two weeks. Ask one question: does this feel like a colleague or a calculator? The tools that feel like colleagues will get used. The ones that feel like calculators will sit open in a tab nobody checks. Adoption rate is the only metric that matters at the end of a tool evaluation.
Second, if you’re building any product with AI inside it, start designing your AI’s character before you design its features. What does your AI care about? How does it handle pushback? What’s its communication style under pressure? These aren’t soft questions. They’re product decisions with hard revenue consequences. Cognition just paid to prove that.
Third, watch how enterprise software categories reshape around this insight over the next 18 months. Project management, customer support, sales tools, every category with an AI layer is heading into a personality competition. The tools that nail character will beat technically superior competitors the same way Slack beat HipChat and Notion beat Confluence. Better personality wins the daily active user count every time.
For founders and operators managing the humans who work alongside these AI tools, Gusto payroll keeps your actual team’s compensation simple and compliant as you shift more operational budget toward AI that handles the repetitive work. You can reinvest the savings without your HR setup falling apart.
The Bottom Line
Cognition didn’t buy a feature. They bought a philosophy: that AI agents who feel real will own the enterprise market. The commodity race is over. The personality race has started. Your competitors are about to figure this out. Build your AI stack around character now or spend twice as much catching up in 2027.
Frequently Asked Questions
Why did Cognition acquire Poke?
Cognition acquired Poke to add a character engine to its Devin AI product. As AI coding tools commoditize on pure capability, consistent personality has become the primary retention driver. According to The Information, pilots using Poke’s character layer showed 40 percent higher 90-day retention than standard AI tools.
What is an AI character engine?
An AI character engine is a behavioral profile system that keeps an AI agent’s personality stable across long sessions, complex tasks, and topic shifts. Unlike a system prompt that can drift, a character engine maintains consistent tone, priorities, and communication style. It makes AI agents feel predictable and trustworthy instead of generic.
Does AI personality actually affect business outcomes?
Yes. According to Stanford HAI’s 2025 AI Index Report, user trust correlates more strongly with behavioral consistency than task accuracy. Tools that feel consistent get used more, generate more internal advocacy, and hold market position longer than technically superior but impersonal alternatives.
How does the Cognition and Poke deal connect to crypto?
The crypto world proved the personality thesis first. ai16z attracted over $100 million in community investment in under 90 days according to CoinDesk, largely because the Eliza agent had a distinct, followable character. Cognition’s acquisition applies the same insight to enterprise software, where the addressable market is far larger.
How should I factor AI personality into tool selection?
Run every AI tool your team evaluates for at least two weeks and measure qualitative feedback alongside performance data. Ask your team directly whether the tool feels like a colleague or a calculator. The tools people want to use deliver more value than the tools with the best specs. Adoption is the real benchmark.
