A great AI agent can be a little like a brilliant houseguest who eats cereal with a soup ladle. The intelligence may be there. The usefulness may be real. But if every interaction feels awkward, people remember the awkwardness first.
That is the frame I keep coming back to with Cognition’s purchase of Interaction, the company behind Poke. Cognition, best known as the maker of Devin, acquired Poke on July 23, 2026, in a deal valued in the low nine figures. Poke is a personal AI agent that lives in your texts, and Cognition bought it to improve Devin’s interaction style, with a focus on conversational AI personality.
For readers of agent101.net who do not spend their day tracking AI company chess moves, this may sound like a strange pairing at first. Devin is associated with coding agents. Poke is associated with texting. One sounds like engineering work; the other sounds like daily conversation. But that gap is exactly why the acquisition matters.
AI agents are not just tools anymore
For years, software competition was often easy to describe: one product had more features, another had a cleaner interface, another had a better price. AI agents complicate that because they are not just menus and buttons. They talk back. They ask questions. They make suggestions. They can feel helpful, impatient, careful, vague, confident, or oddly cold.
That feeling is not decoration. It shapes whether people trust the agent enough to keep using it.
In plain English, an AI agent is software that can take action toward a goal, often through conversation. A coding agent like Devin is built around software development tasks. A texting agent like Poke sits in a much more familiar place: the message thread. That matters because texting is one of the most personal interaction formats people use every day.
By buying Poke, Cognition is signaling that the way an agent communicates may be as important as what it can technically do. Not instead of capability, but alongside it.
Why personality becomes product
When people hear “AI personality,” they may picture jokes, emojis, or a chatbot trying too hard to sound human. That is the shallow version. The more meaningful version is interaction style: how the agent explains itself, how it responds when the user is confused, how it asks for clarification, and how it keeps a conversation moving without becoming annoying.
Those traits matter even more for agents than for ordinary apps. A spreadsheet app does not need to reassure you. A coding agent might. A calendar app does not need to explain why it is asking a follow-up question. An AI agent often does.
This is where Poke’s texting identity becomes relevant. A personal AI agent that lives in texts has to operate in a conversational setting where tone is the interface. There is not much room to hide behind dashboards or panels. The agent has to communicate naturally enough that the user wants to keep the thread going.
Cognition’s decision to bring that experience into Devin suggests a larger shift in the AI agent market: technical skill alone may not be enough. If two agents can complete similar tasks, the one that feels clearer, calmer, and easier to work with may win the user’s loyalty.
Why a coding agent company would care about texting
Coding is often described as a technical activity, but software development is also full of conversation. Developers explain goals, clarify requirements, review progress, change direction, and ask why something broke. If an AI agent participates in that work, its conversational behavior becomes part of the workflow.
A coding agent with a poor interaction style can create friction even when it is capable. It may ask questions in a confusing way. It may provide updates that are too vague. It may sound too certain when uncertainty would be more useful. For non-technical people managing technical work, this becomes even more important. They need the agent to translate complexity into understandable next steps.
That is why this acquisition is not merely about adding a friendlier voice. It is about making the agent easier to coordinate with.
At agent101.net, I often describe AI agents as digital coworkers rather than magic boxes. Coworkers need more than raw skill. They need communication habits. They need timing. They need to know when to speak plainly. The same applies to agents, especially as they move from demos into daily work.
Low nine figures says this is not a side quest
The reported value of the acquisition, in the low nine figures, is important because it shows Cognition placed serious value on interaction design and conversational behavior. This was not framed as a small experiment. It was a major bet on the idea that AI personality can become a competitive advantage.
That phrase can sound soft, but it is actually very practical. A user who feels comfortable with an agent is more likely to give it context. More context can make the working relationship more useful. A user who dislikes an agent’s tone may stop using it before its technical strengths ever matter.
For businesses building agents, this creates a new kind of pressure. They cannot focus only on task completion. They also have to ask:
- Does the agent explain itself clearly?
- Does it ask for help at the right moment?
- Does it sound trustworthy without sounding fake?
- Does it adapt to a user’s communication style?
- Does it reduce confusion rather than add to it?
What this means for everyday users
For non-technical users, the Cognition-Poke deal is a useful clue about where AI agents are headed. The future of agents is not just smarter systems hidden behind chat boxes. It is agents that feel easier to talk to, easier to guide, and easier to correct.
That does not mean every AI agent should pretend to be your best friend. In many cases, that would be exhausting. The better goal is fit. A coding agent should communicate in a way that matches the work: clear, responsive, and aware of context. A personal texting agent may need a different rhythm. The important point is that personality is becoming design, not garnish.
Cognition bought Poke because Devin’s interaction style matters. That single move captures a bigger truth about AI agents: as they become more capable, the human side of using them becomes more important, not less.
The smartest agent in the room still has to know how to talk to the people in it.
đź•’ Published: