Cognition Buys Poke: Why AI Personality Matters for Small Business

On July 24, 2026, TechCrunch reported that Cognition—the company behind the coding agent Devin—acquired The Interaction Company of California, maker of the messaging AI assistant Poke, in a deal valued in the “low nine figures.” Co-founder Marvin von Hagen confirmed the price range. The story is not only about a coding startup buying a consumer chat product. It is about a bet that how an agent talks—and where people already message—can be as strategically important as raw model skill.

For small business owners, the useful lens is customer support, internal ops, and brand voice: when does personality help, when does it hurt, and what does Apple’s Messages for Business path imply?

What happened

According to TechCrunch’s reporting and Cognition’s public comments:

• Cognition acquired The Interaction Company of California in a low-nine-figure deal (von Hagen confirmed the pricing characterization).
• Poke is an AI assistant people text through iMessage, SMS, Telegram, and in some markets WhatsApp. It launched in March 2026 and is known for a familiar, friend-like tone—including slang and humor—rather than a sterile “tool” voice.
• Cognition says users exchanged more than 100 million messages in three months, with hundreds of thousands of users. Von Hagen noted common productivity uses such as email, reminders, and to-dos—and that the product had been expensive to run, making profitability difficult.
• In June 2026, Poke became the first AI agent approved to run on Apple Messages for Business, Apple’s standardized channel for business–customer messaging inside Messages.
• Cognition co-founders Scott Wu and Walden Yan had been angel investors in the company. Wu’s post framed Poke as proactive, personal, and fun—qualities Cognition wants for Devin.
• Near term: Poke stays as-is through the end of 2026 and plans to remain on Apple’s platform. In 2027, the teams expect experiments with Cognition’s software engineering model SWE-1.7, possible orchestration of multiple Devin sessions via Poke, and longer-term product combination is not ruled out.

The strategic pitch is mutual: Devin gains interaction design and personality; Poke gains models and infrastructure for speed and reliability.

Why it matters

Small businesses already live in messaging. Customers text for appointments, order issues, quotes, and “are you open?” Internal teams ping each other on SMS and chat when email is too slow. An AI that works inside those channels reduces the friction of “please download our app” or “open this portal.”

Personality is the second half of the story. Capability without trust feels like a form. Trust without capability feels like a toy. Cognition’s purchase underscores a market belief that agents people enjoy using get more messages—and therefore more chances to complete work. For Devin, “colleague with a personality” is meant to reduce the coldness of an autonomous coding agent. For SMBs, the parallel is a support or ops agent that sounds on-brand and human enough that customers do not bounce to a phone call.

There is also a caution. Casual tone is not a free upgrade. Overly slangy agents can undermine professional services, healthcare-adjacent businesses, legal/finance touchpoints, or any brand that sells reliability. Personality is a product decision with brand-risk consequences, not a default setting.

Messages for Business is the third signal. Being first as an AI agent on Apple’s business messaging surface matters less as a trophy and more as a distribution pattern: regulated, identity-aware business chat may become a serious channel for AI-assisted support. SMBs that already use Apple Business Chat (or plan to) should watch how agent policies, handoff to humans, and approval standards evolve.

How businesses can benefit

Evaluate personality and capability separately. In demos, ask two scorecards: (1) Did it complete the task correctly? (2) Did the tone match our brand and the customer’s situation? A funny wrong answer is still wrong.

Map messaging-native agents to high-volume, low-risk threads. Appointment changes, store hours, shipping status, FAQ deflection, and “where is my invoice?” are better first pilots than refunds with policy exceptions or medical advice.

Use Messages for Business (or equivalent) as a channel strategy, not a novelty. If a large share of customers are on iPhone, business messaging can feel more trustworthy than a random web widget—if you have clear identity, logging, and human escalation.

Write a voice guide before you turn the agent loose. Define: formality level, humor allowed/forbidden, words never to use, when to apologize, when to escalate, and how to handle angry customers. Treat it like employee onboarding.

Watch the Cognition–Poke integration for product lessons. Even if you never use Devin or Poke, the industry is converging on agents that remember context across sessions and orchestrate multiple workstreams. Ask your vendors about persistence and multi-task coordination—not just single-reply chatbots.

Mind unit economics. Poke’s team itself flagged run-cost pressure. Messaging agents that chat a lot can burn tokens. Prefer vendors who optimize for resolution, not conversation length.

Practical examples

Boutique fitness studio. Members already text about class swaps. A messaging agent with a warm but clear voice can confirm cancellations against the schedule and escalate billing disputes to a human. Personality helps retention; correctness protects revenue.

Home services company. A plumber’s office uses SMS for dispatch. An agent can collect job details and photos, then hand a structured brief to the dispatcher. Keep the tone friendly; keep estimates and warranties human-approved.

B2B SaaS support. Cognition’s thesis for Devin—make the agent feel like a coworker—maps to internal IT or CS tooling. A slightly personable agent that remembers open tickets across days can reduce “start over” friction. Still require audit logs for customer-facing commitments.

Brand-risk stress test. Role-play: angry customer, grieving customer, legal threat, and sarcastic teenager. If the agent jokes in the wrong place, tighten the voice policy before launch.

Apple Messages for Business checklist. Confirm business verification, after-hours behavior, human takeover UX, transcript retention, and whether AI replies are disclosed. Do not assume consumer-chat norms transfer to business chat.

What to watch

• Deal terms beyond “low nine figures” were not fully public; treat financial details as reported ranges.
• Poke’s continuity through 2026 is a near-term comfort for users; 2027 product experiments may change the experience.
• Personality without strong tool use and memory becomes entertainment.
• Overly casual agents can create compliance and trust issues in regulated or high-stakes industries.

Conclusion

Cognition’s acquisition of Poke is a clear industry signal: interaction design and personality are becoming competitive advantages alongside model quality. Small businesses should not chase slang for its own sake. They should ask whether agents live in the channels customers already use, whether tone matches brand risk, and whether capability is strong enough that personality is a bonus—not a distraction. Start with narrow messaging use cases, write a voice policy, measure resolution and escalation quality, and treat Messages for Business (and similar channels) as infrastructure worth planning for—not a gimmick.

Sources

• Why Cognition bought Poke: AI personality is becoming a competitive advantage — TechCrunch