The AI agent space has a setup problem. The tools that let you automate tasks across apps - workflow builders, agent platforms, custom GPT configurations - are powerful, but they take real time to learn. Poke, a startup covered by TechCrunch on April 8, is trying to skip all of that by making text messages the interface.
The pitch is simple: send a text, the agent handles the task. No new app to install, no account setup on an automation platform, no programming required. You describe what you want done in plain language, the same way you'd text a friend.
Who This Is Actually For
Poke targets a specific person: someone who's heard about AI automation, maybe even wants it, but found tools like Zapier, Make, or n8n too complicated to build something useful with. The activation energy for those platforms is real. Before you get your first working automation, you need to understand triggers, actions, webhooks, and API credentials. That's manageable for developers. It's a genuine wall for a freelancer who just wants to stop copying data between systems by hand.
The text-message interface has one practical advantage beyond simplicity: no context-switching. You don't open a browser, locate a platform, log in, navigate a dashboard, and then build. You describe the task where you already are.
The honest tradeoffs are real. Text inputs are inherently ambiguous, and agents that operate across your apps need access permissions that don't actually disappear just because the setup feels easy. "No complex setup" often means the complexity is deferred - you'll hit it the first time a task fails and there's no dashboard to debug. The failure mode for simplified agent interfaces tends to be opacity: things break in ways you can't see or fix without help.
The broader bet Poke is making - that the right interface for AI agents is the most familiar one, not the most capable - is worth watching. Most people who would genuinely benefit from AI automation aren't going to become power users of dedicated platforms. If a text-message interface can reliably handle a useful enough range of tasks, that's a real product for a real market.