The agent economy just got its SMS interface, and that changes who gets to build.
The Summary
- Poke launches AI agents accessible purely through text messaging, eliminating the need for apps, coding, or complex setup
- Distribution matters more than features when you're trying to put agents in every pocket
- The lowest-friction interface wins the normie market, and SMS beats app downloads every time
The Signal
Poke is building AI agents that run entirely through text messages, handling tasks and automations without requiring users to download apps or learn new interfaces. This matters because every other agent platform right now requires you to navigate their proprietary interface, connect accounts, build workflows, or at minimum download their app. Poke skips all of that.
The interface innovation here isn't technical brilliance. It's distribution strategy disguised as UX design. SMS is the most universal computing interface on the planet. No app store approval. No onboarding flow. No "create an account" friction. You text a number, the agent texts back, and you're running automations. That's the entire barrier to entry.
This is how agents leave the early adopter bubble. Not through better models or more capabilities, but through meeting people where they already are. Your mom won't download another app. She will text a number if it saves her from manually doing something repetitive. The agent economy doesn't scale through Slack integrations and API documentation. It scales through the path of least resistance.
The Implication
Watch for two things. First, how quickly other agent platforms clone this distribution model once they see it working. SMS as an agent interface is too obvious in hindsight not to become table stakes. Second, watch what tasks people actually delegate when the friction drops to zero. The automation wish list of someone who texts an agent is different from someone who builds workflows in a dashboard. That data tells you what normal humans actually want agents to do.
Sources: TechCrunch AI | TechCrunch AI