What if the most useful AI tool you ever adopt is one you never download?
Most of us have been trained to think of AI as a destination. You open a tab, you open an app, you type into a box designed for talking to a machine. But a quieter category has been forming around a different idea: the AI goes where you already are. And for most people, where you already are is the Messages app.
I find this genuinely interesting, and not because texting an AI is technically impressive. It isn’t. SMS is ancient. The interesting part is what happens to adoption when you remove the download step entirely.
Poke and the Apple approval that actually matters
Poke launched in March 2026 as a general-purpose, text-based assistant. Three months later, in June 2026, it became the first AI agent approved on Apple’s Messages for Business platform.
That second sentence sounds like plumbing. It’s not. Messages for Business is the channel businesses use to talk to you inside iMessage, the same blue-bubble thread where your family argues about dinner plans. Apple gatekeeps it carefully. Being the first AI agent through that door means Apple drew a line and decided an AI assistant could stand in a space previously reserved for airlines, banks, and retailers.
For non-technical readers, the practical translation is this: you don’t install Poke, you text it. No onboarding flow, no account creation screen, no app icon competing for space on your home screen. The friction that kills most new software just isn’t there.
Why the channel changes the product
An AI that lives in your text messages behaves differently than one that lives in a tab, mostly because of how you treat your messages. A browser tab is a place you go to do a task. A text thread is something you glance at in line at the grocery store. Short, scattered, low-ceremony.
That shapes what these assistants are good for. Quick questions. Reminders. Fire-and-forget requests you’d never bother opening an app for. The form factor rewards small, frequent interactions rather than long sessions.
Ollie takes the family angle
The other approach worth watching is specialization. Ollie is one of the first mainstream, family-focused AI assistants to achieve SOC 2 compliance, a widely used security standard.
SOC 2 is unglamorous and easy to skip past, so let me explain why it belongs in an article about text messages. SOC 2 is an audit of how a company handles data: who can access it, how it’s stored, what controls exist. Companies pursue it because enterprise buyers demand it before signing contracts.
A family assistant pursuing it is a different signal. If an AI is sitting in the thread where you coordinate school pickups, share your address, and mention which nights nobody is home, that’s an unusually sensitive stream of information. An external audit of data handling is the kind of thing that should be table stakes for that category. Ollie getting there early suggests the company understood the stakes of the channel it chose.
The pricing tells you something too
Ollie offers a free tier, plus paid plans starting at $25 per month for 150 messages. A $100-per-month plan includes up to 1,000 messages.
Notice the unit. Not hours, not tokens, not seats. Messages. That’s pricing built around a texting habit rather than a software subscription, and it quietly reveals an assumption about usage: that you’ll send a modest number of meaningful requests rather than grinding through hundreds of prompts a day.
It’s also, let’s be honest, not cheap per message. $25 for 150 messages works out to roughly 17 cents a text. Whether that feels reasonable depends entirely on what each message accomplishes. “Remind me Thursday” at 17 cents is a bad deal. “Sort out the three conflicting schedules in this thread” might be a great one.
What I’d actually watch for
If you’re considering one of these, a few questions are more useful than feature lists:
- What does it do with the rest of your thread? An assistant in your messages has proximity to a lot of context you didn’t deliberately hand over.
- Has the company said anything verifiable about security, or just used reassuring language? SOC 2 is a specific claim. “Your privacy matters to us” is not.
- Does message-based pricing match how you’d use it, or will you burn your allotment on trivia?
- What happens when it’s wrong? A bad answer in a tab is an annoyance. A bad answer in a family thread can become a missed appointment.
The shift here is modest in technology and large in behavior. Putting an AI in the place you already check two hundred times a day removes nearly every barrier to using it. That’s a real advantage, and it’s also the reason to be thoughtful about which one you let in. Your text messages are not a neutral space. They’re arguably the most personal software you own.
🕒 Published: