Your texts are getting crowded.
Not with people. With agents. The current pitch from a growing cluster of AI startups is that the chat app you already open 80 times a day is the right home for an assistant, and that downloading yet another app with a glowing orb logo is the thing holding AI back. It’s a decent argument. It’s also a convenient one, because building inside SMS and iMessage means you inherit a distribution channel nobody has to be convinced to install.
So let’s talk about what’s actually been shipped, what it costs, and whether the premise holds up.
Poke got the one thing money can’t buy
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 approval is the interesting part, more than any feature list. Apple’s business messaging rails are a walled garden with a bouncer, and being the first agent through the door is a structural advantage that competitors can’t engineer their way around. It’s not a better model or a smarter memory system. It’s a permission slip. In a space where everyone’s underlying capabilities are converging fast, the permission slip may matter more than the tech.
My honest read: being first through a platform gate tends to buy you a window, not a moat. Apple approved one agent; nothing in that decision suggests it plans to approve exactly one forever. Poke’s job over the next stretch is to turn a timing advantage into a habit advantage, which is a very different and much harder problem.
Ollie is selling trust, and pricing like it
Ollie takes a narrower swing: family-focused, and one of the first mainstream assistants in that category to achieve SOC 2 compliance. SOC 2 is a widely used security standard, the kind of certification that normally shows up in enterprise procurement rather than consumer app marketing.
Putting it front and center for a family product is a smart read of the room. If you’re going to let an assistant sit in the thread where your household coordinates pickups, appointments, and whatever else, “we passed a real audit” is a more persuasive feature than another model upgrade. It’s also the sort of claim that’s verifiable, which makes it rarer than it should be in this market.
Then there’s the pricing, and this is where I’d push back.
The per-message math is rough
Ollie has a free tier. Paid plans start at $25 per month for 150 messages. A $100 per month plan includes up to 1,000 messages.
- $25 for 150 messages works out to roughly 17 cents per message.
- $100 for 1,000 messages works out to roughly 10 cents per message.
- Spending four times as much gets you about six and a half times the messages.
Volume discounts are normal. What’s not normal, for a consumer product, is metering a text message at all. The entire emotional appeal of a texting agent is that texting feels free and thoughtless. You fire off a half-formed question, correct yourself, add context, ask again. That’s how people actually use SMS.
Message caps punish exactly that behavior. At 150 messages a month you’ve got five a day, and a single back-and-forth about scheduling can eat three of them. You end up composing carefully to avoid wasting an allotment, which reintroduces the friction the product was supposed to remove. I’d rather see time-based or task-based pricing than a meter that makes me budget my own sentences.
What I’d actually watch for
The category is young enough that most of the real questions are still open. A few I’d want answered before paying:
- Does a failed request cost you a message? If the agent misunderstands and you clarify, who eats that?
- What happens at the cap? Hard stop, degraded service, or overage billing you notice later?
- How much context does it keep? A texting agent with no memory of last Tuesday is a search box with extra steps.
- Who else can see the thread? SOC 2 is a floor, not a full answer on data handling.
The strategic picture is clearer than the product picture. Poke is betting that platform access wins. Ollie is betting that trust and a specific use case win. Those are both reasonable bets, and they’re not really competing with each other yet.
What neither has proven is that people want a monthly subscription living in their messages app. Texting is the most casual interface we have. Attaching a counter to it is a strange way to prove otherwise, and I suspect the pricing models we see today won’t be the ones that survive contact with real usage. I’d try the free tiers. I’d think hard before committing $100 a month to a thousand text messages.
🕒 Published: