AI Agents in Your Texts: What Marketers Should Do Now
A growing group of AI agents now lives in iMessage, SMS and WhatsApp, booking, shopping and reminding on users' behalf. For brands, that means a new layer between customers and your website.
Rajesh Kumar · 22 Nov 2026 · 3 min read

- TechCrunch AI lists a growing set of AI agents you can text via SMS, iMessage, WhatsApp and Telegram, covering personal, family, work and travel use.
- Instinct, the best-funded, raised $1 billion at a $10 billion valuation in September 2026 and is adding agent email addresses and phone calls while still in private beta.
- Agents that book, buy and cancel on a user's behalf could sit between customers and your site, shrinking the shortlist of businesses people see.
- Clear facts, low-friction booking and checkout, and responsive messaging channels are the practical priorities.
- Most tools are still in beta, so watch adoption before building anything agent-specific.
What happened
A TechCrunch AI roundup lists a growing number of AI agents that people can text like a contact. There is no new app to download. You message the agent, it remembers context, connects to your email and calendar, and completes tasks such as booking a restaurant, researching a trip, sending an email or shopping online.
The best-funded name is Instinct. It raised $350 million at a $2.5 billion valuation, then $1 billion at a $10 billion valuation in September 2026. It is still in private beta. It has also started giving each user's assistant its own email address so it can sign up for services and follow up with businesses, and it is rolling out phone calls.
The field is broad:
- General assistants: Ask Proxy, Folk, Martin, Maverick and Iris
- Work-focused: Comma, which is open source, plus a business version of Ask Proxy
- Families: Fambot and Ohai
- Travel: Miso, which pairs AI with a human travel team
- Personalisation: Haven, which pitches a "digital twin"
Pricing runs from free betas to $1 a day for Ask Proxy, $8.33 a month for Folk Pro and $21 a month for Martin.
Why it matters
Most of these are early, and many are still in beta. But the direction is clear. The customer's first move may no longer be a search box or a social feed. It may be a message to an assistant that then talks to your business for them.
That changes three things for marketers and ecommerce owners.
Discovery gets compressed. If someone texts "find me a restaurant for Friday" or "order groceries," the agent picks from a short list. The source doesn't say how these agents choose. But fewer options shown means a higher cost of not being one of them.
Your customer might not be the one clicking. Agents are described as making reservations, buying things, cancelling subscriptions and calling businesses. Checkout flows, booking forms and phone lines may increasingly be used by software. Friction that a human tolerates, an agent may simply route around.
Email and inbox behaviour may shift. Instinct's dedicated agent email addresses mean sign-ups and follow-ups can come from addresses that aren't a person's real inbox. That could make email lists and attribution messier. The source also notes that this level of autonomy has raised privacy and security concerns.
What to do now
None of this requires a rebuild. It requires tidying up the basics that any agent, or any human, would depend on.
- Make your key facts easy to read. Opening hours, prices, locations, availability, delivery and return policies should be clear, current and on crawlable pages. An agent researching on a user's behalf needs unambiguous answers.
- Fix your booking and checkout friction. Test your flows. Can someone book without creating an account, solving a puzzle or calling during office hours? Simple flows serve humans and automated helpers alike.
- Answer messages and calls properly. Agents like Martin and Instinct can send messages or place calls. Make sure your business can respond clearly, and that your WhatsApp or SMS channels, if you use them, are staffed or well automated.
- Watch your email programme. Don't assume every sign-up is a person. Look at engagement, not just list size, and keep confirmation and unsubscribe flows clean.
- Pay attention to the family and travel niches. Fambot pulls school, sports and calendar information into a nightly summary. Miso handles flights over iMessage. If you serve those audiences, think about how your event, schedule or offer data reaches those tools: invites, calendar files and clear confirmations.
- Don't chase every tool. Most of these products are in beta. Track which ones gain real users before building anything specific for them.
The bigger picture
We've spent two decades optimising for search results and social feeds. The messaging thread is a different surface. It's personal, it's permission-based (Maverick, for example, asks before acting), and it's where an agent can act for the customer rather than only advise them.
The winners will likely be businesses that are easy to verify, easy to transact with and easy to contact. That is not new advice. It just has a sharper edge now.
Raj's takeAfter 27 years in this game, I've seen every new interface promise to change everything, and the real shift is always the same: someone new stands between you and the customer. Here that someone is an agent in the text thread. I wouldn't panic or rebuild, but I would make sure my prices, availability and checkout are clean enough that software can use them without help.
Want this applied to your business?
Related

AI Is Making Marketing Centralization Harder to Get Right
A MarTech piece frames the classic tension: shared teams bring scale and control, local teams bring speed and flexibility. AI makes deciding where each capability belongs more urgent for every business.

AI Agents and CX: Meeting Customer Expectations Wisely
MarTech argues AI agents could help brands meet today's demanding customer expectations, but only if they're used wisely. Here is what that means for ecommerce and marketing teams in practice.

AI Writes Faster, But Campaigns Still Ship Slowly
MarTech argues that faster AI drafts don't automatically produce faster campaigns. The real ROI comes from fixing approval bottlenecks, giving teams decision rights and measuring the gains.
