Could your next AI assistant be a contact in your phone? A TechCrunch survey by Lauren Forristal, published October 3, 2026, maps the growing set of AI agents available through text messages and familiar messaging apps. These tools aim to turn a request in a chat thread into an action: a calendar event, a reminder, a booking, a travel plan, an email, or a follow-up.
Why AI agents in text messages matter
Texting is already part of daily life. An assistant that lives there removes the step of opening a dedicated app and learning a new interface. When users connect calendars, inboxes, and other services, the assistant may also be able to carry out multi-step tasks instead of only answering questions. That makes the experience potentially useful for people coordinating work, travel, or family responsibilities.
Forristal’s roundup includes general-purpose assistants such as Caddy, Folk, Instinct, Iris, Martin, Pally, Poke, and Rene. It also covers products aimed at particular needs: Fambot, Ohai, Ollie, and Orbits for households; Miso for travel; Stanley for creators; Town for work; and Wajo’s Fo for everyday errands. They differ in messaging platforms, integrations, availability, and pricing. Several remain in beta.
The tradeoff: convenience for access
The more an agent can do, the more it may need permission to read messages, manage schedules, send email, contact businesses, or make purchases. Those permissions deserve careful review. Ask what data the service can see, which actions require confirmation, how mistakes are corrected, and whether a human can intervene. Company descriptions of capabilities are not the same as independent performance tests.
This echoes the trust questions we explored in our comparison of AI smart-home assistants: convenience depends on the permissions people are comfortable granting.
One example of the trend is Instinct, which TechCrunch reports has given assistants dedicated email addresses and begun rolling out calling support. Wajo says its agent can use its own email address, phone number, and payment card. These approaches could make agents more autonomous, but they also sharpen questions about accountability when an assistant communicates or spends on someone’s behalf.
What to watch next
The useful test is whether a text-first agent reliably completes real tasks without creating new risks or hidden work. Before giving one broad access to personal accounts, check its current availability, subscription terms, privacy policy, and approval controls directly with the provider. Our podcast and Short discuss the wider market shift and its limits; neither is a hands-on ranking of every product in the survey.
Read the original TechCrunch article and listen to The Daily AI Chat using the embedded episode above.
Watch the Short on YouTube or listen to the full episode on Spotify.
