Your customers' AI agents are starting to text, call and email your business. Is your front office ready?

The next inquiry in your inbox may not be written by a person. It may be written by their agent.

What happened

TechCrunch published a roundup of AI agents that live in text messages. Instead of another app, people text these assistants over iMessage, RCS, SMS, WhatsApp or Telegram, and the agents connect to calendars and email to book appointments, make reservations, shop and follow up. A few details stand out:

• Wajo's agent Fo has its own email address, phone number and payment card, so it can deal with businesses without the user's credentials, and it can call businesses and make bookings.
• Instinct gives its assistant its own email address to sign up for services and contact businesses.
• Poke was described as the first AI agent on the Apple Messages for Business platform.

Separately, The Verge reported on how this is landing with frontline staff. A restaurant manager in Providence says her restaurant has been getting calls and emails from agents trying to make reservations for users, and she suspects a burst of last-minute event requests came from an agent, not the client. Servers and store staff describe customers asking for products, wines or features that do not exist because a chatbot said so.

My take

For small businesses this is a front office problem, and front office is where I spend most of my automation work.

Three things I would set up now:

1. Make your facts machine-readable. Menus, prices, stock, policies and opening hours should live in one clean source that your site, chatbot and booking tools all read from. If an assistant is going to summarize your business, give it the real answer.
2. Confirm before you commit. When a booking or a large change comes in by text or email, send an automatic confirmation to the customer's known contact and wait for a reply before you reserve staff or stock.
3. Tag the source in your CRM. Add a field for how a request arrived and whether it looks agent-made. Within a month you will know how much of your inbound is already machine-written.

None of this needs a new platform. It is intake, confirmation and clean data, built properly.