An SMS AI agent for business is an AI agent that answers and acts on customer text messages (SMS, WhatsApp or iMessage) the way a trained staff member would: it reads the message, checks your systems, replies in seconds, and books, quotes, reschedules or hands off to a person when it should. It is worth building when you get more than a few dozen customer texts a week and some of them wait hours for a reply. A focused first version, covering one job such as bookings or order status, typically takes four to six weeks and runs on the business number your customers already use.
This post is part of our guide to rolling out agents, which starts at wizeb.com/blog/how-to-implement-ai-agents-in-business.
The News: Consumers Are Learning to Text Their Agents
TechCrunch has been keeping a running guide to consumer AI agents you can text, and this month it passed 20 entries: assistants that live in SMS, iMessage, WhatsApp and Telegram instead of in an app. People text them to book travel, cancel subscriptions, plan family schedules and chase down answers. Apple has also started approving AI agents for Messages for Business, so an agent can now reply inside the same blue-bubble thread a customer uses for friends.
The business point is simple. Once your customers get used to texting an agent that answers in ten seconds and actually does the task, a business that replies to texts the next morning starts to feel slow. Texting is already where many customers prefer to reach you: no hold music, no app download, and the conversation sits in their pocket. The agents in TechCrunch's list are training their expectations.
The Problem: Texts Arrive Faster Than Your Team Can Answer
Most small and mid-sized businesses already get customer texts, often on a shared WhatsApp number or a texting add-on to their phone system. The pattern we see is the same almost everywhere:
- Texts pile up outside office hours, and the first reply comes the next working day.
- Most messages are the same handful of questions: are you open, how much, can I move my booking, where is my order.
- Staff answer them in between other work, so replies are inconsistent and sometimes wrong.
- Nobody tracks texts as leads, so an enquiry that goes unanswered simply disappears.
Speed matters more on text than on email. A customer who texts three businesses for a quote usually books the first one that answers properly. A basic auto-reply ("Thanks, we'll get back to you") doesn't help, because it doesn't answer the question.
What an SMS AI Agent for Business Actually Does
The difference between a texting agent and an old SMS chatbot is that the agent can act. A chatbot follows a script and matches keywords. An agent understands the message in plain language, looks things up in your systems and completes the task within limits you set. If you want the full distinction, see wizeb.com/blog/ai-agent-vs-chatbot-vs-automation.
A typical texting agent handles four kinds of work:
- Answers: opening hours, prices, service areas, policies, stock levels, pulled live from your systems rather than from a static FAQ.
- Bookings and changes: offers real open slots from your calendar or booking system, confirms, reschedules and sends reminders.
- Status updates: order, delivery, repair or application status, looked up from your order system or CRM.
- Lead capture and qualification: asks the two or three questions your sales team needs, logs the lead in the CRM and books a call with the right person.
Everything else, such as complaints, refunds above a limit, anything sensitive or anything the agent isn't sure about, goes to a person with the full conversation attached. The customer stays in the same thread and never has to repeat themselves.
How to Set One Up: Five Steps
- 1Pick one job. Pull the last 200 customer texts and sort them. Usually three or four question types make up 60 to 80% of the volume. Start with the one that is most common and easiest to verify, such as bookings or order status. Our guide at wizeb.com/blog/first-ai-agent-for-business covers how to choose.
- 2Choose the channel your customers already use. In the US and UK that is mostly SMS; in the Gulf, India, Europe and Latin America it is usually WhatsApp. Don't move customers to a new number; connect the agent to the one they already text.
- 3Connect the systems the job needs. A booking agent needs read and write access to your calendar or booking tool; a status agent needs read access to orders. Give it only what that job requires.
- 4Write the limits. Decide what the agent may do alone, what needs a person's approval, and what it must never do. For example: book and reschedule freely, offer discounts never, escalate any complaint immediately.
- 5Run it in shadow mode for two weeks. The agent drafts replies and a staff member approves them with one tap. Once 95% of drafts go out unchanged, let it send on its own for that job.
The Rules: Consent, Templates and Opt-Outs
Texting has stricter rules than email, and getting them wrong is expensive. Three things to get right before launch:
- Consent. Replying to a customer who texted you first is generally fine. Texting people who never opted in is not. In the US, business texting falls under the TCPA, and business SMS numbers have to be registered for A2P 10DLC with the carriers.
- WhatsApp's 24-hour window. On the WhatsApp Business Platform, you can reply freely for 24 hours after the customer's last message. Outside that window, you can only send pre-approved template messages, such as reminders and confirmations.
- Opt-outs and disclosure. Honour STOP and similar keywords instantly, and tell customers they are texting an AI assistant, with an easy way to reach a person.
Rule of thumb
Let the agent reply instantly to anyone who texts you. Treat any message the agent starts itself, such as reminders, follow-ups or offers, as marketing that needs consent and, on WhatsApp, an approved template.
What It Costs
Running costs are usually small next to the staff time saved. Messaging fees depend on channel and country, from fractions of a cent to a few cents per SMS segment, with WhatsApp charging per template message rather than per reply. AI model costs for a typical customer conversation are usually a few cents. For a business handling 2,000 conversations a month, total running costs often land in the low hundreds of dollars.
The bigger cost is the build: connecting your systems, writing the rules and testing on real conversations. A single-job texting agent is typically a four-to-six-week project. Each added job afterwards is cheaper, because the channel, logging and approval flow are already in place.
Case Study: A Home Services Firm That Stopped Losing Evening Leads
Consider a 25-person plumbing and HVAC company that received about 1,400 texts a month, roughly 40% of them after 6 pm. Office staff replied the next morning, and the owner estimated that one in five evening enquiries had already booked a competitor by then.
The company connected a texting agent to its existing business number, its scheduling system and its price list. The agent answered service-area and pricing questions, asked three qualifying questions (type of problem, property type, urgency), offered real appointment slots and booked them. Anything involving a gas smell, flooding or a complaint went straight to the on-call technician by phone. After a two-week shadow period, the agent handled 68% of conversations end to end. Median reply time fell from 9 hours to under a minute, and evening bookings rose 31% in the first two months. This is an illustrative example, but the pattern is typical: the gain came from answering at 9 pm, not from clever wording.
If your customers mostly call rather than text, the same approach works on the phone. Our guide at wizeb.com/blog/ai-agent-examples-by-business-function shows where texting, voice and back-office agents each fit.
How WizeB Builds Texting Agents
WizeB builds AI agents that work on the channels your customers already use: SMS, WhatsApp and voice. We start from your real message history, pick the job with the clearest payoff, connect the agent to your booking, order and CRM systems with the limits you agree, and run it in shadow mode before it sends anything alone. You get a log of every conversation and every action, and you own everything we build. See how we work at wizeb.com/services/ai-agents.
Turn your texts into bookings
Book a free scoping call at wizeb.com/services/ai-agents. Bring an export of last month's customer texts and we will show you which share an agent could handle, and what it would take to launch on your current number.