I want to share a story that shows what AI agents look like when they're done right.

It comes from the AI company my treatment centers use. They shared it with me. It isn't my client, and the numbers below are what the vendor reported.

One treatment group they work with gets about 30,000 calls a month.

Think about that volume. Even a great intake team can't answer every call instantly, around the clock. Some calls get missed. Some wait. And in a business where a single call from paid ads can cost about $200, every missed call is expensive.

So the group put AI on the phones and a well-trained AI chat on their website.

Not one of those useless out-of-the-box bots. It handles real intake and answers real questions.

According to that AI company, the result was about 24 new intakes a month.

Each intake is worth about $10K in net revenue after therapist and treatment costs, so that's roughly $240K a month in estimated net revenue.

That one change pays for the AI phone and chat for a long, long time.

The difference between this and the chatbot that annoys you on most websites is training.

Most bots are installed and forgotten. They give canned answers, can't handle a real question and push people to "leave a message." That's not an agent. That's a speed bump.

A well-trained agent knows your services, your process, your FAQs and when to hand off to a human. If you take one thing from this story, take this: the tool matters less than the rules and knowledge you give it.

Some people call. Some people won't pick up a phone but will type a question at midnight. The group covered both. Most businesses cover neither well after hours.

Without an agent, missed inquiries often wait 1 to 12 hours, and I've seen 50-70% of those leads lost when nobody answers. AI on both channels closes that gap.

Treatment is a sensitive, high-trust decision. If a trained AI can handle real intake conversations there, it's worth testing in your business. Just stay within your industry's rules. Healthcare and other regulated industries have privacy and compliance requirements, so check yours before you deploy.

The math only clicked because the value of each intake was clear: about $10K net, per the vendor. Do you know your number? If you know what a new customer is worth after costs, you can tell in minutes whether an AI agent is worth paying for.

You probably don't get 30,000 calls a month. You don't need to. The same principle works at any size:

A lean agent stack typically runs about $100-$300/month. AI phone answering can cost more, so weigh it against what one new customer is worth to you.

I've seen 10-30% revenue lifts in client businesses from AI agents on the front end and back end. Results vary. But the pattern in this story is the same pattern I see everywhere: the money is in answering faster and following up better.

Want to set this up in your business? The Owner's Guide to AI Agents That Actually Grow Revenue walks through the speed-to-lead agent and four others, step by step. Get the guide →

J White, Growth Operator AI