Field Notes · Operations

My Customers Did Not Want to Talk to AI. Here Is What I Changed.

By Published 7 min read

Why did I put an AI agent in front of callers who did not want one?

A lot of owners have already decided their customers will hate talking to AI. Some leave the phones alone and keep missing calls. Others put an agent in place and quietly hope nobody notices it is not a person. I understand both instincts, because I built an agent for customers who felt exactly that way.

Garage Force of Greater Charlotte is a concrete coatings franchise I own and operate. Most of our work is residential garages, with some commercial jobs, and most of the people who call us are fifty five and older. They would rather talk to a person than a machine. That made it a hard place to test AI, and that was the point. Before I recommend this kind of tool to another owner, I want to prove it where my own money and reputation are on the line.

So I set up an AI voice agent named Mark. Mark answers only when I miss a call. It answers basic questions about floors, books a consultation or a discovery call, and never quotes prices. The full system, and the list of everything that went wrong first, is in the Garage Force case study. This post is about one part of it: what I learned from customers who did not want to talk to AI at all.

Why does my AI agent tell callers it is AI in the first sentence?

My callers skew older, and hiding the AI would have broken their trust. So the agent says what it is in its first sentence. It does not wait to be asked.

The reasoning is simple. If an AI is caught pretending to be a person, the customer is gone. If it is honest, it still has a chance to help. A caller who knows they are talking to an assistant can decide to give it their name and the space they want coated, or ask for me instead. Either way they are still a customer.

Honesty only works if there is a real person behind it. When a caller wants to speak with me, the agent texts my phone directly and I call back within 24 hours. Reminders nudge me at four hours and again the next morning, and they stop once I mark the callback done. Telling people they are talking to AI is easier to hear when you also tell them, by how the system works, that a person is accountable for what happens next.

What did older callers actually push back on?

Being honest about the AI was the first decision. The next ones came from watching how real callers responded, and each of them was a choice I had made while building for a demo instead of for the person on the phone.

The agent asked too many questions.

Older callers hung up. I cut the questions to five: first name, phone number, area, the space, and which conversation they want. Confirmations now go by text, and the agent never spells an email address back letter by letter. Every question is friction, and a caller who already doubts the machine will not sit through a long form read aloud.

Asking about timelines felt like pressure.

The agent used to ask whether a buyer was ready in thirty to forty five days, and that landed as a sales push. Now it offers a choice: someone comes out to look at the space, or we talk it through by phone first. The customer chooses the next step.

Every one of these mistakes cost me time, and a few probably cost me jobs. Both fixes came down to the same thing: respecting the caller's time and the caller's right to decide.

What is the better way to read customers who resist AI?

When an owner tells me customers do not want AI, I take it seriously. My own callers would rather talk to a person, and I did not pretend otherwise. But it helps to separate what a customer could be objecting to: the tool itself, or being misled, slowed down, or pushed by it. The last three are design problems, and an owner can fix design problems.

In the 4C Framework, this sits in Connections: the relationships the business depends on. A customer relationship is an asset, and a customer-facing AI tool spends from that account every time it answers. The question is not whether to use AI with customers. The question is whether the way you use it adds to that account or quietly draws it down.

It also helps to be clear about which work the agent should do and which work stays with a person. In my business the agent takes the missed call, captures the five answers, and books the next step. Prices, judgment calls, and the actual relationship stay with me. That sorting is the idea behind Automate, Augment, Advise, or Avoid, and it is what lets an older caller trust the assistant: it is not pretending to be the whole business.

How do you check your own customer-facing AI this week?

You can run this check yourself in under an hour, with no new software. If you use a phone agent, a chatbot, or an automated text follow up, sit down with it and walk through it as your least patient customer would.

  1. Call or message your own business the way a customer does. Write down the first sentence the tool says. If it does not make clear the customer is dealing with an assistant, change that first.
  2. Count the questions it asks before the customer gets something useful. Cut every question you do not need to book the next step, and move confirmations to a text where you can.
  3. Read each question as your oldest customer would hear it. Anything that sounds like a timeline, a budget test, or a push, replace with a choice the customer makes.
  4. Find the way to reach a person. Make sure it exists, that it reaches someone by name, and that you know how quickly that person responds.
  5. Ask three real customers what the experience felt like. Their answers will tell you more than any setting in the dashboard.

If you find that the tool can do things it should not, such as quoting prices or making promises, that is a control problem as much as a customer problem. I wrote about how I handled that in A Prompt Is Not a Control.

If you run a service business and want help setting up phone, chat, and follow up that your real customers will accept, AI consulting for service businesses covers how that work is done. If you are still deciding whether AI belongs in front of your customers at all, the AI Verdict gives you a written answer on where it fits and what to fix first.

Questions owners ask

Should an AI phone agent tell callers it is AI?

Yes. In my business most callers are fifty five and older and would rather talk to a person, so the agent says it is an AI assistant in its first sentence. An agent caught pretending to be a person loses the customer, while an honest one still has a chance to help.

Why do older customers hang up on AI phone agents?

In my business, the agent asked too many questions and older callers hung up. I cut it to five questions, moved confirmations to text, and stopped having it spell email addresses back letter by letter.

How many questions should an AI phone agent ask?

Only the ones you need to book the next step. Mine asks five: first name, phone number, area, the space, and which conversation the caller wants. Everything else can wait for a person or go by text.

How do I keep a person involved when an AI agent answers my calls?

Give the agent one direct way to alert a named person, and set a response time you will keep. My agent texts my phone when a caller wants to talk to me, and I call back within 24 hours, with reminders until the callback is marked done.

Next step

Start with a conversation.

Book a free 20-minute Intro Call with Shenard. We will talk through where AI fits in your business and a sensible first step.