The chatbot mistake: people belong at the counter, AI in the back office

You message your phone provider's AI assistant about a charge on your bill that you don't understand. It replies instantly, impeccably polite: it explains how to read a bill and suggests restarting your router. It cannot see or correct the charge, but it wants to know if you need anything else. You type: “I want to speak to a person.”

You wrote to get one problem sorted out, and now you have two: the original problem and getting someone to help you.

I think that scene captures one of the most tempting mistakes when introducing artificial intelligence into a business. Putting it right in front of the customer. It is the part everyone sees. You can show it at a meeting, put it on the website and announce it as an innovation. Meanwhile, behind that assistant, the same slow processes and the same overwhelmed people may still be there.

I would start there. Inside the business. With the work that eats into the time of the people who should be looking after customers.

Invest in the back office first

Before putting a machine in charge of answering, I would look at what stops people from answering well. The quote that requires looking up prices in several files. The paperwork that arrives in a jumble. The case whose status has to be pieced together from emails. Scheduling, reminders and follow-up that always get left until later.

Each of those tasks competes with a conversation that is waiting to happen. The customer waits while someone finds the document, checks what is missing or works out who was supposed to take the next step.

That is where I would put AI: preparing information, organising documents and getting drafts ready for review. I would start with one specific process and check how much work it actually saves. That is the idea behind matching AI to the processes it serves: choose the work that needs improving before choosing the tool you are going to show off.

I would also change how we measure success. I would want to know whether a quote arrives sooner, whether a call is returned on time or whether the person helping the customer has the case ready. A chat window that replies instantly can leave all the waiting that follows untouched.

What is failing at that phone company lies behind the chat: the account history, the reason for the charge and the ability to correct it are still out of reach. AI working inside the business could pull that information together and prepare the case for a person to resolve in a minute.

The client wants to speak to their lawyer

Think of a law firm. Someone arrives worried about their case. They want to explain what has happened, understand their options and know what comes next. They want to speak to their lawyer. They can fill in a form to provide some details, but that form should not become the relationship.

The lawyer needs to come to that conversation prepared. With the documents to hand, a clear timeline and the outstanding tasks in view. If they have spent the morning gathering all of that, looking after the client will have to compete for whatever time is left.

I picture a law firm organising that work with AI: preparing the paperwork, suggesting a summary to be checked and gathering what is needed before the meeting. The lawyer reviews it, makes decisions and talks to the client. That is one reason to think of the law firm as a production system: what happens before the conversation determines how much the person sitting down to listen can contribute.

One study offers a concrete example of that support. In Generative AI at Work, published in 2025, the authors studied an assistant used by customer service staff at a software company. With that support, the number of enquiries resolved per hour rose by 15% on average. The AI suggested answers; the staff remained in charge of the conversation and could edit or reject the suggestions. 1

That division of roles interests me. The machine helps prepare the answer. There is still a person at the other end.

The time you free up must benefit the customer

There is a business decision behind all of this. Saving time creates an opportunity; you have to choose to use it to serve people better. If every minute saved is filled with new tasks, the person will still be rushing through the conversation.

I would set aside some of that time to listen. To return a call before the customer has to chase it up. To explain something difficult without watching the clock. To follow up on a case even if nobody has sent a reminder yet.

That attention can also be a competitive advantage. In a market where speaking to someone means getting past screens and automated replies, offering a knowledgeable, accessible person can become a reason to choose a business and stay with it.

Small businesses have an opportunity here. They can use AI to increase their capacity behind the scenes while keeping the personal touch that sets them apart. It is one way to put the size paradox into practice: someone who understands the problem and can make decisions has room to change how the work gets done. I would use that freedom to give the customer more attention.

I also think that forcing people to battle with a bot is a terrible way to introduce them to artificial intelligence. If their experience of it is another obstacle when they need help, it will be hard for them to see it as an improvement. In the divide that rejecting AI could create, I discuss the cost of turning away from this technology. Businesses can contribute to that rejection through the way they use it.

They can also offer a much better experience: someone who responds having looked into the case, does what was agreed and has time for one more question.

That is the use of AI I would argue should come first. Taking work off people's hands so they can better meet their customers' expectations.

A business can automate a great deal and feel more personal than before. To achieve that, it has to decide where to put the machines and what it wants to do with the time they save.

Behind the scenes, machines doing the work. Out front, people looking after people.