Customers want to know a person is within reach

My argument in this series is that customers are best served when AI resolves their request in the organisation’s own systems while they wait, and a person is within reach when they need one. There is a strong objection to that argument, and this article is about it. When customers are asked in surveys, a large share say they would prefer to deal with a person even if the outcome and the waiting time were exactly the same. In a survey of 15,000 consumers by HubSpot and SurveyMonkey, published in August 2025, it was four in five.

I take that seriously. An argument that only works if you ignore what customers say is not worth much, so I want to deal with the objection properly.

I think the preference is about being able to reach someone

Most people’s experience of automation in customer service has been poor. It has been a phone menu that does not have their option, or a chat window that can explain the returns policy and cannot do anything about the return. Most of us have been stuck in both. If that is what a customer pictures when you say “AI”, then of course they would rather have a person, whatever the survey says about equal outcomes.

What customers fear, as far as I can tell, is being trapped. They fear a machine that cannot help them and will not let them past. Guidewire asked more than 4,000 European insurance customers, in a survey published in May 2026, what would give them confidence in an insurer’s use of AI. The most common answer was the ability to take an AI decision to a person. I find that reassuring, because it is a promise an organisation can make and keep.

A person should be easy to reach

My own view is simple. A customer should be able to reach a person whenever the organisation’s team is working, and reaching one should take less time than it does today. Outside those hours, the customer should know when a person will get back to them.

That is how we design oHallo. oHallo picks up the call at once and resolves what can be resolved in the organisation’s own systems. When a customer needs a person, the call is transferred with everything said so far, the account and everything already checked on the screen, so the employee picks up where oHallo left off and the customer does not have to explain the matter again. Where the organisation’s team keeps opening hours, a customer who calls outside them is told so, and a follow-up is booked.

The part I think gets overlooked is what happens to the queue. When oHallo resolves the address changes, the meter readings and the payment dates, the queue for a person gets much shorter. The customer who wants a person reaches one sooner than they do today, because they are no longer waiting behind everyone who only needed an address changed.

Today, one queue for every request

Each mark is a request. The mark with the brand border is a customer whose partner has died.

→ An employee

The customer whose partner has died is number 5 in line, behind requests that did not need a person.

With oHallo

→ oHallo

→ An employee

oHallo takes up the other requests as they arrive, always open. The same customer is number 1 in line, and the employee sees the full history.

Some customers will always choose a person, and they should get one. I expect far fewer to make that choice once they have seen a task completed while they wait, and the fewer who do, the sooner each of them is answered.

Being told is part of the trust

Customers should know when they are speaking to an AI. The EU AI Act requires it from August 2026, and I think the law is right. I also think it helps. The same insurance survey found that confidence rises sharply with familiarity, and people who use AI every day were far more comfortable with it than people who never do. Trust grows when customers see what the AI can do for them, and they can only see that if they know what they are dealing with.

There is one finding I want to be honest about. A field experiment on Alibaba’s Taobao marketplace, published in 2026, found that requests resolved by AI alone came back no more often than those handled by employees, but customers rated the AI’s manner lower. The authors put it down to how the AI communicated. An organisation can set that tone, and it should measure how its customers rate it.

Where I think people belong

The same research contains a warning I think every organisation should hear. When the system noticed that a customer was frustrated and only then passed the chat to an employee, it went badly. The chat ran longer, the customer was more likely to come back with the same problem and the rating fell. Where an employee stepped in earlier, the damage was smaller.

I read that as an argument for sending the requests that need a person to one from the start. When someone has died and an account has to be moved, or when a bill is disputed and the customer is worried about money, an employee should take it straight away, before frustration has a chance to build. Those are the moments the organisations that defend the personal contact are rightly proud of, and I want those customers to reach a person faster than they do today.

Where each request goes

The customer asks for a person

Passed straight away while the team is working, with a follow-up booked outside its hours.

→ An employee

Your procedures keep it for an employee

A bereavement, a dispute or a complaint goes to an employee from the start.

→ An employee

An action needs approval

A large refund or a change to a contract waits for an employee to approve it.

→ An employee

Your systems hold no verified answer

An employee takes over and completes the request.

→ An employee

Every other request

Identified, checked against your procedures and completed in your own systems.

→ Resolved, always open

The employee receives the request, the account and everything oHallo has already checked.

My answer to the preference

So my answer to the four in five is a promise with two parts. Every request is taken up when it arrives, at any hour, and the task is completed while the customer waits. And whenever the team is working, a person is within reach sooner than today, with the whole history in front of them. Outside those hours, the follow-up is already booked. I think most customers will take that offer once it is made to them honestly.