Customer Relations: "The AI Bot Behaves Like a Nightclub Bouncer — Unfriendly, Rigid, Making the Customer Feel Unwelcome"

According to the 2025 Ipsos BVA Customer Services Observatory, conducted among 5,000 people across Europe, only one in two chatbot users reports being satisfied. More tellingly: 90% of French consumers would rather wait longer to reach a human agent than interact with a virtual one.

What do people hold against these bots? Off-target answers, robotic forms where the relevant option is always missing, response times that feel endless. Even as the technology has matured, many bots remain incapable of understanding natural-language requests, sometimes trapping the user in loops with no way out.

More often than not, customers have to fight to reach a human being — or even to find a customer service number they end up searching for on Google. The bot then behaves like a nightclub bouncer: unfriendly, rigid, conveying the impression that the customer is not welcome.

Why Are Bots So Dumb?

While many companies boast a large fleet of "virtual assistants," the bot is frequently disconnected from the business. It is unaware, for instance, of the latest promotional campaign or even the product catalogue.

The dumb-bot syndrome strikes companies of every size

Our national railway operator's bot was long unable to say what time the train you were on was due to arrive. The dumb-bot syndrome strikes companies of every size.

What is visible reflects what is invisible. The bot often betrays an obsession with cost-cutting, or the magical belief that AI will unclog customer service at zero cost. Yet a bot demands time: to design customer journeys and to keep content up to date across a fragmented organisation. It is equally naive to think a quick technical fix will suffice. A solid, secure platform with near-instant response times is required, along with baseline integration with internal systems to break the bot out of its silo.

Organisational adaptation is equally essential. A bot imposes cross-functional working, and it introduces new processes for supervising and enriching it in real time. A bot must evolve: "set and forget" simply does not work.

When Savings Cost More Than They Save

The first step — often a surprising one — is to have the leadership team test the bot themselves. Customer service, frequently outsourced or offshored, is not always a management priority; it tends to get attention mainly during crises or budget reviews. The other urgent action is to put in place a quality-monitoring framework grounded in user feedback.

It is also vital to measure the value created or destroyed by good or poor customer relations. Experience shows that a swift response to a request has a significant commercial impact. This is how companies avoid penny-wise, pound-foolish savings that destroy far more value than they preserve.

Finally, a cross-functional project team must be assembled: domain experts, IT (architects, AI developers) and user-experience specialists. Within a few weeks, a bot can gain many IQ points — and project a far more positive image of the company it represents.

Pierre Guimard is a Partner at Takh AI Consulting.