Companies are equipping their agents with artificial intelligence. But a quieter transformation is already under way: their customers are doing exactly the same. They arrive better informed, better equipped, better prepared — but not necessarily better informed on the substance.
A passenger lands four hours late. Until recently, claiming compensation meant finding the rules, checking eligibility, working out the amount, gathering evidence and writing the claim. Today, they tell an AI: “Check my rights and write my claim.” Seconds later, they have a structured letter, a legal argument and a figure.
AI reduces the information asymmetry
The company knows its contract, its procedures and thousands of similar cases; the customer usually discovers all of it when a problem arises. With AI, customers can summarize terms and conditions, compare contracts or identify their options. That is real progress for them — with a paradox: AI can make an error look as convincing as a perfectly valid right.
A well-written claim no longer proves much
In thirty seconds, almost anyone can produce a letter that reads like a legal brief. The reasoning may be right — or rest on a non-existent reference and claim compensation that was never due. The appearance of expertise is now easier to produce than expertise itself, while a customer who writes four clumsy lines has no fewer rights.
At Sanef, on the A13-A14 — the first historic French motorway fully converted to free-flow tolling — nearly 54 million passages were recorded in the first year. At that scale, billing disputes are a daily reality. Tomorrow, customers will arrive with a case already argued and legally framed by their AI. The formal quality of a request must no longer be mistaken for its merit.
Stop trying to find out who wrote the letter
Detecting AI-generated text makes little sense: the line will soon be impossible to draw, and the question has little operational value. Flight, toll or bank transaction: the agent must check the facts — the time, the timestamp, the applicable rule. The more AI polishes the form, the more the company must return to verifiable facts.
Even the Voice of the Customer gets harder to read
When thousands of customers have the same assistants rephrase their complaints, messages start to look alike. Verbatim analysis may end up measuring the assistant’s style rather than the customer’s real state. Voice of the Customer programs will need to cross-check verbatims with journey facts: actual events, successive contacts, relationship history.
Agents will need an AI to check the customer’s AI
Most AI projects in customer service target productivity: summarizing, suggesting, drafting. Agents will increasingly need another capability: instantly verifying the regulation or clause a customer cites. Internal AI becomes less a tool to “write better” than a way to retrieve the right rule, the right document and the facts needed to decide.
Tomorrow, the customer may not even write the claim
Today, customers ask AI to prepare. Tomorrow, they will ask it to act: get a refund, renegotiate a contract. The customer’s AI will deal directly with the company’s systems — and its AI. How far will a company let its AI decide when facing an AI whose mission is to defend the customer’s interests?
The company is no longer the only one augmenting its capabilities
The answer is not to distrust AI-equipped customers, but to make one’s own rules clearer, knowledge more reliable and decisions more verifiable. The question will no longer be whether the customer looks like an expert. It will be to verify, much faster, whether they are right.