New York proposes a right to reach a human behind AI customer service: schools and employers should copy it

New York City's Department of Consumer and Worker Protection (DCWP) proposed a rule on 20 October 2026 to end "doom loops" in automated customer service, including AI chatbots and phone systems. Businesses would have to disclose when a customer is dealing with an automated system such as AI, provide an easy route to a human representative, and give meaningful help if they advertise it. They could not avoid responsibility for unlawful conduct that results from their use of AI. Fines would start at $525, violators could owe restitution to harmed consumers, and the public has 30 days to comment. If adopted, the department says New York would be the first US municipality able to enforce this type of protection.

The department points to everyday situations: bank problems, refunds, returns, package deliveries, prescriptions and student loans. It cites Gartner figures showing chatbot resolution rates ranging from 17% for billing disputes to 58% for returns and cancellations. In plain terms, a customer with a billing problem is the least likely to get it fixed by the chatbot. Commissioner Samuel Levine said: "We are not waiting for Washington to come to the rescue of working people."

The story reaches well beyond one city. Every school office, college helpline, council service and employer now puts an AI system between the public and a decision. Parents chasing a place or a refund, a student querying a fee, an employee with an HR question: each meets the same loop when the system cannot help and no person is reachable. The rule treats that as a failure of the organisation, not the customer.

Dan Fitzpatrick says this is the right test, and nobody needs a law to apply it. AI can handle the routine question well. What it cannot do is take responsibility, make an exception or notice that someone is upset. The organisations that keep trust will be the ones where a named person stands behind every automated front door.

School and business leaders can act now. List every place where AI answers the public or staff. At each one, check that a person can be reached in one step, that the system says it is automated, and that someone owns the outcome when it fails. Test it as a parent or customer would, with a question the system will not understand. Ask suppliers what happens to the conversation when it is handed to a human.

Dan Fitzpatrick works with school leaders and educators on how AI is used in practice, including what happens to pupils, parents and staff when a system gets it wrong. He writes and speaks on AI, learning, work and trust.

Dan Fitzpatrick is available for interview and comment on this story. Available today for live or pre-recorded radio, TV, podcast and print interviews, with a broadcast-quality remote setup for live or pre-recorded down-the-line interviews. Contact: dan@theaieducator.io