New York City wants a human on the line

New York City wants a human on the line
New York City wants a human on the line
On October 20, New York City’s Department of Consumer and Worker Protection (DCWP) proposed a rule aimed at what it calls customer service “doom loops.” If it is adopted, businesses that use chatbots or automated phone systems with New York consumers will have to say so and give customers a real path to a person. The city says it would be the first municipal rule of its kind in the country.

What DCWP proposed

The proposal covers “automated customer service systems,” a category that includes AI chatbots and voice-response systems on the phone. It treats three practices as deceptive or unconscionable:

  • No prompt disclosure: failing to tell consumers right away that they are dealing with an automated system rather than a human representative.
  • Loops instead of help: advertising customer service while repeatedly sending consumers to menus that don’t address their question, or answering with unresponsive, irrelevant or wrong information.
  • No way out: failing to offer a single-step way to reach a human representative or, when no one is available, instructions on how to get human help.

DCWP would enforce the rule citywide. Violators would be liable for restitution to affected consumers and civil penalties, with fines starting at $525. The proposal opened a 30-day comment period, and DCWP will hold a public hearing on November 19. The text can still change before it is adopted.

Part of a broader enforcement agenda

The proposal does not stand alone. On October 1, DCWP and two other city agencies issued a joint policy statement saying that a company cannot avoid responsibility for a violation because it relied on an algorithm, AI or other technology. The same day, the city began enforcing its Click-to-Cancel rule on subscriptions.

Taken together, these steps set a clear expectation in New York City: what an AI agent tells a customer counts as what the company tells that customer.

What changes in the design of an AI customer service agent

Many mid-size companies already use some automation in customer service, from phone menus to website chat. A rule like this turns several good practices into requirements:

  • Disclosure at the start: the first message or prompt says the customer is talking to an automated system.
  • Clear escalation triggers: the agent knows when to stop, for example after a failed answer, a complaint, a billing dispute or a direct request for a person.
  • A one-step handoff with context: the person who takes over sees the conversation and the customer’s data, so the customer doesn’t start again.
  • Answers from real data: an agent connected to the CRM and the help desk is less likely to give irrelevant or wrong answers than one working from generic scripts.
  • Records: each interaction is logged, so the company can show how a request was handled and when it was escalated.
Customer message → Disclosure → Answer from CRM data → Escalation trigger → Handoff to a person with context → Logged outcome

Where to start

  1. Map every automated touchpoint that answers New York customers: chat, WhatsApp, email bots and phone menus.
  2. Check disclosure in each one, from the first message or prompt.
  3. Define and test the path to a person, and make sure it takes one step.
  4. Turn on logging so every automated interaction and handoff can be reconstructed.
  5. Follow the November 19 hearing and adjust once DCWP publishes the final text.

Does your AI customer service know when to hand off?

We design AI agents for WhatsApp, chat and phone that answer from Zoho CRM and Zoho Desk data, say clearly that they are automated, and pass cases to your team with full context.

See how it works