Connect residents with the city workers who can solve their problems. Don’t replace those workers.
In mid-September, I applied for an IDNYC at the East Harlem Neighborhood Health Action Center run by the New York City Department of Health and Mental Hygiene. The worker at the desk was friendly and professional. They took my photo, and I took my application receipt and left.
A few days later, I had a question about the card’s issuance, so I called 311. The customer service representative who handled my inquiry was patient and pleasant and swiftly brought in a Mandarin interpreter (that’s my native language). They told me that 311 did not have a service-request category for IDNYC, so no case number could be opened — it could only be forwarded as program feedback with a reply in 14 business days. I thanked them for the effort and moved on.
As it turned out, the card arrived ahead of schedule. The answer to my question existed all along; there was simply no way for me to reach the person who had it.
The experience got me thinking: The ability to reach a person in government to quickly resolve problems is absolutely essential. My problem was a fairly small one; others, from an apartment without heat in January to a housing voucher stuck in limbo while an eviction case moves forward, are far larger.
Yet today, New Yorkers have one of two options. They can call 911 if they have a genuine emergency that demands the rapid dispatch of the police, firefighters or EMTs. Or they can call 311, where their request is routed to the agency in question, and sometimes they hear back about what became of it.
It’s exceedingly difficult for an ordinary person to get a public official on the phone to listen to their question and engage in actual problem-solving back-and-forth the way private-sector tech support, as frustrating as it can be, often works.
We can change that.
When New Yorkers ask the City for help, a tool should make it easier to actually connect them with the worker responsible to get a real answer from a real person in the know. This is where we should begin in crafting a city’s AI policy: tightening the relationship between those who need a service and those who provide the services.
Step one: listening
It starts with understanding what people care about, whether or not those concerns happen to align with what government wants to hear.
In early January this year, Mayor Zohran Mamdani established the Office of Mass Engagement, promising to build feedback loops between residents and City Hall and asking to be judged “by whether engagement changes outcomes.” But to date the office has mainly been used as a government outreach wing bringing New Yorkers out to Rent Guidelines Board hearings and otherwise organizing them to rally behind the mayor’s biggest priorities.
A more important use of the office would be to listen more systematically to residents’ concerns and problems at scale. There are models to follow.
When ride-sharing rules divided the people in Taiwan, we used a tool called Polis. A participant sees a statement from someone else and may agree, disagree or pass. There is no dunk button and no reply thread, so the incentive is to find language others can live with. Statements that travel across clusters become visible as shared proposals. Those proposals then went into stakeholder discussion and the slower work of regulation. The full vTaiwan record is public, so anyone can see how the method works. The whole point is that government doesn’t control the conversation; people, engaged with one another civilly, do.
The public sector can and should adopt its own version of such a listening tool. Engaged California, run by the state’s Office of Data and Innovation, invited government workers to discuss organizational efficiency enhancements. That project was not a public hearing where people got in long lines to speak for two minutes at a time. Instead, over 10 weeks, 1,469 people took part in a complex, iterative back-and-forth. They left 2,477 comments, from which analysts, using AI and then human review, drew 2,627 specific ideas. The code is on GitHub.
Step two: closing the loop
Once government hears concerns clearly, it has to stay in conversation with the people who raised them.
Germany is testing a version of this nationally. Its federal innovation agency, SPRIND, asked residents where dealing with the state drives them up the wall, and collected more than 27,000 submissions. When residents then voted on what to fix first, “administration as a service rather than control” came in second. Another priority on the list: “find the right office immediately.” That is precisely the gap 311 leaves open. Teams are now building prototypes for the priorities residents chose.
In New York, closing the loop would mean a few concrete changes. Every request should get a case number, even when no category fits. AI can route it not to an agency’s general queue but to the team that actually owns the issue, and keep the thread open so residents can ask a follow-up question instead of waiting 14 business days for a one-way reply. Interpretation, which 311 already does well, should come built in. None of this replaces city workers; it puts residents in front of them.
New York City’s largest public-employee union, District Council 37, has rightfully asked for a seat at the decision-making table as government explores new ways to use AI. That’s understandable: AI should be used primarily to enhance the performance of city workers, not to slash jobs. And technology should never be adopted for its own sake. Tools should have clear purposes — and when they’re not doing their job, they should be retired.
Rather than supplanting human judgment or turning the bureaucracy into an even more faceless force, AI can be harnessed to get more people involved in government and more easily connect people to the officials who can help them with the problems they face. That should be our vision. It’s well within reach.





