How new tech tools can streamline processes and make hidebound systems more rational
When Mayor Zohran Mamdani created a new Office of Mass Engagement, promising to listen more closely to the needs of everyday New Yorkers, it was a welcome first step toward earning the public’s trust. But as Mamdani seems to understand, listening better, and to a broader swath of the public, is only half of the recipe for a more responsive local government. City Hall also needs to solve the irritating quality-of-life problems more nimbly and proactively.
I saw how this played out when I served as Mike Bloomberg’s deputy mayor. Even though the City generally received high marks for performance, I wanted to understand how operations affected New Yorkers in their daily lives. So, I would meet five times a week with small groups of community representatives in each borough, starting each session with a simple question: How could city government do a better job for you and your neighborhood?
Almost every conversation included a common-sense request or complaint that ran into rigid interpretations of City rules. Can the timing of walk signs be adjusted to allow seniors to cross a wide street more safely? The traffic engineering manual said no. Why was the construction at a neighborhood park shoddy? Because procurement rules at the time required choosing the lowest bidder, not the best. A restaurant owner in Queens once hollered about getting fined by health inspectors for leaving his mozzarella out of the refrigerator while he was making pizzas.
To understand how it got this way, you have to go back over a century to the Progressive era. In response to Tammany Hall’s corruption and abuses of power, reformers went overboard with rules, compliance requirements and oversight that have stayed with us to this day. Union leaders in many jurisdictions are eager to protect their workers from supervisory abuses; understandably, they embedded restrictions and limitations in the work rules set out in bargaining contracts. Unfortunately, this leads to compounded restrictions, further limiting flexibility for supervisory staff.
In this model, the traffic engineer looking at the wide street can pull up pedestrian counts, crash data and the age profile of the surrounding residents in seconds and adjust the walk signal without waiting six months for a manual revision.
To be sure, city managers, elected officials and residents all want decisions that follow rules, reflect public values and treat people fairly. Too much control, however, creates nonsensical rigidity. It robs traffic engineers, procurement officials and restaurant inspectors of the discretion to develop pragmatic solutions to neighborhood problems.
Artificial intelligence, paired with updated civil service and management rules, offers a new way to strike a healthy balance. In a recent journal article in Urban Governance, my coauthor Juncheng “Tony” Yang and I call it “accountable discretion”: giving frontline workers room to use judgment on the problem at hand, while using technology to make every decision traceable, reviewable and fair.
In this model, the traffic engineer looking at the wide street can pull up pedestrian counts, crash data and the age profile of the surrounding residents in seconds and adjust the walk signal without waiting six months for a manual revision. The health inspector, upon walking into the Queens pizzeria, can see that this is a first-time issue in an otherwise clean kitchen, issue a warning rather than a fine and offer guidance on how to meet the intent of the rule. The procurement officer can weigh a contractor's past performance against the low bid, with the comparison built into the system rather than buried in a file cabinet.
AI can do the work that used to make discretion difficult, if not impossible: It pulls the relevant data to the moment of decision, handles the routine paperwork that eats a field worker's day and leaves a clear record of who decided what and why. The frontline worker spends less time typing and more time judging — and every judgment is visible after the fact.
Supervision changes, too. Instead of writing rules that force every inspector to act the same way, a supervisor can watch how her inspectors actually use their judgment — and intervene only where the pattern looks off. If one inspector is issuing warnings at twice the rate of her peers in similar neighborhoods, that is a coaching conversation. If fines in one borough track the race of the business owner more than the severity of the violation, that is a problem the old system would have hidden for years. A supervisor responsible for dozens of field workers cannot read every case file; AI can and can flag the handful that deserve a second look.
Accountable discretion only works with strong human oversight, clear limits on surveillance, serious investment in staff training and straightforward ways for residents to question a decision and push back.
None of this is automatic. AI tools are only as good as the data they draw on, and most City data still lives in siloed, aging systems — locked up by privacy rules written for a different era or trapped in software that does not talk to the software next door. Without fixing the plumbing, faster answers will just be faster wrong answers. There are also real risks on the other side: agencies that monitor their own staff too closely, or managers who rubber-stamp whatever the algorithm suggests. Accountable discretion only works with strong human oversight, clear limits on surveillance, serious investment in staff training and straightforward ways for residents to question a decision and push back.
AI can also change how a city listens to its communities, advancing, over time, an intervention in the life cycle of a developing problem. Residents routinely document deficient park construction on Instagram and other public platforms weeks before the issue reaches municipal channels. To close this latency gap, some cities are now deploying anonymized survey instruments and AI-based sentiment analysis to surface neighborhood-level quality-of-life concerns — including dissatisfaction with specific contractors or facilities — earlier in the complaint cycle. AI should monitor sensors to get ahead of problems: A spike in an air-quality sensor in the South Bronx, for example, could trigger a response before the first asthma complaint comes in.
Rather than removing human judgment, AI, when properly used, can free up more time and space for frontline common sense. An AI helper can sort records, summarize documents, flag patterns or handle routine steps. AI can increase responsiveness by dramatically expanding self-service. Why should a family opening a restaurant in the Bronx have to chase permits across a dozen agencies, when a single AI-powered front door could take their information once, in whatever language they speak, and route it to every office that needs it? Residents receive faster, clearer answers and services along with new levels of transparency. These shifts will free up more time for judgment and help public employees focus on areas that deserve more attention.
None of these changes will come easily. Civil service procedures are governed by state and local statutes, encrusted by decades of litigation. Union leaders, often skeptical of both managers and elected officials, will need to be persuaded. Public anxiety about AI misuse must be addressed directly. Yet this is precisely the moment for reform, when mayors can get out in front of changes that are coming regardless.
Cities need rules and protections regarding AI that, for example, require it to be deployed to expose bias, not entrench it. Mayors with strong labor relationships should engage now — demonstrating that humans remain in the loop at every decision point, and that accountable discretion is itself a recruiting advantage with a new generation of tech-fluent public workers. Thousands of New Yorkers will tell you, on any given day, that you cannot win against City Hall. But it is their City Hall, and what they want is common-sense responsiveness. Public employees take these jobs because they want to serve, not to hand out fines over mozzarella. With the right tools and the right rules, both sides win. Struggling New Yorkers have the most to gain, and efficiency can be turned into better and more affordable services.
This piece draws on an academic paper by Stephen Goldsmith and Juncheng “Tony” Yang, “AI and the Transformation of Accountability and Discretion in Urban Governance,” in the journal Urban Governance.
