How AI in government can turn city data into better services

This is Government’s Future Frontiers, the podcast from Deloitte that asks questions today to help create tomorrow. I’m Bill Eggers, executive director of Deloitte’s Center for Government Insights, and in this episode, I’m sitting down with former Indianapolis Mayor Stephen Goldsmith—he’s now the Derek Bok Professor of Urban Policy at the Harvard Kennedy School and director of the Bloomberg Center for Cities’ Data-Smart Cities Solutions program.

Steve and I have collaborated on government reform for nearly three and half decades, including even writing a book together. He’s long been one of the leading thinkers and practitioners of government innovation, especially in cities.

In the conversation you’re about to hear, we talk about the concept of government as a cognitive system, discussing next-generation maintenance technologies and sensors, the role of digital twins, and we look at what characterizes governments that move from analysis to action. 

Bill Eggers: So, governments have invested heavily in data dashboards, analytics, and stat programs. Based on your research, where do you most often see the breakdown between sensing problems and actually improving outcomes? And how can what you’ve talked about in terms of the use of AI may be help to bridge that gap?

Stephen Goldsmith: Yeah, I mean, the famous stat programs that we all celebrated—CompStat with Bill Bratton or CityStat—were top-down systems, right? But increasingly, there was no connection between what was measured and what happened.

Eggers: Right.

Goldsmith: It was measured and didn’t change any of the actions in the city. And, so, what we are proposing is that artificial intelligence can broaden—I don't like to use the word “democratize” because everybody uses it—the use of data throughout the enterprise. So why do you have an outlier? Why do you have more potholes here? Why do you have a drainage problem?

Then, allow a broader number of individuals in the organization to use their discretion to make these natural-language inquiries of the data and solve problems. So, you can identify the problems more clearly, you can identify them in real time, and iterate them more quickly. You can broaden the number of people who can come up with answers, and you'll be more responsive.

Eggers: So, in [the] cities you’ve worked with, what distinguishes governments that turn predictive insights into action from those that are stalling at analysis (as a kind of follow-on to that)?

Goldsmith: My first answer is [that] we’ve just barely scratched the surface, and what’s occurring is that the old regime of 100 years ago is still the prevailing way we deliver government.

So, unleashing the power of data means granting discretion to people who currently are rule-driven rather than problem-solvers. And so, what it means is—and you see this across the country, in the United States—[that] there are cities that have enormously talented chief data officers, right?

Eggers: Right.

Goldsmith: They come up with all sorts of insights, but they need to have a partner who’s the chief administrative officer or the deputy mayor of operations, who has the authority to work across the organization and change the delivery system. So, we’re seeing not too many cities that have really unleashed that power on a day-to-day basis.

Eggers: So, we’re doing a lot of work—a piece—on kind of organizational structures and radically new ways of looking at organizational structures in government, especially around back office, calling it platforms and pods.

Any thoughts on building on that? What organizational structures and incentives would need to change?

Goldsmith: Organizational structures and incentives are two interlocking but different questions—and complex questions. On the organizational question, I have two totally inconsistent answers for you: One is we could use generative AI to run across the agencies, so long as you have somebody with authority over the top, and not reorganize everybody.

That’s what appears to be the suboptimal solution, but the advantage of that is you don’t have the chaos that’s unleashed by organizational change. I answer your question that way only in saying I think, if you had leadership at the top, task-force cross-agency understanding, [and] the ability to set up your data so you can look at it across agency, you can make many of the accomplishments while you reorganize your government.

So, if you put reorganization first, you may delay the power of the solution. So, I think it’s kind of a hybrid solution.

Eggers: Your latest piece looks at parks—which I really enjoyed. And so, you have a section on next-generation maintenance—proactive, efficient, sustainable. Can you speak a little bit on that?

Goldsmith: So, one of the big revolutions in the way government operates will be to unleash the power of IoT [Internet of Things] sensors and connect those to the maintenance and construction, as well. But in the parks example, [it’s about] maintenance. And so, we should be able to anticipate the broken bench, the light out, the broken slide, from vibration sensors and cameras, and the like.

And so, putting that information into action will produce a much higher-quality experience in the parks. And I think that not only will we be able to save on accidents and liabilities, but the experiences of people who use a park will go up dramatically. And just—oh, by the way, we'll also be able to determine usage patterns.

So, if there is a slide that’s not used or a ballpark that’s not used—the ability to kind of change the utilization of parks based on real knowledge of how many people are there and what they’re doing is very exciting as well.

Eggers: Do you see augmented reality playing a role? I’ve always thought, like, in terms of national parks, that there’s so much they could do with augmented reality.

Goldsmith: Well, there’s a lot that can be done with augmented reality in the planning process for sure, right? And one of the problems with community engagement is it’s not really engagement and it’s not really informed engagement.

So, if you could use augmented reality to go through scenario choices, you could have true engagement and make better decisions.

Eggers: And building on that, that’s my next question, we’re seeing governments all over the world use simulations, digital twins, policy models to test choices before rollout. Where have you seen this actually add the most value? And where is it overhyped? Love to get your thoughts on the actual use of this and leading to better outcomes.

Goldsmith: Well, there is a lot of talk about digital twinning. Many cities have somebody capable of doing the twinning. And I think there’s a fair number of planners who are using it.

But, in terms of day-to-day scenarios, I’ve been trying to find cities that take their environmental and pollution data and their traffic data and their public health data, and then use a digital twin to change the way traffic patterns, or types of traffic [that] are in a neighborhood, in order to produce [...] so that you could twin it and look at the implications of it. But on the complex solutions, I would say digital twinning is not used as much there as it is used in the planning department.

Eggers: I mean, I think you look at disaster response, you look at even policy twins, I think there’s so many opportunities, and it doesn’t have to be the very, very expensive kinds of digital twins, necessarily.

But it does feel like we’re at that point where people are spending a lot of money on these in infrastructure, transportation, and so on, and even large-scale population models. You know, there should be opportunities to use them. But not every problem, of course, needs a digital twin, and how should leaders decide when simulation is worth the effort versus when simpler predictive tools might be enough?

Goldsmith: Well, I think there’s a risk here that perfect could be the enemy of the good. Complex simulations would be helpful, but we’ve been doing a lot of work on curbs and sidewalk—taking digital platforms, all the data that relates to curb and sidewalk. You can figure out how many pedestrians are where on a particular moment of the day or day of the week.

You could figure out whether the curb should be altered for commercial loading zones from 2:00 to 4:00 and night-time parking from 4:00, or where should the TNCs drop off. Those don't require complex digital twinning. They just require taking as much of the available digital information and making policy decisions.

And that situation exposes a bigger problem, which is, there’s multiple agencies involved that touch these issues, and there has to be a better way of resolving and utilizing the data cross agency.

Eggers: So, what would be, with that kind of data, how would that impact different policy decisions or operations?

Goldsmith: So let's just play with this for a while: It would mean that, from 4:00 to 6:00 on Saturday night, the price of parking goes up. It would mean that, from 10:00 to 12:00, commercial loading zones would increase in the number of zones, and you’d have an advantage for electric vehicles. It would mean that you couldn’t drop your scooter off in front of a restaurant for people to drop off without a financial penalty.

It would mean that the use of the curb was more connected to the quality of life in the community, with dynamic variations in management. You could look at, as you know, anonymized cellphone data to determine where the density of pedestrian traffic is so that it would change how you handle valet parking in outdoor cafes.

So, if we think about systems, that’s the power of AI—to be able to understand the system and make modifications in different agencies that affect that system.

Eggers: So, speaking on infrastructure, you’ve also written about digital innovation and infrastructure in addition to digital curb management, vision zero, smart signalization, [and] materials innovation, in which I think there are so many interesting things there.

But you write that jurisdictions have yet to fully capitalize on these opportunities. So what’s the promise here, and how can governments better capitalize on this promise?

Goldsmith: Let me start with a problem. I think part of the problem is life-cycle costing. So, if you issue a RFP [request for proposal] to build a bridge, and if the cost of building it goes up because of sophisticated sensors that allow you to maintain it better, then there is a tendency not to pick that bid.

But if you said, “What’s the 20-year life-cycle cost—construction and maintenance—of this piece of infrastructure? Then you would build it with digital infrastructure as part of the physical infrastructure.

So, I think there’s not enough connection between the build and the maintenance. The pricing models are wrong and many of these departments are risk-averse, right? So they’re not excited about new sensors. But if you thought about digital infrastructure as a critical component of every piece of physical infrastructure—yeah, you could improve its operation and maintenance.

Eggers: Lastly, by 2030, what would convince you that this idea of cognitive government and all the things we’ve been speaking about have genuinely improved service delivery, readiness, operations—and not just added smarter tools?

Goldsmith: Well, it’s about to. I’m not sure that it’s fair to say we’ve reached that point yet. But if you think about this, our conversation today, as one not about digital tools, but about responsiveness and trust, so, how does government anticipate and understand the problems of its residents and, in a timely manner, [being] consumer-facing, fix those?

And if we could unleash these tools to understand not just what they’re calling 311 about, but what they’re frustrated about—if we could measure time to response, not just response, but time to response—I think we could listen better with digital tools. We could respond better. If we give public employees discretion to problem-solve, they can do that as well.

So, I think we’re about to the point if we can change the procedures of the bureaucracy, that we have the digital information to make government truly more responsive.