Research Radar: The City as Mesh and New Ways of Organizing for Effective Problem Solving
This week New Jersey Governor Murphy announced the establishment of the New Jersey Innovation Authority (NJIA), stating that it would “ensure state government continues to deliver more efficient and effective services for New Jerseyans long after my time in office.”
For me, this milestone is personal. It formalizes the Office of Innovation (OOI) we set up back in 2018—a “bureaucratic hack” we introduced because the existing structures couldn’t move fast enough to allow us to build a problem-solving team focused on execution and implementation with new technology.
For me, this milestone is personal. It formalizes the Office of Innovation (OOI) we established in 2018—a “bureaucratic hack” we introduced because the existing structures couldn’t move fast enough to enable us to build a problem-solving team focused on execution and implementation with new technology.
The OOI was an innovative institutional design. I served as a member of the Governor’s cabinet, but we set up the Office of Innovation as a nonprofit, allowing us to hire quickly, raise philanthropic dollars, and build what has today become a 100-person team of technologists, designers, and policy professionals who can work across agencies on hard problems using the latest technology and social science.
During the pandemic, this bureaucracy hacking proved uniquely helpful in enabling us to work rapidly across agencies on urgent projects such as standing up the COVID-19 information portal, creating a symptom tracker, requisitioning ventilators, helping people obtain tests and vaccines, establishing a resident mailing list, and sharing data across agencies. As the independent law firm evaluation of the state’s COVID response noted: “Innovation allowed for effective collaboration, delegation, and prioritization of work.”
I wish Mayor Zohran Mamdani’s team had read the paper before announcing their approach to governance.
The City as Mesh: A Panoply of Institutional Forms
The question of what is the “right” way to organize for effective problem solving is at the center of a hugely important paper by Geoff Mulgan and Caio Werneck, "The City as Mesh: Options for Handling Cross-Cutting Tasks in City Governments," prepared as a background briefing for the Harvard Center for Cities and the topic of an upcoming webinar on Next Level Urban Problem-Solving: Institutional Innovation in Cities (Jan 7, 2026, 3 pm).
The City as Mesh responds to a problem that every leader recognizes immediately: the growing mismatch between how governments are organized and the kinds of problems they are now expected to solve. Most city halls still operate through 19th- and 20th-century functional hierarchies—transport, housing, health, education—designed for stable, bounded tasks.
But today’s defining challenges, from pandemics and climate shocks to homelessness, data governance, and care, are inherently cross-cutting. They span agencies, jurisdictions, sectors, and time horizons, leaving leaders with problems that no single department owns and no existing structure can easily address.
Mulgan and Werneck argue that this gap between institutional form and problem type is now a central constraint on effective city governance.
Their central claim is not that cities should adopt a single new structure, but that effective leaders need a repertoire of institutional forms—and the judgment to deploy different designs for different kinds of problems.
Their central claim is not that cities should adopt a single new structure, but that effective leaders need a repertoire of institutional forms—and the judgment to deploy different designs for different kinds of problems.
Centralization in the mayor’s office can sometimes compensate, but it quickly becomes a bottleneck. What’s missing is not leadership ambition, but a more developed understanding of how to organize horizontally—when, why, and with what tradeoffs.
Importantly, the paper is explicit about limits. Horizontal organization, they point out, can blur accountability, slow decision-making, or overwhelm staff if overused. In New Jersey, we surely faced limits with the old Office of Innovation model. The point, they say, is not to replace hierarchies, but to complement them.
Mulgan and Werneck are exactly right: cities can’t keep pretending that yesterday’s organograms will solve today’s cross-cutting problems, and the paper should be required reading for anyone interested in social change and governance.
This is an important contribution because it advances the call for “silo busting” by offering a structured way to think about which institutional designs work for which kinds of problems. The paper outlines a new “grammar” of institutional design: a vocabulary and taxonomy for organizing work horizontally, beyond traditional silos. The paper catalogs a wealth of interesting examples, each of which, on its own, is worth a case study; the richness lies in the comparison of forms.
Mulgan and Werneck are exactly right: cities can’t keep pretending that yesterday’s organograms will solve today’s cross-cutting problems, and the paper should be required reading for anyone interested in social change and governance.
Three Reflections on the New Grammar of Organizational Design
I wish Mayor Zohran Mamdani’s team had read the paper before announcing their approach to governance.
His early executive order organizes the top of New York City Hall into five deputy mayoral portfolios and an expanded set of offices on the topics of most concern to him, like housing and economic justice—important moves, but still largely a reshuffling of boxes and titles.
The deeper challenge Mulgan and Werneck are responding to is not simply “who reports to whom,” but how work gets done across boundaries: how intelligence is orchestrated, how teams form and dissolve, how data moves, how priorities are set, and how accountability is maintained when problems don’t map cleanly onto agencies. You cannot rapidly solve the kinds of large-scale, urgent, and important issues affecting New Yorkers with the same structures, as Andrew Rasiej pointed out.
From that starting point, here are three areas where the paper opens a path and where we can advance the conversation.
1. AI as a lightweight alternative to reorganization (and a powerful complement to it)
The paper makes a strong case for cross-cutting “command” functions—operations centers, common operating pictures, emergency coordination, and bullpens that help leaders see dependencies and act across silos. What’s newly possible, though, is that AI can now provide many of these coordination benefits without moving bodies, chairs, or org charts.
Mulgan and Werneck already nod in this direction when they note that “artificial intelligence will increasingly provide a layer of organisational design,” supporting task allocation, decision-making, and systems management. That line deserves to be operationalized. An “AI-native execution layer” (their phrase, borrowed from private-sector examples) can help a city do, faster and more continuously, what many reorganizations are trying to accomplish episodically: integrate signals, spot risks, flag contradictions, and route issues to the right people across agencies.
In other words: instead of reorganizing to get coordination, we can sometimes coordinate to avoid reorganization—or at least to buy time while testing what kind of structure is genuinely needed. That doesn’t eliminate the need for institutional change. Still, it can reduce the friction and cost of “trying” new forms, and it can help leaders make smarter choices about which of the paper’s menu of designs is actually warranted in a given context.
In Massachusetts, for example, our AI for Impact students, who build AI products with government agencies, created an AI chatbot to help workers across six call centers get access to the same information. In Boston, AI for Impact has designed a tool to ingest, summarize and organize citizen feedback and route it to the right place. The AI redirects the flow of information without the need to reorganize the institution.
2. The missing variable: skills, training, and the methods of work
The paper is fundamentally about organizational form—and it’s excellent at that. But form is only half the story. You can create task forces, innovation teams, new cross-cutting roles, and even whole “mesh” arrangements, yet achieve very little impact if people continue to work with the same mindsets, habits, and toolkits.
Mulgan and Werneck themselves flag the human dimension: collaboration depends on trust, relationships, and learned behaviors, not just structures. They also point to professional “cultures and training” as part of how knowledge and norms travel. But I’d elevate this: skills are not an implementation detail—they are a design choice. This is why we focus on changing mindsets and skill sets through our InnovateUS work, which emphasizes peer-to-peer innovation and AI skills training for the public workforce.
A productive next research question that flows directly from City as Mesh is: Which matters more for outcomes—changing the org chart, or changing the skills and methods of work? In many cases, we should at least test whether targeted capability-building (product management, service design, data literacy, change management, AI fluency, procurement fluency) can deliver the gains that leaders often seek through reorganization. When a city reorganizes, a parallel “skills architecture” prevents the new structure from reproducing old behaviors.
The paper gives us the taxonomy. The field now needs to pair it with training strategies—how to build teams that can actually inhabit these forms and make them real.
3. Beyond the “technocratic vs public-facing” split: engagement as operational capacity
One of the paper’s most valuable sections is the list of design choices cities must make because they “cannot coordinate everything.” But one of those choices is framed in a way that I think we should revise for today’s reality: “Technocratic or public facing,” suggesting that some tasks may be technical while others require greater public engagement.
That distinction is understandable, but it risks smuggling in the assumption that public engagement is mainly for legitimacy and “soft” issues. At the same time, “hard” problems are best handled behind the technocratic curtain. In practice, engagement can be a source of intelligence and an input to better technical design.
Even highly technical domains benefit from structured participation because the public and frontline workers often possess information that systems lack. No one knows more than Geoff Mulgan about collective intelligence, and there are lots of great examples from Peer to Patent to Smarter Crowdsourcing and more recent examples of using AI to source and organize public expert input that demonstrate why these are not in juxtaposition.
If the “mesh” metaphor is right, then engagement isn’t a separate civic layer—it’s part of the mesh itself: a way to tap distributed intelligence across the city (including civil society and residents) and improve operational performance, not just democratic aesthetics. The paper repeatedly emphasizes the importance of “tapping into wider networks of intelligence” and of building coalitions across distributed power. That logic applies as much to technical work as it does to values-laden work.
Where to go from here: the big opportunity is to extend City as Mesh into a more integrated agenda:
(1) AI-enabled coordination as a design layer
(2) capability-building as the missing lever that determines whether new forms actually function
(3) participation as an operational asset—especially for the “hard” problems where governments most need better intelligence.
JOIN US for the conversation about this pathbreaking paper.