What If Public Infrastructure Could Tell Us What It Needs?

When I worked at The Rockefeller Foundation, one of the most compelling frameworks shaping our work was what Judith Rodin -- then president of Rockefeller - called the Resilience Dividend. Rodin laid out the concept clearly in her keynote at the 2014 World Urban Forum in Medellin, and later expanded it into a book of the same name: resilience isn't just about surviving disruption - it's about the co-benefits that come from investing in resilience before disaster strikes. Job creation. Economic opportunity. Social cohesion. And crucially, financial return.

The math she cited was striking: research across U.S. government agencies found that $1 spent on preparedness can amount to $15 in savings in future losses. Every dollar invested in resilience before a disaster could save multiples of that dollar in recovery costs afterward.

The concept was intellectually sound. The data supported it. The challenge was never proving the theory. The challenge was making it financeable.

The Gap Between Theory and Investment

How do you get private capital to fund a bridge reinforcement today based on a flood that hasn't happened yet? How do you build the investment case for infrastructure spending around a future that is, by definition, uncertain?

Rodin was explicit about this tension. She described the need for innovative financing and public-private partnerships as essential to unlocking the upfront capital that resilience requires - pointing to initiatives like RE.invest, which Rockefeller supported alongside the White House and the U.S. Conference of Mayors, to help cities package portfolios of resilience investments that could attract private funding.

The framework was right. The financing mechanisms were creative. But the underlying challenge remained: the savings from resilience investment are future and diffuse, while the costs are present and concrete. Prevention stayed underfunded because the need was invisible until it wasn't. We rebuilt after the flood. We repaired after the collapse. We funded recovery because the need was visible and urgent.

I've been thinking about this framework again lately -- because I think AI is finally starting to build the bridge between the theory and the investment case.

What Changes When Infrastructure Can Talk

A recent study group report from Baukunst, Assembling the Future: Manufacturing + AI, makes a point that applies well beyond factory floors. The report documents how AI-enabled sensor systems are fundamentally changing what it means to monitor a physical asset: machine learning can now analyze continuous streams of sensor data to forecast equipment failures before they occur -- not after a breakdown, not during a scheduled inspection, but in real time, as the conditions that precede failure begin to develop.

In manufacturing, this is already generating clear returns. Avoided downtime. Reduced emergency repair costs. Extended asset life. The ROI is visible because the asset and its failure modes are contained enough to measure.

Now apply that same logic to public infrastructure.

Imagine a bridge embedded with sensors that continuously monitor structural stress, material fatigue, and environmental load. The bridge isn't just standing there -- it's generating data. AI analyzes that data and identifies, months in advance, where reinforcement is needed and what the consequence of inaction looks like. The bridge, in effect, tells you what it needs before it fails.

Or consider a subway system in a city that floods repeatedly. Each flood event generates data: where water entered, what it cost to remediate, how long service was interrupted, what the downstream economic impact was. AI can ingest that history, model the next flood scenario, and produce a specific, costed recommendation: here is what it would take to redesign this section of the system, here is what that costs, and here is what it will save when the next storm comes -- because the next storm is coming.

This is the resilience dividend, but now with the math made visible. Not theoretical math. Operational math, built on real sensor data from the actual infrastructure in question.

Making the Case with Real Numbers

Rodin noted in her keynote that President Obama had announced government partnerships with Google, Microsoft, and Intel to integrate big data with community resilience tools -- from sensors on city buses to simulations modeling coastal flooding threats. That was 2014. The vision was right. The technology wasn't yet ready to fully deliver it.

It's closer now.

The NYC subway offers a concrete example. Hurricane Sandy's impact on the system cost an estimated $5 billion in damage and lost economic activity. The investments that could have mitigated significant portions of that damage -- flood barriers, system hardening, redesigned drainage -- were known before the storm. What was missing was a financial case compelling enough to prioritize them ahead of everything else competing for a municipal budget.

AI-driven predictive modeling, applied to the actual system data generated since Sandy, could now begin to produce exactly that case for the next event. What did the last flood cost, broken down by component and location? What is the modeled probability and projected cost of the next one under different climate scenarios? What does mitigation cost, and what does it save? When you can answer those questions with specificity rather than estimates, the investment conversation changes. The resilience dividend stops being a concept and starts being a line item.

This is the mechanism that was missing in 2014. Not the framework -- Rodin had that right. Not the political will, entirely -- that was always contested but not absent. The missing piece was the ability to make the future savings legible enough, specific enough, and credible enough to compete with present-day budget pressures.

The Opportunity AI Creates Here

The Baukunst report describes AI as the first wave of automation that can operate effectively in messy, unstructured environments -- moving beyond the deterministic systems of prior automation waves into something that can perceive, adapt, and make judgment calls based on complex, heterogeneous data. That capability, applied to public infrastructure, means that the gap between a contained factory floor predictive maintenance case and a complex urban infrastructure resilience case is narrowing.

Rodin described the resilience dividend as having two components: the difference between how disruptive a shock is to a city that has invested in resilience versus one that hasn't, and the co-benefits that accrue from that investment -- jobs, economic opportunity, social cohesion. AI doesn't change either of those components. What it changes is our ability to quantify them in advance, with enough specificity to make the investment case to the private sector partners that resilience financing has always needed.

A bridge that can tell you it needs reinforcement before it fails is a financeable asset. A subway system that can model the cost of the next flood and the savings from mitigation is a financeable project. The resilience dividend was always real. What AI gives us is the ability to see it before we need it.

That feels like a significant shift. And one I'm surprised doesn't get more attention in the conversation about what AI is actually for.