A new evidence firm for social programs.

What we’re working toward

To make rigorous evidence affordable enough, fast enough, ethical enough, and ordinary enough that every social program comes to count on it: to seek it, fund it, build it in, and act on it before the decision is made, so the work improves the lives of the people the money is for.

The evolving Prova story

Where Prova stands now.

The account I’d give today of what I’m building and why. I’ll update it as the venture changes.

I’m building Prova, a venture with AI at its core, to change how funders, governments, and operators measure how well social programs work, and what that measurement costs.

The cost of rigorous impact measurement has always kept it out of reach for almost every program that should have it. In early 2026, that began to change: large language models’ long-range reasoning crossed a threshold.

If it holds, the cost of turning everything a program produces, its case notes, interviews, transcripts, and field reports, into evaluation-grade evidence falls: work that once took a research team six months can be done in weeks, for a fraction of the price, with no drop in rigor. For the first time, evidence like that could come within reach of the funders and operators who were always priced out.

To find out whether that’s true, I’m building this venture.

In spring 2026, working on the problem with frontier models, I built what became an eight-layer operational architecture: 115 densely interconnected nodes, each specified across nine levels. The methods for rigorous evaluation are well established; they live in the literature, not in any one person’s head. What was missing was a structure that runs them at the cost and speed real programs need. Working layer by layer with frontier models, I built that structure, and if it works, it should lower the cost of rigorous evidence for everyone, not just Prova. It’s the foundation of the Prova Method. AI does the heavy lifting at every layer; the judgment that decides what the evidence means stays human. In theory, it lets a few expert humans run a rigorous evaluation that used to need a full research team.

The bet is that the same engine scales, from a fast, bounded diagnostic to measurement embedded in a government program serving millions, for years, without the rigor coming apart at the speed and cost real programs run at. Proving that is the venture.

That’s what Prova delivers today, in two ways: Reads, productized diagnostics, and Field, embedded multi-year measurement attached to programs already running. Both run on the same engine, the Prova Method.

My background

How I came to this.

I have spent more than fifteen years on versions of this problem: why evidence from programs and policy so rarely changes the decisions it is meant to inform. The methods are sound. What fails is the cost of putting them to work fast enough, and cheaply enough, to land before the decision is made. The problem is structural, and I learned it by living it.

I learned it first in India, in the one place that had, almost by accident, run the experiment. The 2014 law requiring the country’s largest companies to spend two percent of profit on social development created a vast pool of private-sector capital that, by law, someone had to judge was working. For nine years I founded and ran a firm built to design and run those programs, in road safety, school sanitation, pandemic response, and urban homelessness, measured to standards a US foundation would recognize, work that competed directly against the Big Four on rigor.

One program I co-designed with a foundation was for families who had been homeless for two and three generations, the kind of poverty that feels permanent. We built it so the outcomes could be measured, and had it evaluated by people outside the work. The evidence showed families better fed, and hours given back each night to women who’d had none. Knowing that, rather than hoping it, changed what the program could become. I’ve never forgotten how rare and satisfying that kind of knowing felt, or how many programs serving people just as deserving never get it. The block was almost always the same: the study a scale-up needed was too slow or too costly to commission at the speed board decisions moved. Abundant capital, an explicit mandate, good evidence, and the handoffs failed anyway.

Later, inside British Columbia’s economic strategy unit, I was designing measurement for the province’s mission-oriented economic strategy: work that brought together public goals, cross-government coordination, and the challenge of measuring progress as policy moves. It showed me the challenge from the public-sector side: serious decisions, time constraints, and policy environments where the work cannot pause while the proof is being built.

I have come to think the same constraint sits beneath even the best-resourced philanthropy, setting the limit on what anyone can actually find out before they have to act. Prova is what I’m building to do something about it. The reason to build it now is that new technology can finally make the cost of credible measurement fall far enough to expand who gets to know how well their work is going. The funders and operators I want to work with are fellow travelers: folks who would rather find out, honestly, than keep hoping. Prova is how we find out together. If that sounds like you, let’s talk.