The founder
I am building Prova to bring credible evidence within reach.
First I spent over fifteen years learning why it isn’t, advising at the decision-making level of capital markets, philanthropy, and government, where the right evidence was the one currency every party accepted and few could produce.
I have done this for shareholders and chairs, for funders who can afford to fail and governments that cannot, for global investment banks and individual donors, for the World Health Organization and the petroleum companies at the same table, for highway authorities and trucking unions, for municipal officials and families who had been homeless for generations, for corporate boards and the frontline nonprofits their money reached, and for survivors of trafficking and the companies in an alliance to end it. Each defined a successful outcome differently, and each had reason to hear that the work was working. My job was to find the strongest claim the evidence could honestly support, and to say so plainly when it supported less.
Where the work has taken me
The background
The incentive systems I have worked across.
Shareholder value, under public scrutiny
In capital markets the incentive is shareholder value, tested in public. I worked with the CEO and CFO of Tata Chemicals on a three-year investor-engagement program built around one quantified outcome, an increase in foreign institutional ownership, which rose from roughly six to eleven percent and brought in more than one hundred million dollars. For four years I prepared the predictive shareholder-question compendium Ratan Tata used at annual meetings, drawn from thousands of data points across operations on five continents. Through the 2008 global financial crisis, the most severe economic downturn since the Great Depression, I ran Bank of America’s India media response and worked with its India CEO, where a single wrong figure could move a frightened market. At that level a claim either survives public questioning or it fails in public.
Philanthropic capital, after the law changed
Philanthropic capital runs on a different logic, and after 2014 part of it ran on a new one: India became the first country to require large companies to spend on social programs. I spent nine years advising the senior leadership and foundation boards of global banks and multinationals on where to direct that capital, building two corporate foundations from a blank sheet and competing for the work against Deloitte, EY, KPMG, and Dalberg, with a firm I built from scratch. The useful funders were the ones who wanted to know what the money produced and could treat an honest failure as knowledge, which is exactly the incentive a government running programs in public rarely has.
Government, and what it cannot risk
Government runs on yet another set of incentives: electoral cycles that punish visible failure, budget years, and accountability rules that keep public money from being staked on uncertain experiments. Inside the Government of British Columbia I designed the measurement for a long-term flagship mission-oriented economic plan: work that advised Cabinet, ministers, and the Premier, often on 24-hour turnarounds. The method became the recurring standard the government kept using across the plan.
The currency
Credible evidence is the one currency a chair, a funder, and a minister all accept, judged against the claim it is asked to support.
Where Prova comes from
The method began before AI.
Producing that evidence well has always been expensive. It rests on scarce expert attention applied to everything an organization produces, its filings, reports, case notes, and field data, which is why credible evidence has reached the largest commitments and skipped the rest. That is the wall I met from every side of the table.
Prova’s method did not begin as AI. Its ancestor is the diagnostic and design framework I built across the social-program years, forty-two factors that made each program explicit enough to test: what it assumed, what it promised, the strongest claim the evidence could honestly support, and where a claim had to be narrowed. The framework worked. What it cost was the expert reading behind it.
One attempt along the way sharpened the question. We tried to build infrastructure that could trace philanthropic and public money from the giver to the person it was meant to reach. It did not ship, and it left a distinction I have not set down since: knowing where a dollar went is a different problem from knowing what the dollar produced. The first is transparency. The second is evidence, and it is the harder problem Prova is being built to address.
AI entered the work later. In 2023, inside the BC government, I began using frontier language models for careful, supervised synthesis: reading long policy texts, drafting baseline guidance, designing measurement proxies, and mapping responsibility across ministries. The tools carried the labor-intensive reading. They did not decide whether a design was sound, what a finding meant, or what leadership should do. That judgment stayed human.
That is the narrow claim behind Prova: AI can carry much of the reading that credible evidence work depends on, which lowers its cost and time; how far it does is under test. It does not lower the standard of evidence, and it does not replace the judgment that gives an evidence claim meaning. A wrong finding is worse than no finding at all. An honest null is one of the results the method is being built to produce and stand behind.
Having worked every side of that table is also what lets Prova stand apart from it. Prova Method grades the evidence and has no stake in the answer: it does not run the programs it assesses, and it does not bid on the evaluations it reviews.
Where this stands
Bring the question you are trying to answer.
Prova is taking on a small number of engagements at this early stage. The first conversation is a mutual read on fit. If you fund, run, or evaluate social programs and want to know the strongest claim your evidence can honestly support, I would like to hear the question you are trying to answer.