Policy analysis in the age of AI
Ariel Kennan · July 22, 2026 · 1 min
Max Ghenis spoke on policy analysis in the age of AI at the Aspen Institute Financial Security Program’s Benefits Leadership Forum in Chicago, where Ariel Kennan also took part — with thanks to Tim Shaw and the Aspen Institute team for the invitation.
The argument, in short: AI capability and usage keep growing, but frontier models still miss on questions about taxes and benefits. We track exactly how much at policybench.org.
Those questions split in two. Some are about what the law is — am I eligible, and for how much? That has a correct answer, set by statute. The Axiom Foundation encodes the law straight from the statute, so every rule traces back to the source that sets it. Others are about what a change would do — the kind of prediction PolicyEngine makes today.
The opening is that both sides get bigger. Just as AI agents let us think bigger on the policy encoding side, they open new doors on estimation and forecasting — predictions across a far wider range of outcomes, from maternal health to air quality to Medicaid wait times.
The policy community is turning to AI for advice faster as AI reshapes society. We can use AI itself to give AI — along with the public and civic institutions — the tools to answer these questions accurately. Technology is both cause and solution, should we choose to build.