AI research for economics, management and the social sciences

AI research you can check.

EstimandLab builds AI research systems for economics, management and the social sciences. They run the analysis, rerun it, and check their own numbers.

Why checking matters

Given a paper’s method and data, the best AI agent in a 2026 test matched the sign of 91% of published results. Run it twice, and about half the individual estimates moved.

Source: summary of an ETH Zurich study of 48 papers verified by the Institute for Replication. That gap between a good first answer and a stable one is the problem our systems are built to close.

Method

Known answers first

Before a system touches client work, we test it on published studies whose answers are already known, and report how often it gets them right.

Scorecards

We publish the misses

Every study we release carries a scorecard: what matched, what did not, how much results moved between runs, and why.

Have a question that needs a defensible answer?

Tell us the question, the data you hold and when you need the answer. We reply within one business day.

Contact us