Ramp

Mercor’s accountant test changes the question from AI accuracy to the cost of accepted work

Frontier models beat a small human baseline on structured accounting tasks. The business question is what remains outside the benchmark.