Every Golem engagement starts the same way: two weeks, a fixed fee, and a written answer to one question. Which of your workflows should an agent run, and what will it return?

Week one: watch the work

We interview the people who do the work, not just the people who manage it. We pull a sample of real cases and time them. We map every system the work touches, and we note where a mistake would cost money or trust.

Week two: size the prize

For each candidate workflow we estimate hours returned, cost per task after automation, risk, and build time. We build a small prototype against real data for the strongest candidate, so the numbers are measured, not guessed.

What you keep

A ranked list of workflows with numbers, the prototype, the first version of the eval set, and a fixed quote for the build. About a third of clients build the first agent in-house from the audit alone. That is fine with us; the audit is priced to be worth it on its own.