Score any AI vendor in 3 minutes.

Ten questions that expose whether a vendor will ship or stall — drawn from how healthcare, legal and university buyers get burned. Answer honestly, get a grade, and unlock the exact contract clauses to demand before you sign. It works on us too.

Grade the vendor you’re talking to.

01 · EvaluationDo they build an evaluation set from your real historical cases before writing agent code?
02 · AccountabilityCan they show you, in writing, who can override the agent and when a human must review its output?
03 · ComplianceWill they sign your BAA / DPA, and do their cloud subprocessors hold current SOC 2 Type 2?
04 · MilestoneDo they commit to a dated milestone where the agent runs alongside your team on live work?
05 · Outcome riskIs there a defined remedy if the agreed metric is not hit — refund, free remediation period, or fee holdback?
06 · Audit trailDoes every agent action record its input, the rule applied, the output, and who reviewed it?
07 · Change controlDo they have a written policy for how the model gets updated — and how updates are tested before they reach you?
08 · ExitIf you leave, do you keep the repository, the evaluation suite, the prompts, and your data — in your own cloud?
09 · ReferencesCan they name a live deployment you can actually try or a reference you can actually call?
10 · PricingAre their prices published or fixed in writing before kickoff — no open-ended time and materials?
0 / 10 answered