A pilot is paid, measured and bounded. It tests two things at once: whether the workers do the work correctly, and whether delivering it is economic for both sides. A demonstration proves neither.
Maps the process and the rules; connects read access; processes an agreed sample of past cases, the hard ones included, a share held back unseen; you compare the outcomes against your reviewed results.
Whether we understand the work, including the exceptions.
Accuracy by severity meets the agreed thresholds on the held-back cases, not only on the cases used to tune.
Processes current cases in parallel with your team, without writing to production; outcomes compared daily.
Whether performance holds on unfamiliar work at live volume and pace.
Thresholds met for three consecutive weeks; cycle time at or better than baseline; intervention hours within budget.
Prepares real transactions for approval through the decisions inbox; runs the exception queue; writes within delegated authority after each approval.
Whether the service saves effort in practice, with real counterparties.
Total human effort both sides below baseline by the agreed margin; zero high-severity errors; incidents handled by the documented path.
Takes responsibility for the agreed workload and service levels, with authority delegated in advance for routine, in-policy work.
Whether you value it enough to renew and expand.
A signed recurring agreement. Supervised production was the transition, not the service.
Set per task and per error severity, with you, before stage one. The definitions are fixed; the rates are agreed per engagement.
An automation figure can hide a great deal of manual work. Ours cannot: each minute a person spends on a case is recorded against one of six causes and reported to you every week.
Cases reaching the agreed verified outcome without Humetry specialist intervention. Your approvals of reserved decisions are reported separately and never count against it.
Humetry and client time reviewing, investigating, correcting, approving and supervising, classified: implementation, ordinary approval, missing external information, new judgment, system failure, worker capability failure.
Correct outcomes and errors by severity; unresolved and unverified work explicitly identified.
End-to-end case duration, with external waiting shown separately and never netted out.
Our direct cost per case against the fee; onboarding effort; the share of components reused from the previous deployment.
Every desk a hill whose height is its autonomous completion, drawn as contours by your own browser from the signed evidence pack; the agreed threshold is the shoreline. Two benchmarks stand on open ground and light only when your browser has verified each signature.
Four desks above the shoreline, one on it, one under: deductions and disputes at 93.2 against 95, with its cause classified in the week's report.
A pilot is never free. The pilot fee covers discovery, process mapping, integration and the three evaluation stages, and is credited against the deployment fee on conversion.
Tell us where the work sits today, the systems it runs in and the volumes. We come back with a scope, a pilot proposal and a date. We reply from a named person.