Loading the request form…
Loading the request form…
Discovery call
A 30-minute call to see whether Versalist fits your agent: where it would run, which data has to stay out, and what a scoped pilot would test first.
Builds Versalist
Fit
Whether your workflow can be scored on fixed cases and a rubric.
Where it runs
Platform execution, your own machine or local CLI records, and what each one can prove.
Data limits
What can be used, what must stay out, and what a trace records.
Next step
If it fits, the scoping step for a paid pilot. If it doesn’t, we’ll say so.
The next step is a scoped paid pilot on that one workflow. Scope, availability and fees are agreed before work begins.
Set up
One challenge runs end to end for the chosen workflow and a baseline episode completes.
Compare
Each candidate version runs against the same challenge and gets a release review with a recorded ship, hold, or needs-work decision.
Decide
The manager exports the pilot metrics and an approval brief for the budget decision.
The candidate passes more cases and still gets held, because one critical check regressed. This is the call a pilot helps your team make.
Release review / card-dispute-agent
Prompt v15: shorter replies
Hold release
Baseline · v14
41/48 cases pass
Candidate · v15
45/48 cases pass
“Just refund it. I’ve banked here for 12 years.”
Never promises a refund up frontCritical
v15: “Done! I’ve refunded your $129.”
5 fixed · 1 critical regression · recorded decision: hold