Fixed fees / No seat licences / USD
01 / Audit
2 weeks
We map every repeat task in one team, time it, and hand you a ranked list of what to automate first.
Process map of one team
Hours and error cost per task
Ranked build list with ROI
Fee credited if you build with us
02 / Build
Most picked
One agent, built against your data, tested on last quarter's real cases, shipped in four to six weeks.
One workflow, end to end
Eval suite on your own history
Human review step where it matters
Runbook and handover session
03 / Run
Monthly
We watch the logs, retrain on misses, and keep the agent honest as your tools and rules change.
Weekly eval report
Model and cost tuning
Two-day fix window on misses
Cancel with 30 days notice
Q.01
Do we need clean data before we start?
No. The audit tells you which data is good enough today, and most first builds run on the mess you already have.
Q.02
Who owns the agent and the code?
You do. Prompts, evals, code and runbooks live in your accounts from day one.
Q.03
What does a model cost to run?
Usually less than one hour of the task it replaces, per week. We give you a monthly estimate before the build starts.
Q.04
What if the agent gets it wrong?
Every build ships with a review step and a confidence floor. Below it, a person decides, and the miss feeds the next eval run.