Seven steps from one process to a supervised go-live
Fifteen days from signed scope to a first pilot agent. Six weeks to a department automation. Ten weeks for a full fleet. The clock starts at scope lock — and nothing is billed before it.
- 1
Free discovery call
About an hourYou walk us through one process. We tell you honestly whether AI is the right answer — and if it is not, you hear it in that hour rather than in month three.
From you: A named process owner and an hour.
- 2
Process map & blueprint
Days, not weeksWe map the process as it actually runs, using a sample of real work — a hundred tickets, orders or documents is usually enough — and write a solution blueprint: which agents, which workflows, which integrations.
From you: Read access to the systems involved and a sample of real work.
- 3
Schedule A scope lock
Signed before any buildScope, acceptance criteria, autonomy levels and the price go into Schedule A. Verbal scope is not scope — if it matters, it is written here, and then it binds us. 50% of the build fee is billed here.
From you: Sign-off from the decision-maker.
- 4
Build & integrate
3 · 6 · 10 weeks by packageWe build the agents and workflows inside your existing stack, with hard-coded guards, floors and approval thresholds. You see progress in your private workspace as it happens.
From you: Timely answers — delays on access move the timeline, not the fee.
- 5
Supervised pilot
30 · 60 · 90 daysThe automation runs on real work under supervision. We set a measured accuracy benchmark, sample 200 decisions a week against it, and any agent that falls below it is demoted to draft-only.
From you: Review of the exception queue during the pilot.
- 6
Acceptance
Deemed at 7 working daysWhen the acceptance criteria in Schedule A are met, you accept — or it is deemed accepted after seven working days. The second 50% is billed here, with handover documentation and admin training.
From you: A decision against the written criteria.
- 7
Run & improve
MonthlyOn a run plan we monitor, clear exceptions, retrain and report. Minimum term three months, then thirty days’ notice. No exit fee.
From you: Nothing extra — a monthly performance pack comes to you.
You stay in control of every agent
AI makes mistakes. These are the controls that make sure a mistake is caught, traced and fixed — instead of discovered by a customer.
Declared autonomy
Every agent is supervised or draft-only, and you choose which. High-risk actions never execute without a human.
Pause & hard stop
Pause any agent from the dashboard; its queue routes to the human review desk instead of being dropped. Spend has a separate hard stop.
Decision log
Every decision is logged with the inputs that produced it and kept for 24 months, so an error can be traced, not argued about.
Weekly QC sampling
200 decisions re-checked at random every week against the accuracy benchmark set in the pilot.
Four named owners
Named people on the account and a five-level escalation ladder — you always know who to call.
Paper first
NDA, MSA and an InfoSec baseline before we touch your systems. Ownership of data and logic is written down up front.
Send us one process. We’ll tell you what it costs to automate it.
Register in two minutes. A solutions lead comes back to you with a fit check and the package we would recommend — no obligation until Schedule A is signed.
