Path A
You know AI should be doing some of the work. You don't know where to start. Discovery finds the workflow that hurts and the system that fixes it.
The engagement
Two weeks. One honest answer.
The Discovery Sprint is a fixed-fee, two-week engagement: $20,000, ending in a build-ready SOW and a go/no-go that you own. If AI isn't ready for your problem, we say so.
Book a discovery callStart where you are.
Path B
You already have a spec or an SOW. We pressure-test it, price it, and start faster.
Either door leads to the same delivery engine.
What $20,000 buys.
Week 1
We sit inside your operation. We read the workflows, the tickets, the spreadsheets, and we find the work AI should own.
Week 2
We design the system, scope the build, and price it.
- A build-ready SOW
- A go/no-go you own
- Everything we learned, documented
Fixed fee. No surprises. Walking away is a legitimate outcome.
Phase 1
The MVP is a production release.
Two-week sprints. You pick the backlog. Working software at the end of every sprint, not a demo at the end of a quarter.
Send
The AI acts on what it has earned.
Draft
The AI proposes; your team approves.
Hand off
A person owns it, with full context.
Phase 2
The system keeps earning.
We tune what runs, add what the numbers justify, and hand your team more of the controls each sprint.
The bar we hold ourselves to.
| Sprint goal achievement | ≥ 85% |
|---|---|
| Defect escape rate | < 5% |
| Documentation at handover | 100% |
| Stakeholder satisfaction | ≥ 8/10 |
Fair questions.
What if AI isn't ready for our use case?
Then the answer is no, and you'll know in two weeks for $20,000 instead of a year from now for much more. The go/no-go exists so you find out cheaply. We have told clients no.
Who owns what you build?
You do. The IP, the data, the prompts, the systems. If we part ways, everything keeps running and everything stays with you.
What does delivery cost after discovery?
It's scoped in your SOW and billed by sprint. Discovery exists so that number is grounded in your operation, not a brochure.
What about Quebec's Law 25?
Every workflow we ship is designed human-in-the-loop: automated decisions are disclosed, reviewable, and reversible. We host in Canada when compliance demands it.
Do you work in French?
Yes, natively. Customer-facing software used in Québec must work in French; everything we build is bilingual by default.
How fast is Phase 1?
First production release in sprints, not quarters. A full platform takes longer; we co-built Operto's over twelve months of sprints.
Do you train models on our data?
No. Your data stays yours, full stop.