01What you receive
Three documents at the end of the week. Not a slide deck, not a proposal dressed as an assessment — documents your own engineers can evaluate and act on.
Architecture document
How the agent would be built: model selection, tool design, where it runs, where your data sits, and what stays with a human.
Integration feasibility report
System by system: what can actually be automated, how brittle each integration is likely to be, and what will break first.
Fixed-price build quote
A firm number for the build, with scope defined tightly enough that it does not move later.
02What happens during the week
01Map the workflow
We sit with whoever does the task today and record exactly what they do — every login, every copy-paste, every judgement call and every point where it goes wrong.
02Establish feasibility
Where a system has no API, we capture and analyse its real network traffic to determine what can genuinely be automated, how brittle it is likely to be, and what will break first.
03Design the architecture
Model selection, tool design, where the agent runs, where the data sits, what is automated and what stays with a human — written down, not described on a call.
04Price the build
A fixed-price quote for the build itself, with scope defined tightly enough that the number does not move.
03What we need from you
- An hour with whoever does the task today — the person doing it knows where it actually breaks
- A screen recording or screenshots of the workflow as it runs now
- Read-only or sandbox access where possible — useful, but not essential; we can complete a discovery from description and recordings alone
- Written authorisation for any system we are asked to access. We never touch a production system without it
04The honest part
Sometimes the right conclusion is don't build this. A workflow that runs twice a month, or a system due for replacement next year, or a process that would be cheaper to fix than to automate — we will say so, in writing, and explain why.
That outcome is a legitimate result of discovery, not a failure of it. You will have spent $750 to avoid spending far more on a build that would not have paid for itself, and you keep the analysis either way.
05Common questions
Why is discovery paid rather than free?
What if the answer is that we should not build an agent?
Is the fee really credited against the build?
What access do you need?
How long does it take to start?
Start with the system that hurts most
Tell us the workflow being done by hand today and the system that blocks it. We reply within one working day to confirm whether discovery is worth running at all — and if it is not, we will tell you that for free.
Start discovery →