The method
Analysis before automation. Always in that order.
The failure mode of AI projects isn't the technology. It's automating a process nobody understood, then discovering the savings were imaginary. So we measure four layers of your operation first, and we write down what we find before anyone touches a tool.
- Duration
- 4 weeks
- Your team's time
- ~8 hrs total
- Fee
- Fixed, quoted up front
- Commitment after
- None
Four layers, measured in that order.
Each layer is only meaningful once the one above it is written down. You can't judge a workflow without knowing which vertical it serves, and you can't judge an application without knowing which workflows depend on it. The order is the method.
- L1
Verticals
Which lines of business you actually run, who owns each, and where margin genuinely comes from.
- Revenue and headcount split by line, from your own records
- Who is accountable for each, and who people actually ask
- Which lines subsidise which — usually a surprise to someone
- L2
Workflows
Every recurring process end to end, with the number of touches and the wait between them counted.
- Traced by walking the work, not by reading the procedure binder
- Handoffs, approvals and the queue time between each step
- The undocumented workaround everyone uses gets written down too
- L3
Application stack
What you pay for, what's actually used, where data lives, and which integrations are really a person retyping.
- Licence count against real active users
- Which system holds the authoritative version of each record
- Every "integration" that turns out to be a CSV and a Tuesday
- L4
Friction points
Where work stalls, is re-entered, or lives in one person's memory — each costed in hours per week.
- Hours per week attached to each, from observation not estimate
- Single points of failure — the person who can't take a holiday
- This layer becomes the build plan
The opportunity register, ranked
Every friction point becomes a line item with a value estimate and an effort estimate. Then they get sorted. The ranking is the deliverable — not the list.
- A.High value, low effort — built inside the first 60 days
- B.High value, real effort — scoped, staged, and sequenced against IRAP
- C.Fix the process first — automating it would only make a bad process faster
We will tell you when the answer is "don't automate this." It happens on most engagements, and it's usually the cheapest recommendation in the document.
The method has a name and a price.
It's called the BusinessOS assessment — four weeks, fixed fee, and the document is yours either way.