Workflow Audit
Bring one workflow. Leave with a map you keep.
A single operational process — the one with the exceptions, the delays, and the approvals nobody can fully explain. We map how it actually runs, measure how it performs today, draw the redesign, and finish on a single word: go, reshape, or stop.
What it is
A scoped engagement on a single workflow. Not a review of your AI readiness, not a survey of your whole operation — one process, followed end to end, from the trigger that starts it to the record it leaves behind.
It hands you a document set you own. It says whether the workflow is worth automating, which parts a model should handle, which parts belong in deterministic code, and which parts stay with a person who can be held to the call.
If the answer is that it should not be built, the audit says so in those words.
Who it’s for
Bring a workflow with a name. The invoice run. The intake queue. The renewal review. The pile of exceptions somebody clears every Friday because the system won’t.
It fits when:
- the process repeats, and someone can describe how it is supposed to go
- it is document-heavy, exception-heavy, or gated by an approval
- somebody owns it and can authorize a change to it
- the systems it touches are systems you can grant access to
No industry qualifier. No size cutoff. The shape of the work decides, not the sector you sit in.
What you receive
- Your Workflow Map.
- The current state, drawn from how the work actually moves rather than how the process document says it moves — systems, handoffs, decision points, and the places where it stalls.
- A measured baseline.
- How the workflow performs today, written down while it is still true. Later, a change has a record to argue with instead of a memory.
- An automation and control design.
- It splits the work — what a model handles, what deterministic code handles, what a person decides — and puts every approval point where somebody has to stand behind it.
- An implementation brief.
- Acceptance criteria, dependencies, and the stop conditions that would end a build before it wastes your money.
- A direct recommendation.
- Go, reshape, or stop.
You keep all five whether or not we build anything. Take the map to another firm, or build it in-house — it still describes your workflow.
Deliverable
The map is yours either way.
Plate I
Recommendation
Go. Reshape. Stop.
Plate II
Plate III
Illustrative layout. No engagement data and no measured values appear in this figure.
The engagement
A single workflow, followed from trigger to record.
Map one workflow
Plate IV
How it runs
-
Written intake.
You describe the workflow in writing before anyone books a call. Written first, so the call is about your process instead of about qualifying you.
-
Access and observation.
We watch the workflow run in the systems it lives in and read what it produces, including the exceptions nobody logged.
-
Design.
We draw the split between model, code, and person, and we set each approval point on somebody’s desk.
-
Brief and recommendation.
The document set arrives whole, and it closes on the call: go, reshape, or stop.
You see the schedule and agree to it before anyone starts.
What it asks of you
Short list, and none of it open-ended.
- Access to the systems the workflow runs in.
- Records — real examples, and the ugly ones especially. The exceptions are the workflow.
- Time from your people — whoever owns the process, and whoever actually does it. That one we cannot substitute.
You hear the whole ask before you commit to any of it. Nothing starts until you arrange the access.
What follows a go
A go points at one of these shapes. None of them starts without a map — ours, or one you already hold.
- Workflow Build.
- We build the design inside the systems you already run, approval points and all.
- Workflow Operation.
- Somebody keeps it running: monitoring, the exception queue, and the measurement cadence that keeps the baseline current.
- Controlled-Data Build.
- The same build, scoped for a workflow where the records are the thing that cannot move.
Map, build, operate, measure. Each pass leaves an accepted, versioned baseline that the next decision stands on — and that a later observation can revise.
Questions we get before the first call
What does it cost?
We quote case by case, against the workflow in front of us. This page will not carry a figure, now or later. We publish what sets the quote, in full, on How scope is set — read it first and you’ll already know what we’re going to ask about.
How long does it take?
You agree to the schedule before anyone starts. We don’t publish a duration, because a duration published by somebody who hasn’t seen your workflow is a guess wearing a number.
Can we skip the audit and start building?
Only if you bring an equivalent map of your own. Building against a process nobody measured is how automation ends up automating the wrong step, faithfully.
What happens if the recommendation is stop?
Then it’s stop, and you keep everything. A recommendation you didn’t want is worth more than a build you’d have regretted.
When can you start?
One operator, and engagements run in sequence. Intake is currently open. If it’s closed you get told that, instead of a slot that doesn’t exist.
Do we have to run your tooling?
No. A build is designed to land in the systems you already run. If you never engage us past the audit, the map still describes your workflow rather than our software.
A stop is a result. You still keep the map.
Map one workflow
Name the workflow, say where it breaks, and tell us who owns it. Written intake first — no call booked to arrange another call. We read it, and you hear yes or no on fit.