The structural problem with every automation audit
The party running the diagnostic is the party who gets paid to build whatever the diagnostic finds.
This is not a scandal, it is just the shape of the industry, and everybody in it knows the shape. It is also why a free audit is priced the way it is: it is a sales meeting with a deliverable attached, and its conclusion is known before it starts.
The only correction that actually works is a stated condition under which the answer is no, decided in advance and checkable afterwards.
The rule
A workflow gets built only after a written ROI estimate, signed by the operator, clears positive at conservative inputs.
If it does not clear, the reasons are written up and the engagement ends.
That is the entire rule. Its value is not in its sophistication. It is in being said out loud before the numbers are known.
What conservative means, specifically
Three choices, all made before anyone sees the answer.
- Volume from a low month. A return that only works at peak throughput is a return that vanishes in January.
- Labour at loaded cost. What the hour costs the business to provide, not what appears on the payslip.
- Build cost at the top of its range, including the tuning that always follows go-live rather than the figure that would hold if nothing went wrong.
None of these is clever. Together they are the difference between an estimate that survives the finance function and one that was written to be approved.
What the client gets when the answer is no
Everything produced so far: the map, the signed baseline, the priority matrix, and the written reasoning.
That last document is the one operators actually use. It names the candidates considered, what each would have cost, and what would have to change for the answer to become yes - a volume threshold crossed, a data-quality problem fixed, a system replaced first. It is frequently used internally to stop a project somebody else was pushing.
Which is a better outcome than a build that never pays for itself, at a small fraction of the price. The reasoning behind pricing an audit this way is in how to read an automation quote.
What breaks without the rule
Every audit returns yes.
Not because anyone lied, but because a process with no stopping condition does not stop. The candidates get generous volume assumptions. The build estimate is the good case. The payback lands just inside the acceptable window. And the client, who cannot check any of it, approves.
Then the automation ships, works, and is abandoned in month eight when its sponsor changes role - and the conclusion the client draws is not that this project was mis-scoped. It is that automation does not work for companies like theirs. The pattern is common enough to have its own literature, some of it in why your last chatbot failed.
Where it sits
The refusal rule belongs to Cut, the fourth of six stages, and it depends on everything before it being real: a measured baseline, an honest map, a scope that stopped moving. The other half of Cut - what gets deleted and in what order the rest ships - is in choosing what ships first. The arithmetic itself is in ROI math for automation projects.
Top comments (0)