Why the biggest process is usually the wrong pilot

The most expensive process in a company is very often the one you should not start with. It is expensive because it is big, and it is big because a lot of people spend a lot of judgment on it, which is exactly what does not automate.

So the shortlist is not ranked by money first.

Three tiers, with a rubric that has to discriminate

Every mapped process is scored high, medium or low on how automatable it actually is:

• High. Off-the-shelf connectors exist for every tool in the chain. The work is moving structured data between systems that already have interfaces for it. Shippable in weeks.
• Medium. Real work: custom integration, scripting, or scraping something never meant to be read by a machine. Bounded, but it is a project.
• Low. The core of the job needs human judgment, or human hands. Only the administration around it can be automated.

The rubric is printed on the report page rather than kept in a methodology document, because a score the reader cannot check is a score they have to take on faith.

A scoring pass where everything comes out medium is a failed pass, not a cautious one. If the tiers do not separate, they carry no information, and the ranking underneath them is the money column with extra steps.

The pilot shortlist

Within the high tier, the shortlist is drawn around the processes worth at least half of the best candidate. That threshold does two things: it keeps the first project small enough to finish, and it stops the shortlist filling up with high-feasibility work that is barely worth doing simply because it was easy to score.

The result is usually not the biggest number on the page. It is the biggest number that can be delivered without a discovery phase of its own.

When nothing scores high

An engagement where no process reaches the high tier is normal. Companies whose work is physical, or judgment-heavy, or running on software with no integration surface, will map plenty of expensive processes and none of them will be quick.

That is a real finding rather than a failure of the audit, and the report says so in words instead of showing a bare zero beside a large annual figure. The shortlist falls back to the best tier that actually contains work, and names the tier it used.

The alternative, quietly promoting a medium process into the high tier so the page looks better, means the first project overruns and the relationship opens with an apology.

Win one, then expand

The first automation for any client is deliberately conservative: one workflow, one clear measure of success, and only systems with real interfaces. Not because bigger projects cannot work, but because the second project is sold by the first one having worked, and nothing sells a second project like a first one that was boring to deliver.