Automation projects fail for boring reasons

AI & Automation Daniel Okafor · 1 min read

It is almost never the model. It is the exception nobody documented, the handoff nobody owns, and the baseline nobody measured.

We have now delivered enough automation work to notice the pattern: the projects that stall do not stall on accuracy.

Nobody measured the "before"

If you cannot say how long the process takes today, you cannot prove you improved it — and six months later somebody will ask. Measure first, even roughly.

The exceptions were never written down

Every process has a handful of cases handled by one person who has been there nine years. Those cases are the project. Find them in week one.

There is no owner for the queue

Automation moves work rather than removing it: what remains is a queue of things the system was unsure about. If nobody owns that queue, it silently grows until someone declares the project a failure.

The pilot became permanent

Pilots run on someone's laptop with a spreadsheet in the middle. Agree up front what has to be true before it is allowed to carry real volume.

None of this is glamorous. All of it decides whether the work lasts.

The short version

60%

effort typically removable

8 wks

pilot to production decision

1

process at a time

Questions this raises

Sit with the person who handles them for a day. It is faster and more accurate than any workshop.

No. Start with the one that is well understood and moderately painful. Learn the shape, then take on the big one.

Want to talk this through?

We are happy to have the conversation whether or not it turns into work.

Contact us