The process works, but only because people keep re-keying it

Slow, inconsistent, and it only scales by hiring. Somewhere in the middle, a spreadsheet is load-bearing.

The problem

The process functions. It functions because people move information between systems by hand, catch each other’s mistakes, and remember the exceptions that were never written down.

The costs are quiet: it is slow, results vary by who ran it, and the only way to do more of it is to hire more people. Errors enter at the joins, where nobody owns the handoff.

Automating the documented process usually fails, because the documented process is not the real one.

How we approach it

Map what actually happens, including the undocumented steps people improvise
Find where errors genuinely enter — usually a handoff, rarely the step everyone blames
Automate the mechanical parts and leave judgement with people
Integrate with the systems already in use rather than replacing them wholesale
Instrument it, so the next bottleneck is visible instead of anecdotal