Nobody thought the process was slow. It still cost 129 man-days a year.

The request started in the store and ended at central HR administration, with two paper signatures along the way. The annex was sent on as a pdf, signed on paper, scanned back in, and sent by internal post for good measure. Whoever did not get a digital version back scanned it themselves.

On top of that, the same data lived in three places: in SAP, on a SharePoint site and in a spreadsheet updated after every request. Each place asked for its own action, and every action was a moment where something could sit still.

The bottleneck was not the people. It was a process built around a paper signature that had long since become possible digitally.

First, write out the process the way it actually ran, step by step, with the time per step and the waiting in between. That produced the numbers a board can decide on: how often the process runs a year, how many man-days go into it, and where exactly that time sits.

Then every step past the same questions, in order: can it go, can it merge with another, can it be placed at the source or run alongside another step instead of after it, and only last: can a machine take it over. So the paper and the double entry went first, leaving a smaller process to automate.

Only then did technology come in. A standardised online form at the start, a digital signature instead of pen and post, and a software robot taking over seven steps: watching the mailbox, reading new requests, storing the form, reading it out, creating the annex through SAP, starting the digital signature and filing the signed document in the personnel record.

The moment it became clear that central HR administration had no step at all in the new design. Not because the role disappeared, but because in the ordinary run there was nothing left to do: the robot did the work and administration only stepped in for exceptions.

That made the discussion concrete. No longer whether automation yields anything, but what you do with the room it frees up.

Across two thousand requests a year, processing time went from 129 to 63 man-days. That is 66 man-days, roughly 0.8 of a full-time role, no longer spent retyping and forwarding. The investment paid for itself in six to nine months.

The paper disappeared from the process entirely, and with it the scanning, the posting and the chasing of whatever had been left lying somewhere along the way.

The gain was not in working harder within the steps. It was in steps that are no longer there.