The same three things, in every accounting team.
None of this is anyone's fault. It is what happens when the people who understand the process cannot build, and the people who can build are committed to the product roadmap for the next three quarters.
A spreadsheet nobody dares touch
Thousands of formulas, built up over years by people who have since left. It is slow, it is fragile, and one wrong paste puts a number in the board pack that should not be there.
Reports measured in days
Pull the raw extract, filter it, clean it, pivot it, paste it into the deck. Every month, by hand, by someone whose actual job was supposed to be analysis.
A ticket that never gets picked up
Accounting asks engineering for help. Engineering is building the product. The request sits in a backlog until someone escalates it high enough to embarrass everyone.
Stabilise first. Automate second. Never the other way round.
The fastest way to automate a broken process is to first stop it being broken. We take the work exactly as it runs today, get it boringly predictable, and only then start replacing the parts that never needed judgment. The candidates are not chosen from a slide — they come from the people running your process, who have done the same manual step forty times a month and know exactly which ones a machine should own.
A month of knowledge transfer with the people who do the work today. We take the process exactly as it runs. No redesign, no opinions, no automation.
We run it as it is until the output stops surprising anyone. You cannot automate a process that is still moving underneath you.
Every step written into an SOP, then sorted into two piles: the steps that need a human judgment, and the steps that only need doing.
We automate the second pile only. Extraction, cleaning, matching, pivots, uploads, report generation. The judgment stays exactly where it was.
The build runs alongside the manual process until the two agree. Then the manual one is retired, and we start again on the next process down the list.
What discovery usually finds.
Different in every company, but the shapes repeat. These are the processes where the work is real, the volume is high, and the judgment required is close to zero.
