Spreadsheet sprawl is a symptom, not the disease

Somewhere in your business there is a spreadsheet that should not exist, that everyone depends on, and that only one person fully understands. There is probably a folder of them. When a new operations leader finds this, the first instinct is usually to clamp down. Ban the spreadsheets. Force everything into the proper system. It feels like restoring order. It almost always makes things worse, because it misreads what the spreadsheets are.

Spreadsheet sprawl is not the disease. It is the symptom, and a remarkably honest one. Every one of those files is a record of a place where the official system failed someone, and a capable person quietly built their own fix rather than waiting for permission. The dispatcher who tracks jobs in a personal sheet is not being difficult. They are telling you, in the clearest possible way, that the real system does not let them see what they need to see. The spreadsheet is the workaround for a gap, and the gap is the actual problem.

This matters enormously the moment you start thinking about AI, because sprawl is exactly the kind of mess a model cannot reason about. The truth is split across a dozen files with different conventions, half of them out of date, the logic living in formulas nobody documented. A model fed that produces confident nonsense. So the temptation is to treat the spreadsheets as the obstacle to clear before the AI can go in. But if you just delete them without understanding why they exist, you destroy knowledge and break the workarounds people were relying on, and the sprawl quietly grows back somewhere you cannot see it.

The better move is to read the sprawl like a map. Each spreadsheet marks a need the system is not meeting. Taken together they are a free, detailed specification of what your operation actually requires, written by the people who do the work, paid for in their own unpaid time. When we run a Discovery Sprint, those files are some of the most valuable things a client can show us. They tell us where the real system fell short and what good would look like, far more reliably than any formal requirements document, because nobody builds a workaround for a problem they do not have.

So the cure is not enforcement. It is to give people a system that does what their spreadsheets do, only better, in one place, with the truth shared rather than copied. When that lands, the sprawl does not need banning. It evaporates, because the reason for it is gone. People do not cling to spreadsheets out of affection. They cling to them because they work.

And only then, with the truth consolidated and the gaps filled, does AI have something to stand on. The model can finally see one coherent picture instead of a dozen contradictory fragments. The lesson holds more widely than spreadsheets. Before you remove a workaround, understand the problem it was solving. The mess is usually pointing straight at the thing worth fixing.