Digital Transformation 28 June 2026

Retiring the spreadsheet: a migration playbook

Somewhere in your company there is a spreadsheet nobody is allowed to touch. It has 34 columns, six colors and one person who understands it. That spreadsheet is your real ERP.

Somewhere in your company there is a spreadsheet nobody is allowed to touch. It has 34 columns, six colors, and exactly one person who understands it. That spreadsheet is your real ERP — and the day that person goes on vacation, so does your operation.

The first step is respect, not replacement. That sheet encodes years of decisions: every column is a requirement, every color a workflow state, every "don't touch this cell" a business rule nobody ever wrote down. We map all of it before we build anything — and this is where AI earns its keep: we use Claude to extract and document the hidden logic from legacy sheets, which is faster than interviews and surprisingly good at spotting the rules people forgot they made.

Then: migrate in stages, never big bang. Phase one, the new system runs beside the spreadsheet and mirrors it — zero risk, everyone keeps working. Phase two, one process — say, scheduling or invoicing — moves over completely and the corresponding columns are frozen. Phase three, the sheet becomes a read-only archive. At no point does the business stop.

The human part decides everything: the person who owns the spreadsheet becomes either the new system's administrator or its saboteur. Involve them from day one — their expertise is the specification. When they demo the new system to colleagues themselves, the migration is already over.

The goal was never to kill a spreadsheet. It is to free the data and remove the single-person risk. Done right, the moment of success is quiet: three months in, someone asks how a report was ever assembled by hand — and nobody can quite remember.

← View all insights