Is the spreadsheet you want to get rid of the problem, or the only thing holding the process together?

When a team decides it has outgrown Excel or Google Sheets, the instinct is to start shopping. Search “spreadsheet alternative” and you'll find a long list of tools that promise to replace it. Most of them can. The trouble is that nobody has worked out what “it” actually does yet.

“A messy spreadsheet is usually a very honest record of how the work really happens.”

The part I pay attention to

Every odd column has a story. There's a “notes” column that's really a status, and a hidden tab that fixes prices because the old price list was wrong. Colour-coding means “don't ship yet”. None of it is written down anywhere else. Rebuild the file in a new tool from the headers alone and you keep the data but lose the rules that made it work.

WHAT'S ACTUALLY HAPPENING
The file holds data and rules together
The rules live in colours, comments and habits
A straight copy keeps the data and drops the rules

How I'd read it

Sit with the person who maintains it for an hour and go column by column. I ask 4 questions of each one:

  1. Where does this come from: typed, pasted or calculated?
  2. Who reads it, and what do they do next?
  3. What happens when it's wrong?
  4. Would this column exist if 2 systems talked to each other?

Mark anything that's a rule rather than data. That list, usually a page rather than a binder, is your real requirements document. It also tells you whether you need a new system at all, or just 2 fixes to the one you have.

Copy the rules, not just the columns.

Where AI fits

The same goes for AI. An AI agent given the columns but not the rules will be confidently wrong. The rules you write down while reading the file are exactly what let AI do a job in it safely, with a person checking the exceptions.

Give AI its first job →

Kathy has read a lot of spreadsheets that only one person understood. If yours is one of them, she'd genuinely like to go through it with you. Show us the workaround →