Excel is a gift and a curse

No software has done more for small and mid-sized companies than the spreadsheet. It needs no IT department, no budget line and no training beyond a colleague leaning over your shoulder. A salesperson can model a discount in five minutes. A founder can build a cash forecast on a Sunday. An accountant can reconcile a bank statement, an operations lead can plan a week of deliveries, a buyer can compare three supplier lists side by side. All in the same tool, all before lunch.

That versatility is real and we do not pretend otherwise. When a business is figuring out what it is, the spreadsheet is the right instrument: cheap to change, impossible to break in a way you cannot undo, visible to everyone. Most of the systems we run started life as one, and the sheet is where the rules of the business were first written down, even if nobody called them rules.

The trouble starts the day the spreadsheet stops being a calculation and becomes the record. The place where the truth about customers, orders, prices and payments lives. A spreadsheet is a terrible database, and it is terrible in ways that stay invisible until they cost money.

• It accepts anything. A price typed in the wrong column, a customer spelled three ways, a date that is actually text. Nothing refuses the row. A database would.
• It has no memory. Overwrite a cell and the previous value is gone. Nobody knows what the price was last Tuesday, who changed it or why.
• It forks. "Final_v3_REAL.xlsx" exists in every company. Two people edit two copies and both are right, until the numbers are compared in front of a customer.
• It does not connect. The sheet does not know the invoice was paid, the order shipped or the supplier changed the list. A person has to carry that news across, by hand, every day.
• It hides the logic. The formula that computes margin lives in cell K14, written by someone who left two years ago. Nobody dares touch it; nobody can test it.
• It does not scale in people. One person can run a business from a spreadsheet. Three cannot, and the third hire spends half their time reconciling the other two.

Ask three questions about the sheet. Does more than one person write to it? Would a wrong cell cost you money or a customer? Does anything downstream depend on it: invoices, dashboards, another sheet? One yes means it is a calculation that grew up. Two or three mean it is your database, and it is failing quietly.

The symptoms are familiar: a weekly hour spent "cleaning" the data, a person who is the only one who understands the file, reports that are right the second time, and an unwillingness to change anything because something else might break.

Keep the sheet

As a calculator, a scratchpad and a view. People are fast in it and that is worth protecting. Let it read from the record instead of being the record; the numbers show up already checked.

Move the record

Customers, products, orders, prices, payments and accounts go into a real database with the constraints turned on. Postgres is free and will outlive every tool you plug into it.

Move the rules

The formulas in K14 become code with tests and a version history. When the margin rule changes, it changes once, and you can see who changed it.

Fix the entry

The reason companies drift back to spreadsheets is that the database made typing slower. AI data entry closes that gap: the sheet, the PDF or the chat message is read, a person confirms a preview, the row is written clean.

We are not against Excel. We are against Excel as the single source of truth for a business that has outgrown one person's memory. The move is not "replace the spreadsheet"; nobody wins that fight and nobody should. The move is to take the record out of it, leave the calculation in it, and let the two talk. Done well, the team notices only that the numbers stopped arguing with each other.