Written by: Luke Sobieraj, Founder & COO, Digital Colliers
You know the one. It started as a quick solution three years back. Now it has seventeen tabs, conditional formatting that nobody remembers how to edit, and a comment thread in row 247 that reads "DO NOT DELETE THIS COLUMN."
When Sarah is out sick, the monthly close waits. When the file size hits the limit, someone exports older data to a second spreadsheet. The audit team keeps a printed copy of the formulas because they're not sure the digital version will survive the next Office update.
This is how systems of record happen in real companies. Not by design. By accretion.
Signs it's gone from tool to infrastructure
You'll know a spreadsheet has become business-critical when more than one department needs it to do their job. When your planning cycle can't start until someone runs the numbers. When version control means emailing files back and forth with datestamps in the filename.
The pattern I keep seeing: one person owns it. They built it, they maintain it, they're the only one who fully understands how the calculation works. Everyone else just consumes the outputs. That's a single point of failure dressed up as a productivity win.
Other signals: it feeds into board reports. It's referenced in compliance documentation. Someone has built downstream processes that assume it exists and updates on schedule. People ask "can we add this to the spreadsheet" rather than questioning whether the spreadsheet should still be the answer.
The cost of leaving it alone
Data quality degrades slowly, then suddenly. Small errors compound. Rounding differences accumulate. Someone copies a formula down wrong and nobody catches it for six weeks. By the time you spot the issue, you're debugging three months of downstream decisions.
Then there's the bus factor. If the maintainer leaves, retires, or just gets tired of being the spreadsheet person, you're stuck reverse-engineering intent from cell references. I've watched teams spend weeks reconstructing business logic that used to live in one person's head.
Compliance exposure grows too. If that spreadsheet feeds into financial reporting or regulatory submissions, you're carrying risk you probably haven't quantified. If it touches personal data and you're in Europe, GDPR fines reach up to €20M or 4% of global turnover. You're one audit away from explaining why your system of record is a .xlsx file on someone's laptop.
The rush-to-AI trap most companies fall into
Here's where it gets expensive. Companies ignore the problem for years, then panic and grab at solutions. Right now that usually means an AI pilot.
The numbers are brutal. 88% of AI proof-of-concepts never reach production. For every 33 AI POCs a company launches, only four graduate to actual use. Another study found 95% of enterprise GenAI pilots deliver zero measurable P&L impact.
The failure mode isn't the technology. It's that teams skip the boring work. They don't map the actual workflow. They don't interview the people who use the spreadsheet every day. They don't document what the system needs to do before they start building.
So they end up with a solution that doesn't fit the problem. Or they ship something that works in demo but breaks with real data. Or they build features that sound clever but nobody actually needs.
How to decide if building is worth it
Not every spreadsheet needs to become a system. Some should stay spreadsheets. The question is whether the cost of maintaining it as-is exceeds the cost of building and maintaining something proper.
Start with usage. If fewer than five people touch it regularly, probably leave it alone. If it updates monthly and the workflow is stable, a spreadsheet might be the right tool forever.
But if you're seeing daily updates, multiple editors, downstream systems that depend on it, or meaningful risk if it breaks, then you're past the point where a spreadsheet makes sense.
When you do decide to build, boring wins. Map the current workflow first. Document every calculation. Talk to everyone who touches the data. Then build the simplest thing that could possibly work. A basic CRUD app with proper validation beats a sophisticated AI-powered solution that takes two years and never ships.
The real system of record isn't the one with the cleverest technology. It's the one people actually use, that doesn't break, and that you can hand off to someone new without a three-week training programme.