All perspectives
ArticleData Curation & Governance

Which Spreadsheet Would Stop the Business If Its Owner Left?

Find the workbooks the business cannot run without. Make their inputs, logic, owners, and backup process dependable before adding automation.

An employee reviewing business data on computer monitors.

Most operating leaders can answer that question without checking.

It might be the weekly cash forecast, the production schedule, the commission calculation, or the workbook used to prepare a customer quote. Everyone depends on it. One person knows how it works.

The problem is not that the business uses Excel. The problem is that a critical operating process has become dependent on a file whose logic, sources, and controls are not understood by the people accountable for the result.

Adding AI to that file does not remove the dependency. First, the business needs to make the process visible.

Start with the decision the workbook supports

Ask the business owner what happens if the file is unavailable or wrong. Does payroll wait? Does a customer receive the wrong price? Does leadership make a cash decision from an incomplete balance? Can production continue?

That establishes priority better than counting spreadsheets across the company. A simple workbook used for a high-consequence decision deserves attention before a complicated file nobody relies on.

Walk through a full operating cycle with the person who maintains it. Identify the source exports, manual entries, formula changes, adjustments, and copies sent to other people. Ask which steps depend on memory.

Pay particular attention to cells someone changes because “the system never gets that right.” The adjustment may contain valuable business knowledge, a recurring data defect, or both.

Make the information and logic traceable

Data Curation & Governance should establish the approved source for each input, the meaning of key fields, and the owner of the calculation. Document assumptions separately from observed facts.

Choose an authoritative version. Control who can change formulas and who can enter operating data. Maintain a practical change record and a recoverable copy. Decide how the output will be reconciled before it is used.

The business owner remains accountable for the result. IT should support access, recovery, and the environment in which the file operates. Neither responsibility should depend entirely on the person who originally built it.

Test continuity before replacing the application

Workflow Optimization means having a trained backup perform the process from the approved inputs and instructions. The owner should compare that output with the established result and investigate differences.

A successful test includes the exceptions. What happens when an export is late, a new customer code appears, or a formula range no longer covers all records? Who decides whether the report is fit to release?

Some workbooks will remain appropriate once these controls exist. Others may need to become a maintained application or move into an existing business system. Make that decision from the operating requirements, support burden, and cost, rather than assuming every spreadsheet needs replacement.

Use AI only where the calculation can be checked

AI can help draft documentation, explain candidate formula logic, or prepare a review checklist. Those outputs require validation against the actual workbook and the owner’s knowledge. Use approved data handling practices when supplying business information.

Do not allow generated formulas or scripts to overwrite a production workbook without review and testing. A plausible explanation of a formula does not establish that the calculation handles every business condition.

Over the next two weeks, have the finance or operations leader identify the three workbooks with the greatest operational consequence. Within 30 days, assign a primary owner and backup, document the inputs and controls, and complete one supervised backup run for the highest-priority file.

Measure recovery capability, reconciliation differences, and time spent preparing and correcting the output. The impact should be lower risk exposure, more dependable operational visibility, and less recurring rework. Require a defined cost and benefit before authorizing replacement or an AI layer.

The real test is straightforward: can the business produce a reliable result without depending on one person being available?

Start with a Business Systems Assessment

Related perspective: Your Best Employee Shouldn’t Be Your Only Operating Manual