Everyone called it an Excel error. So was the one that cost Mouchel its CEO. Blame the spreadsheet and you miss what actually failed.
TransAlta is an electricity company. Every month it placed bids in a New York power-market auction to buy transmission hedging contracts. One rule governed that auction, and it is the rule that matters here: the price you typed in was the price you committed to pay, and every bid you won locked you into buying that contract. So you bid high only on the few contracts you actually wanted, and low on everything else. The bids were sealed. There were no take-backs.
One month in 2003, the person preparing the bid file sorted the spreadsheet, and the rows slipped out of line. The high prices TransAlta meant for the handful of contracts it wanted got pinned instead to the contracts it did not want. The file looked finished. Nobody caught the slip. The bids went in.
When the sealed results came back, the finance team felt the floor tilt. TransAlta had won a stack of contracts it never intended to buy, at prices far above what they were worth. Here, winning was the catastrophe, because winning meant paying. The rules barred the company from pulling back a single bid. It had to honor every one.
So TransAlta paid, and TransAlta lost. The overpayment came to 24 million dollars, about 10 percent of the company's profit for the entire year. The chief executive stood in front of shareholders and called it what it was: "literally a cut-and-paste error in an Excel spreadsheet."
Sit with what came after the press release, because that is where the real cost lives. The board meeting where every other number the company produced was suddenly suspect. The analysts who quietly trimmed their estimates. The year spent clawing back a loss created in a single click. One misaligned row, and a company with billions in revenue spent twelve months explaining itself.
You might tell yourself that is a controls problem. That your team would have caught it. So here is the one that should keep you up at night, because the leader did everything right and it ended him anyway.
In October 2011, the chief executive of Mouchel, a UK services group with more than 600 million pounds in revenue, resigned. Not over a scandal. Over a spreadsheet. An outside firm of actuaries, credentialed professionals the company paid precisely so it would not have to second-guess them, had made an error in a spreadsheet that valued a pension fund. The mistake wiped millions off the company's profit. Within a day of the disclosure, the shares fell by a third. Within a week the company had lost most of its market value, its lenders had called in an outside review, and a business that had recently fielded takeover offers north of 170 million pounds was worth a fraction of that.
The chief executive did not make the error. He inherited it. He trusted a number from people whose entire job was to be trusted, and nothing stood between that number and the decisions built on top of it. Nine years in the chair ended in twenty-four hours.
Notice what neither of these is. Neither is a story about people who could not use Excel. TransAlta's team were professionals. Mouchel's actuaries were experts. The software did exactly what it was told, instantly and without complaint. That is the whole problem. The tool had no opinion. It could not tell a 24-million-dollar triumph from a 24-million-dollar catastrophe, because nothing in the system was built to know the difference.
That is the part worth staring at. The failure was never in the cell. It was upstream of the cell, in the empty space where an architecture should have been. A number was produced. A decision was built on it. And nobody, nowhere in the chain, had created the layer that asks whether the number deserved to be believed.
Every operation runs on numbers it has stopped questioning. The pricing sheet someone built years ago. The inventory forecast that drives every purchase order. The margin model the whole quarter rests on. The one a person who left two roles ago first wrote, that everyone now trusts because it has always been there. That is your Revenue Hemorrhage in its quietest form. Not theft. Not fraud. A slow bleed from decisions made on numbers no one is architected to verify. The intelligence your operation throws off every day, the Digital Exhaust of every transaction, vents into the air because nothing was built to catch it and check it.
And here is what should change your week. Every leader reading this is about to hand a far more powerful tool than a spreadsheet the same ungoverned authority TransAlta and Mouchel handed Excel. They will call it adopting AI. They will deploy it on top of the same empty space, point it at the same unquestioned numbers, and trust it for the same reason: it answers fast and it sounds certain. A spreadsheet that is wrong costs you 24 million dollars in a morning. A learning system that is wrong compounds the error every single day. And the competitor who architected his correctly does not wait for you to catch up. That is the Catch-Up Penalty: the late mover does not start from a smaller gap. He starts from zero.
That pattern, the one that runs from the first misaligned spreadsheet straight into the AI system you are about to deploy, is the subject of my new book, Death To Excel! The title names Excel. The book is about something far larger than Excel, and far more expensive. It is available now on Amazon.
The pattern in your operation is not a story. It has a dollar figure. The Strategic AI Intelligence Diagnostic™ exists to put that figure in front of you: what your current architecture is already costing you, in dollars, before you spend a cent on AI. No products pitched in the room. No obligation created. One number you will not be able to unsee.
If you want yours, reach me directly: [email protected].
Carl J. Peterlin Jr. is the only Strategic AI Intelligence Architect™ in the NEPA market, with 42 years of Fortune-10-to-startup pattern recognition across pharmaceuticals, defense, energy, and high-growth startups. He is the author of Your SMB AI Revenue Ratchet and Death To Excel!
Edition 8 of The Fourth Intelligence, first published June 30, 2026. The argument stands as written. Offers, pricing, and product names referenced in the original have since advanced; current ones are at carlpeterlin.com.