Most COOs are about to celebrate the wrong win. AI killed the prep behind the operating review. Prep was never the job.
Think about what your weekly and monthly reviews actually were. Two things wearing one name. The artifact: the deck, the bridge, the six-pager. And the mechanism: a forcing function that made each team stare at its own input metrics until someone built real intuition and owned their variance. AI dissolves the first across every function at once. It barely touches the second. If you run operations, that gap is now your problem.
The reporting tax is dead, and good riddance
Nobody should spend the weekend rebuilding the same narrative by hand. Pulling the data, building the deck, writing the story for why a number moved. That was never the value. It was the toll you paid to reach the value.
But the toil had a hidden benefit. Someone had to touch the numbers, and in touching them they learned the business. Take that away and something quiet breaks. Auto-draft every team’s narrative and you can run a clean operating cadence where no function actually looked. Plausible is not correct. The models are exceptional at plausible. This is the same trap that turns a clean process into a hollow one, which is why bad process, not process itself, kills culture.
The real risk is review theater
Name the failure mode out loud. Review theater, scaled to the whole org. A polished auto-narrative in every meeting, challenged by no one, in a company that outsourced not just the prep but the thinking about why. That is worse than the old grind. The grind at least guaranteed one human had eyes on the data.
Here is the version that should worry you. The auto-narrative tells you conversion softened and recommends a price test. It does not tell you the softening was one product going out of stock for nine days. That is the difference between a good operator and a dashboard, and AI makes the difference easy to skip.
What the operating review demands of the COO now
So the operating review does not die. It inverts. The job stops being assemble-and-present and becomes interrogate-and-decide. Less time admiring the bridge, more time pressure-testing the machine’s bridge down to ground truth. The scarce skill is the adversarial read. Show me the anecdote behind this driver. That is a correlation, where is the controllable input. Reproduce it before we spend on it.
Your job changes with it. You stop being the consumer of the cleanest deck and become the person who installs that adversarial read as a discipline. Change what gets rewarded in the room. Not the smoothest story, the operator who found where the auto-story was wrong. Installing that discipline is exactly the work of a fractional operating partner, and you can see what a week of it looks like in practice here.
Two things you own get bigger, not smaller. First, choosing which input metrics each team is accountable for, because AI will bridge whatever you point it at and will not tell you which lever actually moves the output. Second, protecting the muscle, making sure someone on every team still touches ground truth instead of forwarding the model’s version of it.
The operating review is now a judgment test at the org level. You find out fast which teams can still read their own business when the report is wrong.
