AI won't run your business. An operating rhythm will.

There's a quiet assumption in a lot of AI conversations: that if you buy the right tool, the business will start running itself. It won't.

AI is genuinely good at a narrow set of things: turning messy input into structured output, summarising, drafting, classifying, connecting. None of that runs a business. What runs a business is an operating rhythm — a repeatable way the work gets seen, decided on, and done, week after week.

I've written elsewhere about what "AI as an operating model" actually means — the loops AI rewires when it's built into how a company runs rather than bolted onto what a product does. This piece is about the other half of that equation: the rhythm those loops run on. Because a loop without a cadence is just a diagram.

What an operating rhythm actually is

An operating rhythm is the small set of recurring moments where a team looks at reality, makes decisions, and commits to work — plus the rules for what happens between them. It's not the meeting calendar. Most meeting calendars are what grows when a company doesn't have an operating rhythm: recurring sessions nobody can name the decision rights for, attended defensively, producing minutes instead of movement.

The test for a real rhythm is blunt: for each recurring moment, you can say what enters it, what it decides, and what would break if it didn't happen. If you can't answer all three, it's not rhythm — it's habit.

The three cadences I'd install in any product team

This is the rhythm I run my own work on and would install on day one anywhere. Three cadences, each with one job.

Weekly: the signal review

The job: look at reality before opinions form about it.

Once a week, the team sees the synthesised signal — what customers said, what usage did, what the market moved, what sales and support ran into — clustered and connected, not summarised into three comfortable bullet points. With AI doing the ingestion and synthesis (the heavy lifting it's actually good at), this costs the team thirty minutes, not a day of deck-building.

What it kills: the status meeting. When everyone has seen the same decision-ready picture, "going around the room" becomes what it always secretly was — a read-aloud of things that could have been read.

The rule between beats: signal flows in continuously and raw. Nobody curates it on the way in; the review is where humans interpret.

Fortnightly: the decision forum

The job: make the calls that need making, against the evidence, with the deciders in the room.

Every two weeks, the open decisions get framed — options, evidence attached, a recommendation if someone has one — and someone with decision rights decides. Not discusses. Decides. The framing is work AI accelerates enormously; the deciding is the part that stays stubbornly, properly human.

What it kills: the alignment meeting, and its expensive cousin, the decision that drifts for six weeks because it was nobody's job to force it. Most "slow" product teams aren't slow at building — they're slow at deciding, and they hide it under the word alignment.

The rule between beats: decisions get written down — what was decided, why, and what would change our mind. That record is what makes the rhythm compound: next quarter's questions get answered against this quarter's reasoning instead of everyone's memory of it.

Monthly: the bet review

The job: look up from the work and ask whether the bets are still the right bets.

Once a month, the team reviews what it's betting on — not task progress, but the assumptions underneath: is this still the problem, is the signal still pointing here, has anything we've learned made a bet bigger, smaller, or dead? It's the moment the rhythm connects the week-to-week to the strategy, and the only cadence where killing something is a normal outcome rather than a crisis.

What it kills: the quarterly roadmap theatre — the ritual where three months of drift gets retrofitted into a narrative. When bets are reviewed monthly against continuously synthesised signal, the quarterly conversation becomes a summary of decisions already made, which is what it should have been all along.

Where AI sits in all of this

Notice the pattern across the three cadences: AI compresses everything around the decision — the gathering, clustering, framing, and recording — and humans occupy the decision itself. That's the design principle. The rhythm gets faster and richer because the expensive preparation work approaches free; it stays trustworthy because judgment never left the room. A rhythm where AI also makes the calls isn't an operating rhythm, it's an outage waiting for a root-cause review.

Why the rhythm matters more than the tools

Two teams can own identical AI tooling and get opposite outcomes, and the rhythm is usually the difference. Without it, AI output piles up as more unread material — summaries nobody asked for, drafts nobody reviews, dashboards nobody opens. The tools work; nothing changes. With a rhythm, every piece of synthesis has a destination: this lands in Monday's signal review, that feeds the forum on Thursday week, this updates a bet. The cadence is what turns capability into consequence.

That's also why the honest starting question isn't "which tools?" It's the less comfortable one: how does the week actually work today — and what would change if the boring stuff stopped eating it?

How to install it (without a reorg)

Start with the weekly signal review only. Get the raw feeds flowing, let AI do the synthesis, and hold the thirty-minute review for four weeks. It will feel redundant with existing meetings — that's the point. In week five, kill the meetings it made redundant. Add the decision forum once people trust the picture; add the bet review once decisions are being written down. Rhythm installs one beat at a time, and each beat has to earn the next.

---

Field notes, not theory: this is the rhythm I run, and the version above is the survivor of several worse ones.

Wrestling with your own team's rhythm? Tell me what you're building — or start with what AI as an operating model actually means.