The operating model for the AI era

In Why are you doing that? I argued that most companies are running a modern business on systems built for a world that's gone; status meetings, monthly reports, scorecards, cascades and quarterly resets, all designed before Slack, before Zoom and long before AI. A lot of people came back to me with the obvious follow-up, which is completely fair: fine, so what would you do instead?

This is my answer. It's the operating model we've built Clearahead around, and I want to set it out plainly here, because I think the model matters more than any single feature in the product.

An operating model sounds grand, but really it's just a company's answer to three questions that every company answers, whether it means to or not: what are we trying to do, how do we decide what to work on, and how do we know how it's going? The old answers all assume that information is expensive to gather and slow to move, which is exactly the assumption that has changed.

So here's the short version. The operating model for the AI era is one loop that the whole company runs together: set the goal, find what's holding it back, move on it, and when it clears, find the next thing. The machine does the asking, the gathering and the admin; the people do the thinking and the deciding.

It keeps the ideas that have always worked, like having one clear goal, Eliyahu Goldratt's insight that a system is held back by its constraints, and Lean's war on waste; and it drops the rituals that only ever existed because information used to be hard to get.

Seven principles it runs on

Everything else in the model follows from these, so if you only take one part of this piece into your next leadership meeting, take this list.

1. One goal at a time, per team. A goal is an outcome with a rough when, not a list of eleven priorities. A company can work towards several goals at once, but each one gets its own space and its own loop, so no goal quietly hides inside another.
2. Find the cause, not the symptom. Most department to-do lists are symptoms of one or two underlying causes, so the first job is to ask: what's holding us back? Symptoms are evidence; constraints are the causes the team can actually do something about. More on this in Remove constraints, don't add initiatives.
3. Work on what needs clearing now. The team puts its constraints in order, and that order is the priority. We recommend three to five active constraints at a time, though you can add as many as you need. This isn't a six to twelve month plan; the effort goes on what's in the way right now.
4. Every move earns its place. A move is an action with one owner and a due date, and it eases a named constraint. Work that doesn't ease a constraint is still allowed, but it's labelled as such, so everyone can see what's really moving the goal.
5. The machine asks, people decide. AJ gathers the updates, writes the weekly read and spots what looks done, but it never makes the decision. AJ's view is information, not a gate; if the team wants something AJ doesn't think fits, it happens, and AJ says once what it would keep an eye on. I go further into this in The machine asks, people decide.
6. Updates are a sentence, not a report. Anyone records progress in their own words when it happens. Where things stand is visible to everyone all the time, so nobody has to prepare it, present it or chase it.
7. It's a loop, not a plan. When a constraint clears, you record what cleared it and find the next one; when the goal is reached, you celebrate it, keep the story, and set the next goal. Nothing waits for an away day. Why this beats the annual plan: A loop, not a plan.

The parts

There are really only three things in the model: a goal, the constraints holding it back, and the moves that ease each one. Everything else is there to keep those three clear and up to date.

Every item answers the same five questions

Goals, constraints and moves all carry the same five fields, so anyone in the company can open any item and understand it without needing a meeting to explain it:

• Short description: what it is.
• Why it matters: what it unlocks, or how it holds the goal back.
• Why we believe it: the signals behind it, with how sure we are in words (a hunch, fairly sure, sure), never a score.
• How we'll know it's done: what would be true, or what we'd stop seeing.
• Current reality: where we stand today, written as today's truth rather than a changelog.

The last two do most of the heavy lifting. When the current reality meets how we'll know it's done, the item is finished, and AJ says so.

What AJ does

AJ is the AI inside Clearahead, and its job is to replace the coordination machinery, not the people. In practice that means five things:

1. It shapes. It helps set a sharp goal, pushes past symptoms to the real constraints, and proposes moves.
2. It gathers. It turns each person's update, in their own words, into evidence, and keeps the current reality up to date all the way up from the move to the constraint to the goal.
3. It reports. Every Monday morning it writes This week from AJ: where you stand, what moved, what you learned, and up to three things that need a decision. When the goal changes, it re-checks every constraint against the new one.
4. It notices. It says when a move looks done, a constraint looks cleared or the goal looks achieved, and offers the one-click way to close it.
5. It remembers. It keeps how each constraint was cleared, so it can tell you that a few months ago you cleared something similar, and what worked.

The people

There are three roles, and they're enforced by the system rather than by etiquette. Architects, usually the founder, the MD or COO and the heads of department, set the goal, confirm and complete constraints, and make the calls AJ says need them. Contributors are anyone doing the work; they own moves, record progress in a sentence, and propose new constraints and moves. Viewers, such as a board, advisers or investors, can read the whole model and ask AJ questions without anybody preparing a pack for them.

Two things matter more than the labels, though. Every move has exactly one owner, so nothing is owned by "the team"; and the Architect's real job changes, with a lot less chasing and collating and a lot more deciding.

The rhythm

Here's the part I most want to get across: most of this rhythm isn't meetings. It's what AJ does on a schedule and what people do in Clearahead as the work moves, and the team only needs to get together when there's a decision to make.

It's also a suggested way to use Clearahead, not a schedule the app forces on anyone, so each team can decide how best to fit it around the rhythm it already has. The as-it-happens, daily and weekly parts are built into Clearahead today. The monthly and quarterly steps aren't in the app yet; for now they're habits a team runs for itself, using what Clearahead already shows. To see how it would fit your own team, try the cadence planner.

As it happens: updates in a sentence

When something moves, the owner says so in their own words, in about a minute, to AJ. AJ files it as evidence, brings the current reality up to date and tells the owner if it looks done. Nobody saves their updates up for a meeting, and nobody has to chase them.

Daily, if you want it: your summary

Each person picks the days and the hour they want a summary email. It leads with what's theirs, such as anything they've been made owner of, changes to their moves and what's due this week, and then what changed across their workspaces. If nothing changed, nothing gets sent.

Weekly: the Monday brief

On Monday morning, This week from AJ lands on the goal page and in everyone's inbox: a read on where you stand, what moved (biggest win first), what you learned, and up to three things that need you, each with a button that opens the right place. Most of those decisions get made right there, by the person who owns them. Only when something needs the leadership team to decide together do they talk, and then the conversation is about that one decision, never a round of status updates. The brief also refreshes itself after activity, so on Thursday it's as current as it was on Monday.

Monthly: the constraint review (not in the app yet)

Once a month, an Architect looks over the constraints in Clearahead, on their own or with the people who own them, and asks one question: is the one at the top still what most needs clearing? It's a check, not a meeting. Re-order the constraints so the top one is what matters most now, and keep three to five active; confirm the suspected ones the evidence now supports and archive the ones it doesn't; look at anything that has gone quiet and ask whether it's still true or has quietly cleared; and plan the next moves under the top constraint, so it never runs out of work.

Quarterly: the goal check (not in the app yet)

Once a quarter, the leadership team steps back and checks the goal itself. It's a pause to look at the bigger picture rather than a scoring session, and it doesn't need a set meeting. Is the goal still the right goal? If it changes, AJ re-checks every constraint against the new one and says which still fit. Look back at the constraints you've completed, how each one was cleared and the wins since last time. And run the four questions from Why are you doing that? on every recurring meeting and report, and drop, shrink or change at least one of them.

When it happens: the turning points

A constraint clears. An Architect marks it complete and records what cleared it, which becomes evidence on the goal and part of the company's memory, and the next constraint moves up.

The goal is achieved. The team marks it achieved, AJ writes the story of how you got there, people add their reflections, and the workspace becomes a record you can look back on. Then you set the next goal, ideally the same week.

What it replaces

Every old ritual existed to move information around. In this model the information is already where it needs to be, so each ritual either goes or shrinks down to the decision it was hiding.

The principle survives in every one of those. Focus, visibility and accountability are all still there; what goes is the hours spent carrying information to them.

Starting it in a week

A team can be running the loop within a week, and the setup itself takes a few hours rather than months.

1. Day one, an hour with AJ. An Architect sets the goal with AJ: the outcome, a rough when, and how you'll know it's done.
2. Day one, the question: what's holding you back? AJ asks it and pushes past the symptoms to the causes. You come away with a short list of constraints in your order of priority, ideally three to five.
3. Day two, moves under the top constraint. AJ proposes moves; the team keeps the ones it believes in, and each one gets an owner and a due date.
4. Day two, invite the team. Contributors pick up their moves and start updating in their own words, and Viewers can read along without anyone preparing anything for them.
5. The first Monday, the first brief. This week from AJ lands first thing; work through the things it says need you.
6. End of week one, stop one thing. Pick the recurring meeting or report that the brief has just made redundant, and drop it for a month.

The whole idea

None of this is complicated, and that's rather the point. The systems most companies run on were built when the only way to know what was going on was to ask everyone, in a room, every week. That's no longer true, and the companies that notice first get back the hours they've been spending on the coordination tax.

One goal, a clear view of what's holding it back, the whole team moving on it, and the machine doing the asking and the admin. That's the operating model for the AI era, and it's what we built Clearahead to run. If you want the beliefs underneath it, they're set out alongside our vision.