Why 'What's Your AI Strategy?' Is the Wrong Question

The wrong question

At some point this year, probably at an offsite, or in a board pack that landed the night before, someone is going to ask what your AI strategy is. A board member, an investor, a chair, maybe your own head of operations who has started to feel exposed on the subject. It tends to arrive phrased like a due diligence question, as though AI were a discrete initiative sitting alongside market expansion and cost reduction, waiting for its own line in the deck.

It is a fair question to be asked. It is the wrong question to spend real time answering, because it assumes the thing that is actually broken in most businesses is a decision nobody has made yet, rather than a structure that has never existed.

Here is the structure that is missing. In almost every business we work with, the commitments the business has made, to a customer in a contract, to the board in a plan, to investors in a set of targets, live in one place. The tasks people actually do this week live somewhere else entirely: a project tool, a spreadsheet, an inbox, or nowhere written down at all, just carried around in someone's head. The outcomes get reviewed later, usually quarterly, in a room with a clear view of neither the original commitment nor the daily work that was meant to deliver it.

Ask almost anyone in that business, on an ordinary Tuesday, why they are doing the task in front of them, in terms the business would recognise as commercial rather than procedural. Watch how long it takes them to answer. Most people can tell you what they are doing. Fewer can tell you which commitment it ladders up to. Almost nobody can do both quickly, in language that would mean anything to whoever is reviewing outcomes next quarter.

Where the thread actually snaps

We see the shape of this in nearly every engagement, regardless of sector or size. It shows up as a scramble the week before a board meeting, when someone has to reconstruct, from memory and a scroll back through Slack, what actually happened against a commitment made months earlier. It shows up as a dropped ball nobody notices until a client asks about it. It shows up as a leadership team that is, in the politest sense, flying blind between review cycles: not lied to, just structurally unable to see the connection between what was promised, what has been done, and what is left.

None of this is because the commitment, the task and the outcome are unrelated. They are deeply related. The account manager doing the task usually knows exactly why it matters. The problem is that this knowledge lives in her head, and nowhere else. It is not written into the CRM. It is not written into the project tool. It is certainly not written into the quarterly deck. The moment she goes on leave, changes role, or simply forgets, the thread goes with her, and the business is left holding three separate records of the same piece of work with no connective tissue between them.

Good people, broken architecture

It would be easy to read that and conclude the fix is more discipline: better handover notes, stricter reporting, a programme manager whose job is to nag. We would resist that conclusion, because we do not think this is a discipline problem, and treating it like one mostly just makes good people feel blamed for a structure they did not design.

The commitment lives in a proposal or a contract, owned by sales or the leadership team. The task lives in whichever tool a delivery team has settled on this year, or the one before that, or both, because the migration never finished. The outcome gets assembled by whoever is closest to the board pack, usually from whatever fragments they can find in time. Three different owners, three different tools, three different points in time, and no single place where the line between them is drawn and kept current. That is not a people problem. That is an architecture that was never built to hold the thread in the first place.

We have tried this for forty years

If reconnecting strategy to daily execution sounds familiar, that is because it is genuinely old, and if you carry scar tissue from a previous attempt, that scepticism is earned, not a misunderstanding. Management by objectives set out to do this in the 1950s: cascade agreed objectives down through a hierarchy and review performance against them. Hoshin Kanri, developed inside Japanese manufacturing and exported west from the 1980s onward, tried to do it with more rigour, using a formal back-and-forth between levels, known as catchball, to align strategy and tactics on one working document. OKRs, refined at Intel and popularised out of Google, were the most recent serious attempt: a small number of objectives, each with measurable key results, reviewed on a fixed cycle.

All three were built on a correct insight. All three decayed, in company after company, for the same reason. Keeping the thread current is clerical work: updating the tracker, chasing the check-in, running the catchball session, reconciling this week's tasks against last quarter's objective. That work is unglamorous, it produces nothing anyone can point to on its own, and it is the first thing dropped the moment whoever owns it gets pulled towards something more urgent, which in most businesses is always. The scorecard goes stale within two quarters. The X-matrix becomes a poster nobody updates. The OKR tool becomes one more login people avoid. Often, the decay lines up almost exactly with the consultants leaving, because they were, in effect, the clerical labour holding the thread together while they were still in the room.

The idea was never wrong. The economics of maintaining it were.

What has actually changed

Nothing about this problem has changed in seventy years except one thing: there is now, for the first time, a worker who does not get bored of clerical reconciliation. An AI agent can sit at the boundary between the contract, the task tool and the outcome review, and do the unglamorous linking work every day, not just for the two quarters after the consultants leave. It can read what was committed, read what has actually been done, and surface the gap before the board meeting instead of during it.

None of this removes the need for a person to decide what to do when a gap shows up. It just means the gap shows up continuously, while there is still time to act on it, instead of once a quarter, when it is usually too late to matter.

This is the actual argument for AI in most businesses, and it has almost nothing to do with the version of "AI strategy" that shows up on an offsite agenda. It is not about a chatbot that drafts your emails faster. It is about closing, for good, the specific structural gap that management by objectives, Hoshin Kanri and OKRs all correctly diagnosed and all eventually lost to clerical fatigue. To be clear what this is not: it is not a way to run the same business with fewer people watching it. It is how a business gets five to ten times the output from the people already in it, because those people stop spending their week reconstructing context a machine could have been holding continuously all along.

Why your current tools cannot do this either

There is a second, quieter reason nobody has closed this gap yet, worth naming because it explains why simply buying another tool will not fix it. Monday, ClickUp, Asana, the CRM your sales team already lives in: every one of them was designed around a single kind of participant, a human who logs in, gets assigned a card, and updates a status by hand. None of them were built with any native concept of an agent operating on those same records as a first-class participant, reading and updating the same task, the same deal, the same objective the humans are working from.

Bolting a chatbot into the corner of a tool built entirely around a human clicking through screens does not make a business AI-native, any more than fitting an engine to a horse-drawn carriage made it a motorcar. The carriage still had a bench built for someone holding reins and a chassis built for a horse's gait. The engine changed what powered it, not what it fundamentally was. A lot of businesses experimenting with AI right now have fitted the engine and kept the carriage, then wondered why the ride does not feel any different.

The question worth asking instead

So set aside "what's our AI strategy" for a moment, because answering it well still leaves the actual structure untouched. The harder, more useful question is this: where, in your business, has the thread between what someone does on an ordinary Tuesday and what the business is actually trying to achieve been cut, and what would it genuinely take to reconnect it in a way that survives past the point where anyone is paying close attention.

You can find out where your own thread is cut with a five-minute exercise rather than a workshop. Pick one live commitment, a client contract, a board target, whatever matters most to you right now, and try to trace it forward to this week's task list for the people responsible for it. Then try it in reverse: pick five tasks someone is doing today and trace them back to the commitment they are meant to serve. If either direction takes longer than it should, or dead-ends in someone's memory, you have found the actual gap. It has nothing to do with whether you have licensed the right AI tool yet.

Closing that gap for good is real work, and it takes months, not a weekend, and it is not a workshop you can outsource in an afternoon. It means deciding, in language precise enough for a machine to check against, what a commitment and an outcome actually are in your business, which most companies have genuinely never written down. It means deliberately bringing together systems that currently hold the pieces separately, because nobody designed them to talk to each other. But it is, for the first time, work worth doing properly, because what you build will not need a programme manager to nag it back to life every quarter. It will still be checking the thread long after everyone who was excited about AI this year has moved on to whatever comes next.