Your company has amnesia. The cure is a new kind of operating system.
Why the next advantage is not using AI more, but building a company that finally remembers.
Here is a preference most people share without examining it. If I am seriously ill, I want a human doctor working with an AI, not a human doctor alone, and not an AI alone. The reason is not sentiment. It is that no human, however brilliant, can read and remember all of medicine, hold every trial and interaction and edge case, and stay perfectly objective at four in the afternoon on their tenth patient. And no machine, however capable, should be the one that looks me in the eye and decides. The pair beats either half. The human brings judgement and care and the taking of responsibility. The AI brings a memory nobody could carry alone.
Now look at your company through that lens, and something uncomfortable comes into focus. Your organisation is the patient, and it has amnesia.
What your company cannot remember
Think about everything a company knows and then loses. Every hypothesis it has ever made about its customers, its market, its own operations. Every experiment it ran, and what the result actually was. Every decision, and the reasoning behind it, which evaporates the moment the meeting ends. Every win worth repeating and every loss worth not repeating. The external signals someone noticed but nobody wrote down. And all of it sitting in the heads of people who, sooner or later, leave, and walk out of the building with the memory in them.
No leadership team can hold this. Not because they are not clever, but because it is not humanly holdable. You cannot remember what your company hypothesised three years ago, whether you tried this exact thing before, what you learned, what the market was doing at the time, and simultaneously bring a cool, unbiased head to the argument in the room today. Nobody can. So companies decide the same things twice, relearn lessons they already paid for, and defend positions with confidence and no memory. That is not a talent problem. It is an architecture problem, and it is the one almost nobody is solving.
Why “we use AI” misses the point entirely
Most companies, asked how they use AI, will tell you they chat with it, they code with it, they have it run some routines, they use it for design and drafts. All useful. All real. And all of it treats AI as a better tool: a faster hammer, a tireless intern. A tool is something you pick up, use, and put down. It does not remember your company. It does not get wiser about you between uses. Tomorrow it starts from the same blank the average always starts from.
The re-imagination is not “use the tool more.” It is to build something that has never existed cheaply before: an operating system with a self-improving architecture, that remembers your company, reasons over that memory, brings the objectivity no insider can, and gets sharper every quarter, with humans in the loop owning every call that carries stakes. Not a tool you chat with. A system that thinks alongside you, and does not forget.
That is the doctor and the AI, made corporate. The machine carries the memory and the recall and the outside view. The humans carry the judgement, the stakes, and the care. Neither half is the product. The pair is.
The quiet part: you are already generating the memory, or you are wasting it
Here is where it stops being science fiction and becomes a Monday-morning question, and where our own work sits. What would this OS feed on? Not a heroic, separate “knowledge management programme” that everyone ignores. It feeds on the exhaust of running the company well, if you happen to run it well.
Watch what good goal practice actually produces. Teams set outcomes tied to strategy, not activity. Those outcomes sit on a metric model, so the numbers mean the same thing everywhere. The work underneath is named honestly as initiatives, recurring commitments, and experiments. And, done properly, people write the narrative: why this goal, what we believe, what we are betting, and at the end, what happened and why, the wins and the losses, explained. Every one of those is a small, structured, dated act of memory. A quarter run like this is a company quietly writing its own training data, in its own words, about its own reality.
Most companies throw all of that away. The goals get scored and filed, the narratives never written, the learning never captured, the reasoning lost. Run the same quarter with the capture wired in and the identical effort produces a compounding asset instead of a bin. That is the hinge. Good goal-setting was never only about better goals. It is the richest, cleanest source of institutional memory a company has, and it is generated as a byproduct of work people are already doing. The self-improving OS is what happens when you stop letting it evaporate.
What the new OS actually does
Four things, none of which a chat window does:
It remembers, so the memory survives the people. When someone leaves, what they knew stays, because it was captured as they worked, not extracted in an exit interview nobody reads.
It surfaces, so the relevant prior hypothesis, the experiment you already ran, the lesson you already paid for, and the external signal that matters, arrive in the room at the moment of the decision, not six months after it.
It stays objective, so the debate has a participant with no ego, no politics, and fewer biases, that can say “you tried a version of this in 2023 and here is what happened,” and “here is the case against what you are all nodding along to.” The outside view, on demand, from the inside.
And it improves itself, so every cycle’s evidence, every graded retrospective, every closed experiment, feeds back and makes the next answer better. The loop tightens. The company gets smarter as a company, not just as a collection of smart individuals who will eventually leave.
Why this is the advantage, not a nice-to-have
Because it compounds on the one thing no competitor can copy. We proved the copyable part on ourselves: we handed an AI our public website and asked it to rebuild our method, and it scored 68 out of 100. Our beliefs cloned at around 90. But the evidence, the memory of what we specifically tried and learned, cloned at zero, because it never happened to anyone but us. A company that remembers is building the asset that a rival cannot photograph, and that grows every quarter it runs. Impossible to catch up with is a different game from impossible to copy, and this is how you play it.
And it is humane, which is the part leaders feel before they can justify it. When the machine carries the remembering and the recall, people are freed to do the work only people can do: the judgement, the creativity, the relationships, the care, the hard calls with something at stake. You are not replacing the doctor. You are giving the doctor a memory of all of medicine, so the human parts of the job get more time, not less.
The invitation
So the question for a leader is not “how do we use AI more.” It is bigger and more interesting: what would it mean to run a company that remembers everything it has learned, surfaces it exactly when it matters, argues with itself honestly, and gets wiser every quarter, with your best people in the loop where it counts?
That is a new kind of operating system, and it is buildable now, from the work you are already doing, if you stop throwing the memory away. Most companies will keep treating AI as a faster intern and wonder why the advantage never arrives. The ones that re-imagine the operating system itself will pull away, quietly at first, then not.
Your company does not have to keep forgetting. That was only ever an architecture problem.
Build the company that remembers.
This is the work we do: designing that operating system, wiring the memory and the learning loops, and keeping humans in the loop where judgement lives. If you want to see what it looks like on your own company, that is a conversation, not a download.
Talk it through with Matt →A free, structured briefing that gives ChatGPT, Claude or Copilot our method and your context, so the AI you have starts giving strategy-grade answers. No sign-up, no charge.

A UK-based strategy and OKR consultant and two-time SaaS founder with a venture-backed exit, Matt turns strategy into execution for teams scaling from tens to thousands. He co-founded ZOKRI in 2018, having previously co-founded Linkdex, a venture-backed enterprise SaaS platform he led to a trade sale. He writes the methodology behind these notes.