Tienviitta, jonka nuolet osoittavat eri suuntiin kohti Kilimanjaroa ja Mont Blancia, taustalla klassinen kivijulkisivu ja pilvinen taivas.
By /

After the Triangle: Where Decision Rights Go When AI Removes the Cost of Knowing

I have given roughly this same talk to more leadership teams than I can remember. I walk through the argument: that the strategy–tactics–operations pyramid was a clever solution in its time, that AI is dismantling its foundation, that decision-making power would be best held closer to where value is actually created. And almost every time, the same thing happens. There is nodding around the table. Someone says this is exactly what they should be talking about more. The mood is genuinely enthusiastic. And then — nothing. Six months later the org chart looks exactly as it did before.

For a long time I assumed the fault was in my talk: not convincing enough, not concrete enough. Eventually I realised it was not about understanding at all. Everyone understood. They simply couldn’t. That gap — between a structure whose rationale is visibly crumbling and an organisation that cannot move — is the most important operating-model question most enterprises are not asking. So let me ask it plainly: if the logic is this clear, why does nothing change?

The triangle is an invention, not a law of nature

Start with where the triangle came from, because almost nobody in those meetings knows. It is not the natural form of organisations. It is an invention, and not an old one — roughly the age of the photocopier. The best-known source is Robert Anthony, who in 1965 divided management into strategic, tactical and operational work to help managers steer post-war corporations that had grown too large for any one person to know in full.

Here is the part that stopped me. Anthony did not draw a hierarchy. He described three kinds of work — not three layers of people — and he said outright that the same person does all of them. The head of a shipping line, he wrote, practises both strategic planning and operational control: one person moving between kinds of work, not up and down a staircase. Somewhere between Anthony and now, processes became layers, layers became people, and people became a hierarchy of status. “Strategic” came to mean important and high up; “operational” routine and low down. That was a misreading — and the misreading set into an org chart.

And it worked, for a real reason. Anthony’s triangle was the right answer to a world where information moved slowly and expensively. When no single person could know everything in time, it made sense to gather information at the top, decide there, and send instructions down. The layers were a machine for managing the scarcity of information. The whole structure rested on that one assumption.

AI breaks the assumption the triangle was built on

AI strikes at exactly that assumption, from three directions at once.

It collapses the cost of moving information. Organisational layers are, at bottom, machines for compressing and filtering knowledge — someone gathers the figures, someone condenses them, someone turns them into a slide. Language models now do that in seconds. A filter whose existence used to be expensive becomes the press of a button.

It resets the speed. When the machine accelerates the decision loop, the pace is no longer yours to set. It is like cycling in the lead group: the front dictates the speed, and the slow one is dropped from the loop of their own field.

And it eats the operational base first. The routine “what” — collation, reporting, first drafts, retrieval — is what AI takes over soonest. Harvard and MIT researchers followed over five thousand customer-service agents and found AI raised productivity by around 15% on average, most of all for the least experienced. But it does not eat evenly: on tasks beyond the jagged technological frontier Dell’Acqua and colleagues identified, people using AI produced the correct answer markedly less often than people with no machine at all. Knowing which side of that boundary you are on is human work.

When those three constraints break at once, the levels stop needing different people. In a 2025 Procter & Gamble field experiment with nearly 800 professionals, one person with AI reached roughly the level of a two-person team without it — and the proposals crossed silos that unaided experts stayed inside. This is Anthony’s original insight, restored: one person can again stand on all three levels within a single sitting. The levels do not disappear. They stop being different people.

Decision rights follow knowledge — to the edge

So where does decision-making power go? The sharpest answer is from economists Philippe Aghion and Jean Tirole, who separated formal authority — the right to decide, what the chart says — from real authority, actual control over the decision. Their observation: these rarely sit with the same person. The busy manager rubber-stamps the specialist’s proposal, because they cannot know the matter as well. Real authority follows knowledge. It drains to where the matter is known, whatever the chart says. When AI gives the person at the edge the same whole picture the centre used to monopolise, real authority drains to the edge — whether the chart likes it or not.

This is not a new management fad. Helmuth von Moltke built the Prussian general staff on it: give intent, not orders — the objective and the why, not the how. General Stanley McChrystal rediscovered it in Iraq, pairing radical transparency (shared consciousness) with pushed-down execution; the tempo of operations rose from a few a month to around three hundred, and decision quality rose with the speed, not against it. Netflix’s Reed Hastings distilled the same into context, not control.

For a leader this means three changes that sound easy and are hard. Give intent, not orders. Share the whole picture, not a filtered version — AI finally makes that sharing cheap. And, hardest, tolerate the edge deciding differently from how you would have. If every decision must be made your way, you have not decentralised power; you have outsourced the typing. Empowerment without the right to disagree is no empowerment at all.

The honest counter: flattening reverses

Here intellectual honesty demands a stop. Taking power to the edge can go badly wrong, and the data on flattening cuts both ways. Yes, the average number of subordinates per supervisor has risen, on Gallup’s figures, from just over 8 in 2013 to just over 12 in 2025; Amazon and Bayer have publicly cut layers. But every previous wave of flattening has reversed. When offices were computerised in the 1980s and 1990s, middle management was cut on exactly today’s logic — and then came back, bigger, as coordination proved harder than expected. Over the decades, managers’ share of the US workforce did not shrink but grew, from around 9% in the 1980s to over 13% by the early 2000s. Gartner and Forrester already forecast that many roles cut in AI’s name will be quietly hired back.

Is this time different? Perhaps — AI does the very coordination the middle rung existed for, not some vague efficiency. Or the cycle repeats, because part of that middle work turns out to be a human quality the machine doesn’t do. I don’t know which it will be. And I don’t trust anyone who claims to.

There is a deeper reason the pyramid survives its own obsolescence, named best by James C. Scott: legibility. The chart makes the organisation legible from above — who is responsible, who reports to whom, whose desk a problem lands on. A flat, networked organisation may be more efficient, but from above it is blurry, and you cannot get a grip on the blurry. We don’t hold onto the pyramid because it works best. We hold onto it because it is legible. Redrawing the chart is easy. Dismantling what the chart represents — power, pay, career, esteem — is something else entirely.

What is built can be rebuilt

That is also the reason for hope. If the pyramid were the gravity of organisations, there would be nothing to do about it. But it was built, for a particular world — and what is built can be rebuilt when the world changes. The move that matters is not swapping the triangle for one new correct picture; it is giving up the idea that the top layer is the most valuable, when value is created at the edge, where the customer is actually met.

The technology has made this possible. It will not do it for you. Whether decision rights drain to the edge or pile up ever more tightly at the top is not a technical question but a leadership choice — and that choice depends entirely on people: whether leaders dare to give intent and to trust, and whether the person at the edge will carry the responsibility that comes with the power. Technology gets you to this point. People decide which way you go from here. Competitive advantage isn’t in the machines. It’s in people.

Lenni Laukkanen is the founder of Astu Labs and helps leadership teams turn AI into a competitive advantage through people. For speaking and advisory enquiries: lenni@lennilaukkanen.fi — I reply within one business day.

Sources

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top