A strategy team that once spent an entire Tuesday morning wrestling its way through a single memo now produces four versions before the coffee has gone cold, and picks the one that reads best. Months later a board member asks why the third assumption holds, and the room discovers that nobody can reconstruct the argument, because nobody ever really made it. It was already there while the coffee was still warm.
This is not a story about laziness, and not one about bad AI either. It is a story about two unwritten principles that every company understandably runs on, because they seem too obvious to say out loud: results should get better, and effort should get smaller. Outcome-obsessed and effort-avoidant at the same time.
For most of economic history these two wishes pulled against each other, because better results usually cost more effort, and less effort meant worse results. Then a tool arrived that, for the first time, promised to serve both at once.
The break nobody notices
If a single instrument promises to lift the numerator and lower the denominator, a rational company should pull both levers with equal energy. It should ask what results are now possible that used to be simply too expensive, with the same stubbornness it applies to asking what can now be done more cheaply. Theory says: symmetry.
Practice says something else. In McKinsey's State of AI 2025, around 80 percent of surveyed companies name efficiency as a goal of their AI initiatives, and it is precisely the highest-performing organizations that additionally state growth and innovation as an explicit goal. The majority pulls mostly on one lever, the denominator, and the small group that also works the numerator turns out to be the same group that reports the larger benefit.
You could explain this with corporate caution, with quarterly pressure, with the fact that savings are easier to write into a balance sheet than hypothetical growth. All of that is true and still not enough, because it does not explain why the imbalance shows up so reliably, so universally and so rarely questioned. A deliberate decision would show variation. What we see looks more like a reflex.
Why effortlessness feels like a win
Here it is worth taking a detour through the brain, because that is where the real lever sits, and it sits below the level on which strategy is negotiated. No reward is evaluated on its own; it is evaluated against the effort it took to get it. The brain constantly computes a ratio, and the emotional price sits in the denominator of that calculation.
One substance runs this, and almost everyone believes the wrong thing about it. Dopamine is taken to be the pleasure molecule, the stuff released when something is nice. In fact Schultz, Dayan and Montague showed in Science in 1997 that dopamine is mostly a prediction error, a signal that reality turned out better than expected. It fires most strongly when a good outcome arrives more easily than anticipated, and it stays silent when the outcome comes exactly as you already expected it to.
From this follows something uncomfortable: anyone who drops the effort for an unchanged reward to nearly zero is not building a particularly efficient workplace, but the neurological profile of a slot machine. Reward at minimal input, again and again, with just enough variability in the outcome that the next pull is appealing once more.
AI becomes a slot machine at the individual level, not after someone explains it, but after you have experienced it a few times.
The board that swaps four effortlessly generated memos for one laboriously fought-for thesis has no worse work ethic than its predecessors. It has a tool that, every time the lever is pulled, produces a small upward prediction error, and its brain treats that stimulus the way it treats every other stimulus of that build.
The Behavioral Misfit
Here is the point where two logics separate that everyone takes to be identical. The ROI logic claims to maximize a ratio of output and effort. The behavioral mechanism underneath, however, rewards not the ratio but the shrinking denominator alone. Both look at the same formula and mean different things, and exactly this gap is what I call the Behavioral Misfit: the distance between what a system claims to optimize and what it actually rewards.
As long as the two quantities pulled in the same direction, the misfit stayed invisible. AI separates them cleanly for the first time, because it can lower effort so radically that the effort reduction becomes the actual reward and the result is only the pretext that justifies it. You believe you are optimizing ROI, and in fact you are pressing down the emotional price, because that feels better in every single moment.
A deliberate decision to use AI for cost cutting is legitimate and often right. The problem is not the decision, but that the same shift also happens without any decision, driven by a reward mechanism that is never recognized as one and therefore never questioned.
What feels like entrepreneurial cleverness is, over long stretches, a collective dodging of exertion that only reads like a strategy by accident.
The question almost nobody answers with yes
There is a simple test for the suspicion, and it fits in one sentence. Who do you know who, because of AI, now puts in more effort than before, because they are aiming at a result that used to be out of reach, instead of producing the same result more cheaply?
Most people who hear this question pause and then name nobody. They know plenty who do the same thing faster, cheaper and with fewer people. They know almost nobody who reinvested the freed-up effort to get at a harder question. That is not a character flaw of an industry, but the predictable consequence of a mechanism that rewards effort reduction immediately and output improvement only much later, if at all.
Whoever wants to close the misfit has to shift the benchmark, and to do so before the brain shifts it on their behalf. What decides is not the question of how much faster a task was completed, but the question of which task AI made solvable in the first place. The saved time is then not a gain you book, but a budget you spend, on precisely the exertion the system is so pleasantly offering to take off your hands.
The honest answer to the benchmark question is uncomfortable, because it demands that you voluntarily put effort back into a system that has just learned to reward you for avoiding it. That is intentional.