Who's Driving your AI Transformation?
You may not be in the driver's seat. But here's three important ways to keep your hands on the wheel.
There is a story that business leaders tell themselves about transformation, and it is not entirely wrong. The story goes like this: technology creates pressure, but culture is a choice, and the organizations that navigate disruption well are the ones whose leaders see clearly, decide deliberately, and bring their people through with integrity. The story is flattering and mostly true. It has also, in every prior chapter of forced enterprise transformation, been accurate enough to be useful.
I want to suggest that in this chapter of transformation, it is less accurate than it has ever been — and that the leaders who navigate this one well will be the ones honest enough to say so.
What History Actually Taught Us
Every major wave of technology-driven enterprise transformation shares a basic structure. Competitive pressure makes adoption effectively compulsory; firms that don’t move lose ground to firms that do; the workforce is reorganized around new capabilities. Leaders experience this as a choice — and in important ways, it is — but the direction of travel is determined by the market, not the boardroom.
What varied across chapters was how much time leaders had, how deeply the technology reached into the core of the work, and how much individual firm behavior could diverge from the competitive mean.
Consider electrification. When factories converted from steam power to electric motors in the early twentieth century, the productivity gains were real but slow to materialize — so slow that economists puzzled over them for decades. The reason, we now understand, is that most firms simply bolted electric motors onto factory layouts designed for steam. The firms that redesigned their operations from scratch captured gains an order of magnitude larger. The lesson often cited is about redesigning rather than retrofitting. The lesson less often cited is that leaders had thirty years to figure this out. The technology waited for them.
ERP systems in the 1990s compressed that timeline considerably. SAP and Oracle effectively compelled adoption among large enterprises — firms that didn’t implement enterprise resource planning found themselves at structural disadvantages that became competitively untenable. The workforce effects were significant: entire functional roles were reorganized, some eliminated. But ERP implementations ran on three-to-seven year cycles. Firms could observe competitors, pilot cautiously, fail, remediate, and retry. Deliberate leadership made a measurable difference.
The internet wave was faster still. But it was fundamentally a transformation of connectivity and distribution — it changed how firms reached customers and coordinated internally while leaving the cognitive core of knowledge work largely intact. The judgment, the expertise, the human relationships that produced value remained essential and in many cases became more valuable.
Each of these chapters offered leaders a version of the driver’s seat that was real enough to be meaningful. The direction of travel was not entirely chosen, but the pace, the form, and the human consequences were substantially within reach of leadership judgment.
This chapter is different in three ways that compound each other.
Speed, Depth, and the Trap
The first difference is speed. The competitive clock on AI transformation runs in months, not years or decades. The adoption pressure that took a generation to build in electrification, a decade in ERP, several years in internet transformation — is already acute and accelerating. The time available to observe, pilot, adapt, and course-correct is categorically shorter than any prior chapter offered.
The second is depth. Every prior wave of enterprise transformation reached into process, into task, into the operational and clerical layers of organizations — and left human judgment as the essential residual. The knowledge worker, however reorganized, remained necessary. This wave automates the verifiable cognitive layer for the first time: the drafting, the analysis, the pattern recognition, the synthesis that was previously thought to be the safe harbor. The technology is reaching into the core of what knowledge workers do in a way that has no real precedent in the history of enterprise transformation.
The third is the hardest to say, and the most important.
The rate at which your organization automates is not entirely your decision.
Recent economic research on AI-driven labor displacement describes what it calls an automation arms race — a dynamic in which rational, fully informed firms are locked into automation decisions by competitive structure, not by choice. The mechanism is precise: each firm that displaces workers captures the full cost saving of doing so, but bears only a fraction of the demand destruction it causes, because displaced workers are also consumers whose spending supports every firm in the market. The rest of the demand loss falls on competitors. No individual firm can afford to be the one that holds back while rivals move.
What makes this structurally different from prior chapters is that this is not a failure of vision or courage. Every prior transformation produced leaders who resisted and suffered for it — Kodak, Blockbuster, the retailers that watched e-commerce arrive and looked away. The lesson drawn from those failures was: see clearly and act decisively. What the current dynamic describes is something that lesson cannot address.
The firms that see clearly, that understand exactly where this leads, that can model the collective consequence of everyone automating simultaneously — they automate anyway, because the game theory is the game theory. Seeing the cliff does not change the calculus.
There is a further turn of the knife. The natural assumption is that as AI improves, the disruption eventually stabilizes — productivity gains create new demand, new roles emerge, the labor market finds a new equilibrium. In prior chapters, this assumption was broadly correct. In the current dynamic, it is precisely inverted. Better AI worsens the competitive pressure rather than relieving it. Each capability improvement adds a market-share motive on top of the demand externality: every firm perceives a gain from moving faster than rivals, but at the symmetric equilibrium those gains cancel, leaving only the additional displacement. The technology getting better tightens the trap. This has no precedent in the history of enterprise transformation.
What Remains in Your Hands
I want to be careful here, because the point is not that leadership doesn’t matter. It does, and I believe it does. The point is that leadership’s genuine discretion operates over different things than the driver’s seat metaphor suggests.
Job architecture is in your hands — but only before the unbundling happens. Recent research on how AI redraws job boundaries shows that the critical variable is not which tasks AI can perform, but how costly it is to separate those tasks from the human roles that currently contain them. Jobs where the codifiable and the contextual are deeply integrated — where what you learn doing one part of the work makes you better at the other — resist displacement in ways that loosely assembled task collections do not. Deliberately designing roles around that integration is a genuine strategic choice that most organizations are not treating as one.
The window for building displacement-resistant jobs is before the competitive pressure fully arrives, which means now.
The human pipeline is also in your hands. What happens to the people whose roles are reorganized — whether they are treated as a liability to be shed or a pipeline to be maintained — is a leadership decision that competitive pressure does not fully determine. The organizations that invest seriously in internal mobility and reskilling are doing something that matters both for the people involved and for the long-term health of their own expertise base. The junior pipeline that gets destroyed in this transformation does not rebuild quickly. Understanding that before it is gone is the difference between foresight and regret.
And there is a form of agency available to leaders that prior chapters of transformation did not make as necessary: the willingness to name the structural problem publicly. The economic analysis of this dynamic is precise about what individual firm virtue cannot accomplish, and equally precise about what could. No amount of internal goodwill, voluntary restraint, or worker equity programs changes the dominant-strategy structure of a competitive market. The intervention that actually reaches the problem has to operate at the market level. Leaders who understand this and say so — who are willing to make that argument rather than retreating to the comfort of individual firm virtue — are exercising a form of agency that actually reaches the problem.
The leaders who will look back on this period with integrity are not the ones who believed most confidently that they were in the driver’s seat. They are the ones who were honest about which things were actually in their hands — and didn’t confuse the illusion of agency over the automation rate with the genuine agency they had over everything else.
What Good Leadership Looks Like Here
The car is moving. The question worth asking is not whether you’re driving. It’s what you’re doing with what you actually control.
That means designing jobs before the unbundling, not after. It means treating internal human capacity as a pipeline to maintain rather than a cost to eliminate. It means being honest with your organization about the competitive forces that are constraining your choices — because people can distinguish between a leader who is making hard decisions thoughtfully and one who is pretending that the hard decisions were freely chosen. And it means being willing to say, in public, that some of the most important decisions affecting your workforce cannot be made by any individual firm alone.
Every prior chapter of enterprise transformation rewarded leaders who moved deliberately and designed thoughtfully. This one will too. The difference is that the margin for error is smaller, the time available is shorter, and the things worth designing are different from what they used to be.
The historical record should give leaders humility about agency, not paralysis. Leaders navigated every prior transformation — some well, some badly, most somewhere in between. The ones who navigated well were not the ones who felt most confident in their control. They were the ones who understood, clearly and without flinching, exactly what was and wasn’t in their hands.
That understanding has never mattered more than it does now.
Sources
Hemenway Falk, B. and Tsoukalas, G. “The AI Layoff Trap.” arXiv preprint arXiv:2603.20617v1, March 2026. Source for the automation arms race, demand externality mechanism, and dominant-strategy analysis of competitive AI adoption.
Garicano, L., Li, J. and Wu, Y. “Weak Bundle, Strong Bundle: How AI Redraws Job Boundaries.” Working paper, March 2026. Source for bundle theory, job architecture as strategic variable, and the displacement mechanism in weak-bundle occupations.
Brynjolfsson, E. and Rock, D. “The Productivity J-Curve: How Intangibles Complement General Purpose Technologies.” NBER Working Paper, 2021. Source for the electrification precedent and the thirty-year lag in productivity gains from factory redesign.
Catalini, C., Hui, X. and Wu, B. “Some Simple Economics of AGI.” MIT, February 2026. Source for the Hollow Economy framework and the structural risks of AI-driven labor displacement.