The AI Revolution

This is a revised version of a post originally published on February 25, 2025.

There is a version of the AI story that is almost irresistible. You deploy automation, your costs fall, your margins rise, and your shareholders are pleased. For a quarter or two, perhaps even several years, the numbers look exactly right. What this version of the story omits — what it is, in fact, designed to omit — is the ending.

I have spent more than two decades working across technology companies — some R&D-intensive, others built around software and services — and one pattern I have observed consistently is that leaders who optimize for near-term efficiency at the expense of long-term capability tend to find themselves with organizations that are leaner, faster, and increasingly brittle. The AI wave presents this temptation at a civilizational scale. The question worth asking is not whether artificial intelligence will transform how we work — it will, and in many respects already has. The question is whether the people making those deployment decisions are thinking far enough ahead, and about enough of the right things.

The Math That Leaders Choose to Ignore

The economic case for aggressive automation is straightforward, and I have no interest in pretending otherwise. Companies adopt these technologies because they work — they increase output, reduce error, and, yes, lower labor costs. The McKinsey Global Institute estimates that up to 30 percent of hours currently worked globally could be automated by 2030, and that between 75 million and 375 million workers may need to switch occupational categories entirely as a result.¹

The economic case against treating automation as primarily a cost-reduction exercise is equally straightforward — and considerably less discussed in boardrooms. When workers are displaced faster than new roles emerge, their purchasing power contracts. The customers who were supposed to buy the products made more efficiently by machines are the same people whose wages have been suppressed or whose jobs have disappeared. Henry Ford demonstrated this logic in 1914, when he more than doubled his workers' wages to five dollars a day. His stated reason was unsentimental: a roughly 370-percent annual turnover rate was destroying productivity, and higher pay was, as Ford himself later put it, "one of the finest cost-cutting moves we ever made."² The byproduct was equally instructive — workers who could now afford the cars they built, expanding the very consumer base that made Ford's volumes possible. The principle embedded in that outcome has not changed; only the scale of the test has.

The International Monetary Fund's 2024 analysis of AI's labor market impact concluded that in most scenarios, the technology is likely to worsen overall inequality — both within countries and between them.³ Capital income and wealth inequality increase in virtually every modeled scenario, because AI-driven productivity gains flow disproportionately to those who already hold assets. That is not an argument against AI. It is an argument for taking the distributional consequences seriously, which most deployment strategies currently do not.

Societies under sufficient economic stress become unstable. Instability is bad for markets, bad for regulatory environments, and bad for the long-term confidence that businesses require to plan and invest. The leaders who focus only on their own cost structure while the broader economy absorbs the disruption are, in effect, free-riding on a commons they are quietly degrading.

What Machines Cannot Own

There is a tendency, in discussions of AI's limits, to make claims that are grander than the evidence currently supports. I want to be more careful here. The assertion that machines are categorically incapable of creativity or judgment is harder to sustain in 2025 than it was even a few years ago. Language models write, compose, and reason in ways that were not anticipated; their capabilities continue to expand.

What I am more confident in is a narrower and perhaps more important claim: that the application of those capabilities — deciding what to build, why it matters, who bears the cost, and what kind of organization we want to be — requires human judgment, human values, and human accountability. An AI system can optimize toward a metric. It cannot be held responsible for choosing the wrong one. It can identify patterns in data at speeds no human analyst can match. It cannot feel the weight of what it means to tell a thousand people that their roles have been eliminated.

This is not a small distinction. The organizations that will navigate the AI transition well are those that invest in leaders capable of making those harder calls — people with the intellectual curiosity to understand what the technology can and cannot do, the moral clarity to decide how it should be applied, and the relational depth to bring their teams through uncertainty with trust intact. None of those qualities emerge from a software deployment. They are cultivated, over time, in people — and they require sustained investment to develop.

The Stewardship Imperative

The word I find myself returning to, when I think about how AI should be governed within organizations, is stewardship. Not ownership, not optimization — stewardship. A steward holds something in trust: not simply for their own benefit, but for the benefit of those who depend on them and those who will come after.

Applied to the AI transition, stewardship means asking questions that pure return optimization tends to skip. Who in the organization will be most affected by this deployment, and what support will they receive? Are we eliminating roles because the work is genuinely no longer needed, or because automation makes it easy to do so and we have not thought carefully about the consequences? Are the productivity gains being shared in ways that reinforce organizational commitment, or captured entirely at the top? What precedent does this decision set for the kind of employer — and the kind of institution — we want to be?

These are not soft questions. They are questions with real economic consequences. Research consistently shows that organizations with high levels of employee trust and engagement outperform their peers on productivity, retention, and innovation — and that the cost of losing that trust, once it is lost, is substantial.⁴ An automation strategy that alienates a workforce may look efficient on paper for several quarters before the hidden costs become visible.

A More Durable Kind of Growth

None of this argues for resisting automation or pretending that the transition can be managed painlessly. Some displacement is real, some roles genuinely will not return, and the organizations that fail to adapt to AI will simply be outcompeted by the ones that do. The answer is not to slow down — it is to be more intentional about direction.

The most sustainable path forward, for organizations and for economies, is one that treats AI deployment and human capital investment as complements rather than substitutes. This means committing to genuine reskilling programs — not performative ones designed to satisfy public relations requirements, but substantive investment in helping people develop capabilities that remain valuable in an AI-augmented workplace. It means looking seriously at how productivity gains are distributed: whether reduced labor requirements are used to broaden economic participation through shorter hours and retained headcount, or simply extracted as margin. It means engaging honestly with the policy conversation about social safety nets and workforce transitions, rather than leaving governments to manage the fallout of private-sector decisions.

It also means accepting that the companies best positioned for the long term are not those that have stripped their operations to the minimum viable headcount, but those that have built organizations with deep institutional knowledge, genuine loyalty, and the adaptive capacity that only human creativity and commitment can provide. Resilience, in the face of ongoing technological change, is not a product of optimization. It is a product of people.

The Question Worth Asking

I am not a pessimist about artificial intelligence. The potential for these technologies to expand access to expertise, accelerate discovery, and raise living standards — particularly in parts of the world that have historically been left behind by economic growth — is real and significant. I am an optimist, in the end, about what human beings can do when they are given the conditions to flourish.

What I am less sanguine about is the assumption that those conditions will emerge automatically, or that the market will sort it out in time. They require leadership — the kind that takes a long view, that measures success in years and decades rather than quarters, and that holds the prosperity of the broader human community as part of its mandate rather than an afterthought.

The companies and the leaders who grasp that early will not just do better by the world. They will do better, full stop.

SK102 helps Guam organizations navigate AI-assisted development — building custom software with local teams who stay accountable after delivery. See how we work.