The Resistance Is Understandable and Costly

Every significant technological transition has produced a population of people who resist the new tool and a population who adapt. The resistors are not always wrong — some technologies fail, and early adopters bear the costs of that failure. But when the transition is large enough and durable enough, resistance has a compounding cost. The people who declined to adopt the internet in the late 1990s did not ruin their lives, but they did spend the following decade catching up to a baseline that others had already established. The transition now underway with AI is large enough and durable enough that this dynamic applies again — and the compounding cost of falling behind is accelerating because AI capabilities themselves are improving faster than most people's ability to observe them.

What Thriving With AI Actually Involves

Thriving with AI does not require enthusiasm about the technology or comfort with uncertainty — though both help. It requires developing a set of practical competencies that allow you to leverage AI outputs while preserving your own judgment. The first competency is task decomposition: the ability to break down complex goals into components and identify which components AI handles well versus which require human input. This is a meta-skill that transfers across AI systems and use cases because it is about understanding the division of labor, not any specific tool. The second is output evaluation: reading AI-generated content critically, checking factual claims, noticing when the system has misunderstood the prompt, and knowing when to accept, redirect, or discard what the system produced. Fluency in the output language is not the same as quality, and developing the habit of critical evaluation is protective against the most common AI failure modes. The third is iterative refinement: understanding that initial AI outputs are starting points, not endpoints. People who get the most from these tools treat them like junior collaborators rather than vending machines — expecting to direct, redirect, and revise before the output becomes genuinely useful.

The Research on Adaptation and Resistance

Studies tracking workforce outcomes through major technology transitions show consistent patterns. A longitudinal study from the National Bureau of Economic Research tracking workers through the adoption of enterprise software in the 1990s and 2000s found that workers who adapted to new tools within the first two years of adoption earned significantly more over the following decade than those who adapted later — even when controlling for industry, education, and initial salary. The mechanism is not just that early adopters learned a useful skill. It is that they developed a framework for approaching new tools that accelerated their adaptation to subsequent tools. Technology literacy compounds in the same way that other forms of learning compound.

A Tangent on What Resistance Protects

It is worth being honest about what resistance to AI sometimes protects. For people whose professional identity is built on skills that AI now performs — certain kinds of writing, coding, analysis, research synthesis — resistance can be a way of protecting something real: not just a livelihood but a sense of worth, mastery, and contribution. That is not irrational. The problem is that protecting the feeling of mastery over a skill is different from protecting the capacity to contribute meaningfully. The former requires keeping AI away. The latter often requires learning to deploy AI well.

The Structural Advantage of Early Movers

In most fields, early AI adopters are accumulating structural advantages that will persist. They are building workflows that multiply their output. They are developing intuitions about what AI does well and where it fails. They are establishing reputations as people who can navigate this transition effectively, which affects who gets assigned to high-visibility projects and eventually who gets promoted. Research from Harvard Business School's Technology and Operations Management unit, examining professional services firms in the early stages of AI adoption, found that individuals who integrated AI tools early became de facto internal resources on AI-augmented workflows — a role that conferred visibility and influence independent of the AI work itself.

The Adjustment That Is Not Optional

The harder truth is that in fields where AI can match average human output on the core task, doing that task at the average human level will eventually not be enough. The floor is rising. What was a differentiating skill becomes baseline. This does not mean everyone needs to become an AI expert. It means that people who can do the core work AND direct AI to amplify it will have a durable advantage over people who can only do the core work. This is uncomfortable. It is also a more honest framing than the alternative, which is to wait for the transition to resolve itself in a direction that does not require adaptation. That resolution is not coming. The transition is the steady state.