Watch your tasks, not your job title

What the historical pattern of technology adoption actually tells us about what's happening right now.

Listen to the long-form version or the short-form version of this piece on your commute via Spotify.

Over the past two years, I've noticed a pattern in how people talk about AI at work. Most professionals are getting two messages at the same time, and the combination creates confusion

The first: AI is going to replace a big chunk of knowledge work. The studies and reports vary, but the promises are the same. Something is coming, and you should already be getting ready.

The second: AI is a huge opportunity, but nobody can quite explain where that opportunity is for you, specifically, in the work you do every day.

Both messages feel real. Neither tells you what to do next. And that gap — high stakes, no clear direction — tends to point to a deeper problem with how most people are understanding this moment and thus, what the narrative pushes.

The AI conversation keeps mixing up three things that actually move at very different speeds:

what AI can do technically what companies are doing with it day to day what is showing up in the job market.

Most people treat these as roughly the same thing happening at the same time. They are not.

AI's technical ability has moved fast. When ChatGPT launched in November 2022, you could type a question and get a generic text answer. When it was given internet access, it would fabricate information and provide inaccurate sources. Today, the same tool can search the web in real time and provide accurate information, read and summarize a large documents, create hyperrealistic images and in some configurations complete tasks inside other applications on your behalf plus much, much more.

Things that used to require a trained specialist can now be done by a general AI tool in seconds. But here is what is also real: as of late 2025, only about 10% of American workers use AI daily at work — despite the headlines and the investment (Gallup, 2025). A Federal Reserve study found that generative AI accounts for roughly 5.7% of total working hours in the U.S. (Bick et al., 2025). More people have access to AI tools. Far fewer have actually changed how they work because of them.

The capability gap between then and now is immense. What has not kept pace is how most workplaces have actually changed because of it.

News coverage focuses on what AI can do and the decisions executives are making because of its promises. The slow, messy process of changing how organizations actually operate rarely makes headlines. That gap is where most of the confusion lives.

When personal computers arrived in offices in the 1970s and 1980s, worker productivity actually went down — dropping from over 3% annual growth in the 1960s to roughly 1% through the 1980s. The economist Robert Solow put it bluntly in 1987: "You can see the computer age everywhere but in the productivity statistics." Computers were everywhere. Results were not. The real gains showed up two decades later, between 1995 and 2005 (Brynjolfsson & Hitt, 2000).

The reason was simple. Most companies used computers to do the same old work in a slightly different way. They moved spreadsheets from paper to screen. They digitized filing cabinets. But they kept the same structures, the same processes, the same assumptions about how work should flow. The productivity gains only came when companies rebuilt their workflows around what computers actually made possible.

Office jobs tell the same story at a smaller scale. In the 1970s and 1980s, experts predicted that computers would quickly reduce the number of administrative and clerical jobs. The opposite happened at first. Those roles grew from about 12% of U.S. employment in 1950 to nearly 17% by 1980 — right in the middle of the computer boom (Autor, Levy & Murnane, 2003). The jobs did not vanish. They changed. Workers took on more complex, less repetitive tasks. Job losses came later, slowly, as the tools improved and spread further.

The same pattern shows up every time: a new technology arrives, the predictions about disruption turn out to be too dramatic too soon, perhaps executives make bold decisions, companies adapt more slowly than expected, and work shifts in ways that are hard to see clearly while they are happening.

AI is on the same path, but the timeline isn't just compressed—it’s collapsing. While the PC took twenty years to reshape the office, AI is moving into workflows in months because the infrastructure (the internet and the cloud) is already built and waiting for it. It is worth being honest, though, that this pattern has not been kind to everyone in past transitions. The workers hit hardest by task changes have usually been those in middle-skill, middle-wage jobs — administrative work, coordination roles, clerical positions — often with the least support to adapt. The transition is manageable for many professionals. It is much harder for people in roles where the work that shifts does not get replaced by anything better.

Watch Your Tasks, Not Your Job

One honest note before the practices: changing how you work is easier when your manager supports it and your workplace gives you room to try things, make mistakes, and adjust. The three practices below are harder to apply in organizations that have not created those conditions. If that describes your workplace, building that environment may be the more important first step.

The confusion most professionals feel right now is not a sign that AI is uniquely difficult to understand, or that you are behind. It is what the middle of any major technology shift feels like from the inside. And if history is a guide, the middle is always the hardest part to read clearly.

The question that matters most is not whether AI changes work. It will. The question isn't whether AI changes your work; it’s whether you are willing to let go of the repetitive 'armor' of your old job title to reclaim the high-stakes judgment that made you a professional in the first place. The map is being redrawn—are you holding the pen, or just waiting for the final print?