What AI Is Actually Changing About Work (It's Not What You Think)

When ChatGPT launched at the end of 2022, the prevailing prediction was that white-collar jobs would be automated within 18 months. Here we are, over two years later. The reality is more nuanced, more interesting, and more important to understand clearly.

The automation wave that didn't happen (yet)

Mass job displacement hasn't materialized — at least not in the way the forecasters described. What's happening instead is task displacement within jobs. The same roles exist. The people doing them are just spending less time on certain sub-tasks.

A marketer still does marketing. But they might spend 2 fewer hours per week on drafting, an hour less on research, 30 minutes less on formatting reports. The role isn't gone — the texture of the role has changed.

What's actually being disrupted: the expertise gradient

This is the change nobody's talking about clearly enough: AI is compressing the expertise gradient in many fields. The gap in output quality between a junior and a senior AI engineer is narrowing — not because seniors are less valuable, but because AI is raising the floor.

A junior copywriter with Claude can produce first drafts that would have taken a mid-level writer to produce two years ago. A junior developer with Cursor can navigate codebases and write boilerplate at a pace that used to require years of experience.

"The question isn't "will AI take my job?" — it's "will people who use AI effectively take opportunities from people who don't?""

The unexpected consequence: decision fatigue is getting worse

Multiple AI engineers have told me a version of the same thing: AI has made them faster at producing options, but slower at making decisions. When your AI can generate 12 different versions of a headline, you now have to choose among 12 instead of 3. The output volume goes up. The cognitive load of curation also goes up.

This is the kind of second-order effect that doesn't show up in the productivity benchmarks. The hours saved on generation can be partially eaten by the hours spent evaluating.

The global adoption gap

One of the most significant dynamics I'm tracking is the differential adoption rate across economic contexts. In high-cost labor markets, AI is often used to maintain output while reducing headcount (or avoiding hiring). In lower-cost labor markets, AI is more often used to punch above weight — enabling individual AI engineers or small teams to compete with larger organizations.

This has real implications for global economic dynamics that nobody is modeling seriously yet. The person in Lagos who can now deliver enterprise-quality work solo — what does that mean for traditional outsourcing? For talent flows? For what kinds of companies can get built where?

What this means for you

The AI engineers who are thriving are the ones who figured out which of their tasks AI can accelerate and systematically off-loaded those tasks, freeing up time for the higher-judgment work that still requires a human. They're not threatened by AI. They've turned it into leverage.

The AI engineers who are struggling are the ones waiting to see what happens — hoping the disruption passes without requiring them to adapt. That strategy has a poor historical track record.