Over the past year, every interview, panel, podcast and off-the-record conversation I’ve had eventually lands in the same place:
“What is AI going to do to jobs?”
Not in some abstract policy sense. In a personal sense. People want to know about their job, their team, their company, their kids’ future.
So I did what I always do when the signal gets buried under noise. I dug in. I read as much serious research on AI and employment as I could stand — academic papers, policy reports, real usage data, and economic analyses. Then I layered that with what I’m seeing firsthand at Techstrong and what executives, practitioners and founders are telling me privately.
This is about as close to a 360-degree view as you can get right now.
And it’s still fuzzy.
There is no doubt AI will reshape work. But how, how fast and for whom is far from settled. The research is extensive. The conclusions are not.
We Are Watching the Opening Moves, Not the Endgame
The Hamilton Project’s assessment is refreshingly blunt: Research on AI’s labor impact is “still in the first inning.” Most studies measure short-term signals — hiring trends, task exposure, productivity experiments — not long-term structural change.
Different methodologies are answering different questions:
- Which jobs could be affected
- Which jobs are actually using AI
- Whether AI substitutes for labor or complements it
- How companies respond to productivity gains
- How workers transition over time
AI capabilities are moving faster than labor markets, education systems and corporate structures can adapt. Forecasting decades of change from a few years of data is inherently uncertain.
The First Real Warning Sign: Entry-Level Knowledge Work
The strongest empirical signal so far comes from Stanford’s Digital Economy Lab study, Canaries in the Coal Mine?
Using high-frequency payroll data covering millions of workers, researchers found that young workers in highly AI-exposed occupations are already losing ground.
- Ages 22–25 in exposed jobs saw roughly a 16% relative employment decline
- Experienced workers remained stable or grew
- Effects concentrate where AI automates tasks
- Adjustments show up in hiring, not wages
Entry-Level Employment Shock
Data adapted from Stanford Digital Economy Lab analysis of ADP payroll data.
Two occupations illustrate the pattern clearly: Software development and customer service. Entry-level hiring drops sharply after late 2022 — roughly the post-ChatGPT inflection point — while mid-career and senior employment continue rising.
The explanation offered in the paper rings true. AI substitutes for codified knowledge — structured tasks junior workers traditionally perform — but struggles to replace tacit knowledge built through experience.
Fewer apprenticeships mean fewer pathways into the profession.
Automation vs Augmentation: The Fork in the Road
Not all AI adoption produces the same labor outcomes. The Stanford analysis cross-references real usage data from Anthropic to distinguish between automation and augmentation.
Automation vs Augmentation Outcomes
Data adapted from Stanford Digital Economy Lab and Anthropic Economic Research.
The pattern is consistent:
- Where AI automates tasks, entry-level employment declines
- Where AI augments workers, employment remains stable or grows
AI itself is not inherently job-destroying or job-creating. Its impact depends on how organizations deploy it and how work is redesigned around it.
Exposure Does Not Equal Displacement
Several studies converge on a finding that rarely survives headline writing: Being exposed to AI does not mean being replaced by AI.
Brookings research estimates that about 37 million U.S. workers are in highly exposed occupations, but most are likely able to adapt successfully. A much smaller group faces high risk due to limited mobility, skills or financial resilience.
Adaptability depends on:
- Transferable skills
- Education and training access
- Financial cushion
- Age and career stage
- Local job markets
- Willingness to reskill
Who is Actually at Risk?
Data adapted from Brookings Institution estimates of AI exposure and adaptability.
The NBER analysis reinforces this point: Outcomes vary dramatically across individuals within the same occupation. Two people with identical job titles can face completely different futures.
History Suggests Transformation, Not Collapse
The Economic Innovation Group’s analysis places current fears in a historical context. Previous technological waves displaced tasks and occupations but ultimately expanded productivity and created new work.
Generative AI is unusual because it targets cognitive tasks long considered automation-proof. This time, the disruption is hitting office work before factory floors.
The Constraint Few People Talk About: Physical Labor
While white-collar disruption dominates headlines, the physical side of the AI economy tells a different story.
Ford CEO Jim Farley has warned that the U.S. lacks enough skilled tradespeople to build the data centers and manufacturing facilities required for large-scale AI deployment.
We may automate knowledge work before we can scale the infrastructure needed to support that automation.
White-Collar Now, Physical Jobs Later
Synthesis based on multiple studies of AI adoption and labor market effects.
For now:
- Digital AI disrupts cognitive work first
- Physical AI lags behind
- Skilled trades may face shortages rather than displacement
What a True 360-Degree View Suggests
Across all these sources, several patterns emerge:
Entry-level knowledge work faces the earliest pressure.
Routine cognitive tasks are easiest to automate.
Experienced workers often gain leverage.
AI amplifies judgment and domain expertise.
Automation precedes augmentation in many contexts.
Especially where work is standardized.
Employment effects appear before wage effects.
Hiring slows before salaries fall.
Individual adaptability drives outcomes.
Mindset and behavior matter as much as job category.
Physical work remains harder to automate — for now.
Shimmy’s Take
Here’s where the research stops and the real world begins.
AI is not arriving as a single event. It’s arriving as a multiplier. It amplifies capability, competence and initiative. It also amplifies mediocrity and inertia.
Two people in the same role, at the same company, with access to the same tools, can diverge rapidly.
One experiments, learns how to use AI effectively, integrates it into daily workflows and rethinks how work gets done. That person becomes dramatically more productive.
The other waits for direction, treats AI as a novelty or a threat and keeps working the old way.
Over time, that gap compounds.
What I’m hearing from leaders across industries is not “AI is replacing everyone.” It’s “our strongest people are becoming far more effective, and we need fewer entry-level staff to support them.”
That aligns almost perfectly with the research showing declining junior hiring alongside stable or rising senior employment.
Survivability, in practice, comes down to behavior more than exposure scores. Embracing AI is not optional because the environment is already changing around you. The people who lean into it gain leverage. The people who resist it lose relevance — not because AI directly replaces them, but because someone else becomes capable of doing far more.
Physical jobs will face their own wave eventually when robotics catches up, but that timeline appears longer. For now, cognitive work is the frontline.
The opportunity hidden inside this disruption is that AI lowers the cost of competence. Individuals can accomplish what previously required teams, budgets or specialized support. That creates enormous upside for people willing to rethink how they work.
None of this guarantees stability. None of it guarantees catastrophe either.
It does guarantee change.
If there is one conclusion that cuts through all the studies, charts and forecasts, it’s this:
Your mileage may vary.
But standing still is not a strategy. Embracing the tools, building new skills and using AI to extend your capabilities is the clearest path not just to survival, but to thriving in the opportunities this transition will create.
Because the future of work is not being decided solely by technology.
It is being decided by how people choose to use it.
