The Stanford Institute for Human-Centered AI released its 2026 AI Index on April 13. The headline number getting passed around is 20%. Software developers aged 22 to 25 have lost roughly one in five jobs since late 2022. That timing lines up exactly with ChatGPT reaching mainstream use. The number is real. It is also not the story.
The Number Right Next To It
In the same firms, over the same period, developers aged 30 and older saw their employment grow 6 to 12 percent.
Same companies. Same job titles. Same years. Two different outcomes separated by one variable: experience.
Employment change for developers under 25 versus developers 30+ at the same firms since late 2022.
Source: Stanford AI Index 2026 / ADP Payroll Data, April 2026What The Data Actually Shows
The AI Index draws on ADP payroll records covering millions of American workers. The underlying analysis was led by Stanford economists Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen. Their first paper, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence," published in August 2025, put the decline at 13 percent. The AI Index 2026 updates that figure to 20 percent with another year of data.
The pattern is not confined to software. Call center hiring dropped 15 percent. Similar age-based divergence appears in accounting, marketing, and customer service. The roles that were not exposed to AI, things like health aides and production supervisors and skilled trades, show steady or growing employment across every age group.
The through line: wherever AI can do the textbook version of the work, the textbook-era workers are getting compressed out. Wherever the work requires knowledge that lives in someone's head and was never written down, the workforce is fine.
Productivity Is Up. Headcount Is Down. Both Are True.
The same report shows AI is producing real gains:
Productivity improvement in customer service (14%) and software development (26%) from AI adoption.
Source: Stanford AI Index 2026, April 2026And one-third of organizations expect AI will cause workforce reductions in the next year. In AI-adopting companies over 10,000 employees, 33 percent expect headcount cuts. Only 30 percent expect growth.
This is what "AI is a net positive" looks like in the actual data. Some workers get a 26 percent productivity boost. Others get replaced. The deciding factor keeps coming back to one thing.
Why Experience Wins
Brynjolfsson put it plainly in a Fortune interview last August:
"Older workers have a lot of tacit knowledge because they learn tricks of trade from experience that may never be written down anywhere. They have knowledge that's not in the LLMs, so they're not being replaced."
Tacit knowledge is the stuff you cannot transfer through a training document. The sense that a client is about to churn before they say so. The instinct that a deal is off before you can explain why. The muscle memory of knowing which person to call when a supplier ghosts you.
A language model can read every manual you ever wrote. It cannot read the thing you never wrote down because you did not know you knew it.
What This Means For A 47-Year-Old Executive
If you are 35 to 60 and have been quietly worried that AI makes your twenty years of experience irrelevant, the payroll data says something specific:
The opposite is happening.
Your experience is becoming more valuable, not less. It is the one asset in your career that AI systematically cannot copy. The workers getting displaced are the ones whose work is most easily codified. The workers getting more productive are the ones whose work sits on top of twenty years of judgment.
But there is a condition.
Experience stays a moat only if you can direct AI with it. Experience alone, with no AI literacy, is still at risk over a ten-year horizon. Experience plus AI literacy is defensible for the length of most careers.
The Index also reports a quieter number: nearly 40 percent of AI time savings in organizations get lost to rework. Companies investing in skills and learning see materially better outcomes. Translation: the AI gains go to the people who learned to operate it, not the ones who bought it.
What Indiana Professionals Should Do This Week
Indiana's economy is built on tacit-knowledge industries. Manufacturing. Logistics. Insurance. Healthcare. Legal services. Agriculture. Every one of these is a field where the best workers know things the textbooks do not capture.
That is structural protection against AI displacement.
It is also a closing window. The protection holds while experienced workers learn to use AI. It does not hold if they skip the learning.
Three concrete steps:
Measure where you actually are. Guessing at your AI proficiency is easy. Getting a real baseline takes about ten minutes. Take the 7 Levels of AI Proficiency assessment.
Direct AI, do not delegate to it. The 26 percent productivity gain in software development did not come from handing projects to an AI. It came from experienced engineers using AI to move faster on work they already knew how to do.
Train the layer under you. If entry-level roles are automating, the pipeline to produce the next generation of experienced workers is broken. Teaching the 24-year-old on your team how to use AI well is now part of your job. It was not two years ago.
The Story In One Sentence
The Stanford AI Index 2026 is the first large-scale payroll evidence that experience, not youth, is the protected asset in the AI age.
There are two questions worth asking right now. "Will AI replace me?" has no answer you will like. "Do I know how to direct AI with what I already know?" has an answer, and the answer is the only thing separating the 20% decline from the 6 to 12% growth in the exact same jobs.
Related reading: Level 4: Commander.
Frequently Asked Questions
How is AI affecting workforce employment in 2026?
Stanford's 2026 AI Index, based on ADP payroll data covering millions of workers, shows AI is already moving entry-level headcount. Software developers aged 22-25 lost nearly 20% of employment since late 2022. Call center hiring dropped 15%. Similar patterns appear in accounting, marketing, and customer service. At the same time, experienced workers aged 30+ at the same firms saw employment grow 6-12%.
Which workers are most at risk from AI displacement?
Workers in roles where AI can perform the textbook version of the work are most exposed. Entry-level positions in software development, customer service, accounting, and marketing show the steepest declines. Roles requiring tacit knowledge, hands-on expertise, or judgment built from years of experience show steady or growing employment across all age groups.
Does experience protect against AI displacement?
Yes, according to the Stanford 2026 AI Index data. Experienced workers hold tacit knowledge that AI cannot replicate because it was never written down. Stanford economist Erik Brynjolfsson notes that older workers have learned tricks of the trade from experience that may never appear in training data. However, experience alone is not sufficient. Experience combined with AI literacy is the combination the payroll data rewards.
Sources
- Stanford HAI. "Inside the AI Index: 12 Takeaways from the 2026 Report" (April 13, 2026)
- IEEE Spectrum. "Stanford's AI Index for 2026 Shows the State of AI" (April 13, 2026)
- Brynjolfsson, Chandar, Chen. "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence" (Stanford Digital Economy Lab, August 2025)
- Fortune. "First-of-its-kind Stanford study says AI is starting to have a 'significant and disproportionate impact' on entry-level workers" (August 26, 2025)
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes, research-backed scoring.