The short version: A new Stanford and King’s College London study of AI and entry-level jobs across 41 countries found that companies adopting generative AI ended up with 3.3 percent more employees than similar companies that did not, and 6.7 percent more senior employees. The junior job loss that led many headlines, 2.5 percent, is not statistically significant in the paper itself. So far, AI has made experienced people more valuable rather than made anyone redundant, and the question it leaves a small business owner is who trains the next experienced person.
The paper is “How Does AI Change Labor Demand? Evidence from 41 Countries” by Bharat Chandar of Stanford and Bouke Klein Teeselink of King’s College London, dated September 20, 2026. We read the full version on September 25. It draws on 1.25 billion job postings and 154 million employment records from January 2021 through March 2026, and its most useful line for anyone who signs paychecks is in the introduction: AI adoption “changes which workers an affiliate employs more than how many.”
What did the study on AI and entry-level jobs actually find?
Bloomberg’s report on September 21 led with the shift toward senior hiring, and several outlets rounded the junior figure up to a “3%” fall in their headlines (People Matters is one). Here are the paper’s own numbers, measured in March 2026 against matched companies that had not adopted AI:
- Total employment: up 3.3 percent, which the authors call marginally significant.
- Senior employment: up 6.7 percent.
- Junior employment: down 2.5 percent, “an estimate that is not statistically significant.”
- Junior share of the workforce: down 1.9 percentage points.
Read those together and the story changes shape. The junior share fell mainly because the senior side grew, not because junior people were shown the door. The companies that adopted AI grew, and nearly all of that growth arrived at the top of the ladder while the bottom rung thinned.
The pattern travels: of 31 countries with enough data to test alone, the junior share fell in 23, significantly in seven, the United States among them. None showed a significant drop in total employment.
Why would AI make experienced workers more valuable?
The authors split AI’s effect into two parts. The first is a scale effect: AI lowers costs, the business can take on more work, and demand rises for everybody. The second is a task effect: when AI takes over tasks, demand falls for whoever used to do them, and when it makes someone’s work more valuable, demand for that person rises. On the evidence so far, they write, AI “is labor saving for junior workers and labor expanding for seniors in exposed occupations.”
A completely different data set points the same way. Stanford’s Digital Economy Lab updated its “Canaries in the Coal Mine” study on August 12 using ADP payroll records: workers aged 22 to 25 in highly AI-exposed occupations sit about 19 percent below where they would be had they kept pace with peers in less-exposed jobs. But the decline is concentrated where AI automates tasks. Where it complements workers, employment is flat or rising, and it has grown among experienced workers in occupations that lean heavily on tacit knowledge.
Tacit knowledge is the phrase to underline if you run a trades business. It is what your best tech knows about a unit from the sound it makes, the thing no manual contains. AI can draft the estimate, the follow-up email and the service report. It cannot stand in the attic. The research says the people who carry that knowledge are the ones AI makes more productive, and that is good news for most small businesses, because that knowledge is usually what they sell.
Does this study describe a small business?
Not directly, and the paper is candid about it. Its unit of observation is a foreign affiliate of a multinational company; the authors’ own example is Amazon Web Services’ operations in the United Kingdom. Worker records come mostly from LinkedIn profiles, supplied by the analytics firm Revelio Labs, which the authors note over-represent managers and professionals. And a company counts as an AI adopter only if it posted a job ad involving generative AI. In the paper’s words, “companies that use the technology without ever advertising AI jobs count as non-adopters.”
That sentence describes nearly every small business in America. A six-person plumbing company where the owner writes every estimate with ChatGPT has never posted a job ad asking for generative AI skills, so in this data set it is a non-adopter. The authors say this likely makes their estimates too small rather than too large. It also means the percentages above come from large, internationally connected employers. What carries over to a shop your size is the mechanism, not the exact numbers.
The authors also warn that an adopter outgrowing its rivals is not the same as the economy adding jobs, since part of the gain can be work won from competitors that did not adopt. For an individual business, that is a reason not to be the competitor who sat it out.
What did AI adopters do with their junior staff?
The detail most coverage skipped sits inside the junior numbers. Adopting companies did not simply hire fewer juniors; they moved them. Demand among juniors shifted away from the most AI-exposed roles toward the least-exposed ones by 2 percent, while demand among seniors shifted toward the most exposed roles by 5 percent. Put plainly: experienced people took on the work AI touches, and newer people were steered toward the work it does not.
A small business can copy that pattern on purpose. If the paperwork your apprentice or front-desk hire used to spend mornings on is now drafted by an AI tool and checked by you, the hours it frees are best spent on the part of the job that only comes from standing next to someone who knows it: the ride-along, the second pair of hands on a tricky install, the call where they hear how you calm an upset customer. It is the same split we described when we looked at which work you can safely stop watching: work with a reliable check goes to the machine, and judgment stays with people.
Who trains your next senior tech?
This is the part of the paper not to skip. The authors write that the shift away from juniors “may have implications for the pipeline of experienced workers that firms will draw on in the future,” and they cite a World Economic Forum report urging employers to keep hiring at the entry level. Every senior worker gaining ground in this data was a junior once.
A multinational can buy experience on the open market. A small business in a mid-sized town usually cannot, which is why so many owners grow their own. If AI makes an experienced worker more productive, the sensible response for a small business is not to stop hiring beginners. It is to make the beginner’s first year faster and less tedious, so the pipeline keeps running while the business grows. Anthropic’s economic scenarios pointed toward rising wages for the trades in one of its possible futures, as we found when we multiplied out its numbers. If you have not started with AI at all, our 30-day plan for using AI in a small business begins with one task and one person, which is about the right scale for this.
The study covers just over three years since ChatGPT launched, and the authors call it an early reorganization, not a settled outcome. So far, AI has been kinder to know-how than many feared. What it does for the people still acquiring that know-how depends on the employers who keep hiring beginners.
Frequently Asked Questions
Is AI eliminating entry-level jobs?
Not on the evidence in this 41-country study. Junior employment at AI-adopting companies was 2.5 percent lower than at similar companies that had not adopted, a difference the authors found not statistically significant, while the junior share of staff fell 1.9 points mainly because senior employment grew 6.7 percent. A separate Stanford study using ADP payroll data finds a sharper gap for workers aged 22 to 25, concentrated in occupations where AI automates tasks rather than complementing the worker.
Did companies that adopted AI hire more people?
Yes, modestly. Affiliates of AI-adopting companies had 3.3 percent more employees by March 2026 than matched companies that had not adopted, an estimate the authors describe as marginally significant. Most of the gain went to senior positions.
Does this research apply to small businesses?
Indirectly. The sample is foreign affiliates of multinational companies, tracked mostly through LinkedIn, and a company counts as an AI adopter only if it posted a job ad involving generative AI, so most small businesses that use AI tools would be classed as non-adopters. The mechanism, AI making experienced workers more productive and changing what junior staff do, is the part that transfers.
Should a small business stop hiring junior staff because of AI?
The research does not support that. The adopting companies kept employing juniors and moved them toward work that AI is less exposed to, and the authors warn that thinning the junior ranks may shrink the future supply of experienced workers. For a small business that grows its own talent, using AI to shorten a new hire’s learning curve protects that pipeline.
If you run a crew, has AI changed what you hand your newest hire in their first month? Tell us in the comments.
