Should a small business still hire entry-level staff now that AI can do so much of the junior work? In most cases, yes, but not for the same job. The tasks a trainee used to spend a first year on (first drafts, data entry, routine reconciliations) are the tasks AI now does in seconds. What AI cannot do is turn a 22-year-old into the experienced person you will need in five years. Keep hiring juniors, and rebuild the role around that second job.
Most owners have never had to name that second job, because it always came free with the first. It is worth naming before you freeze the next graduate hire or let the apprentice vacancy lapse, because the evidence says thousands of businesses are quietly making that call right now.
What is actually happening to entry-level jobs?
Entry-level hiring is falling fastest in the occupations where AI is most capable, and the people losing out are new starters, not the people already in post.
In the UK, the Department for Science, Innovation and Technology published a snapshot of entry-level hiring on 8 June 2026, built on LinkedIn data. Of the 38 entry-level occupations it tracked, 30 were declining and only 8 were growing. The steepest falls were accountant (down 29%), graphic designer (down 28%) and software engineer (down 27%), while sales and customer-facing roles grew. The department is careful about cause: the pattern is "consistent with AI having an impact on entry-level hiring in these roles, but this should not be considered causal evidence."
In the US, Stanford's Digital Economy Lab has tracked the same question through ADP payroll records covering millions of workers. In the August 2026 revision of "Canaries in the Coal Mine?", Erik Brynjolfsson, Bharat Chandar and Ruyu Chen report that employment of workers aged 22 to 25 in AI-exposed occupations "now stands 19% below where it would be." The gap comes mainly from fewer young people being hired, not from people being let go. And it splits by how AI touches the work: where AI substitutes for tasks, young workers lose ground; where it complements people, employment is flat or rising, especially for the experienced.
Small businesses sit inside this picture. Gusto's New Grad Hiring Report 2026, published on 30 April 2026 from payroll data on more than 500,000 US businesses, expects firms with 1 to 49 employees to hire about 974,000 new graduates this year, slightly up on last year's 962,000 after a 29% fall from the 2021 peak. Underneath that steady total, the mix is shifting. Small-business employment grew 9.6% between January 2023 and November 2025, but employment in highly AI-exposed occupations grew just 3.4%, and workers aged 22 to 28 in those occupations saw their headcount fall. Recruiter dropped out of the top 20 job titles for new graduates. Service technician and field manager gained ground.
UK smaller firms are moving the same way. A March 2026 working paper from the University of Essex's Institute for Social and Economic Research, drawing on the British Chambers of Commerce's Business Outlook Survey, found that more than half of firms now use AI, up from around a third in 2025. Among the roughly one in ten that have moved beyond off-the-shelf tools to customized systems, about a fifth report staffing reductions they attribute to AI.
Here is my reading, kept separate from what the sources found. No owner sat down and decided to stop training young people. Thousands of owners each made a sensible-looking decision about one vacancy, and together those decisions are sawing through the bottom rung of the ladder.
Why would a small business keep hiring juniors at all?
Because a junior role was never only about the junior's output: it was also how your business grew its next experienced people, and AI does nothing for that part.
Picture a trainee bookkeeper's first year. They keyed in invoices, chased missing receipts, ran the first pass of the bank reconciliation and drafted letters a manager then rewrote. The work had some value. The real return came later, when the trainee had seen enough wrong invoices to smell the next one without being told, and could take a client's call alone. That judgment was built out of the dull work. The dull work was the classroom.
AI now does a decent first pass at most of that list. Treat the junior role as a bundle of tasks and the logic says cut it. Treat it as the place where experience gets made, and cutting it means you have stopped making experience. You will find out in about five years, when a senior person retires or resigns and nobody is standing behind them.
The Gusto data hints at where this leads. Older workers in AI-exposed roles kept gaining while younger ones lost ground, which suggests employers are choosing experienced people who use AI as a tool over new entrants. Any single business can make that choice. If every business makes it, the experienced people of 2031 will not exist, and the few who do will be scarce and hard to keep.
The Second Job Rule
The Second Job Rule is how I suggest a small business decides what to do about entry-level roles now that AI can do much of the starter work: every junior job does two jobs, the work and the making of your next experienced person, and AI only takes the first. Before you cut or freeze a junior role, write down who your experienced people will be in five years and where they will learn. Then rebuild the junior role around the second job: let AI do the first draft, and give the junior the checking, the customer and a named person to learn judgment from.
It is a way of thinking I am proposing, not a research finding. It has three working parts.
The checking. Move the junior to the far side of the AI draft. Instead of producing the first version, they review it. The obvious objection is that a new starter does not yet know what wrong looks like, and that is exactly the point. The junior checks first and writes down what they changed and why. Then an experienced colleague checks the junior's check and explains what was missed. Two checks, one short conversation, and that conversation is where judgment passes from one head to another. It also answers the first question of the Loop Test, "Would that person notice a wrong result without being told to look for it?", by deliberately growing people who will.
The customer. Give juniors contact with real people earlier than you used to. The UK data shows customer-facing roles growing while desk roles shrink, and that fits: as AI takes more of the desk work, the scarce learning moves into conversations. A junior can sit in on client calls in month one and lead the simple ones within a few months. The Scarcity Rule puts it plainly: when everyone has access to the same AI, a skill matters in proportion to how much scarcer it becomes as AI improves, not how useful it is today. Framing, judgment and ownership are those scarce skills, and they are learned with customers and colleagues, not from a template.
The named person. Every junior gets one experienced colleague whose job includes teaching them, with that time blocked in the calendar. Without it, the junior asks the chatbot, gets a confident answer and learns nothing about why it is right or wrong. "Outsource the doing, not the thinking" matters more for a 22-year-old than for anyone else in the building. A trainee who outsources the thinking in year one never builds it.
What does a rebuilt junior role look like?
A rebuilt junior role hands the first draft to AI and spends the hours it frees on checking, customers and coaching, so the junior sees more of the work, not less.
Take a hypothetical 18-person accountancy practice that has hired one trainee every September for a decade. This summer a partner proposes skipping the hire, because AI now does a first pass at transaction categorization and drafts routine client emails. Under the Second Job Rule, the practice asks the five-year question first, finds that two of its three senior bookkeepers expect to retire within six years, and hires the trainee into a different job.
The old first year:
- Keying transactions and receipts for most of the week
- Running first-pass reconciliations for a manager to correct
- Drafting routine letters that someone else rewrote
- Meeting clients only after the first year
The rebuilt first year:
- Reviewing AI-categorized transactions across several client files and flagging anything that looks wrong, with a one-line reason for each flag
- A 30-minute weekly session where a senior bookkeeper goes through the flags, including the ones the trainee missed
- Sitting in on client calls from month one and leading routine ones by month four
- Writing the briefs for the practice's AI tasks, using the New Hire Test, which turns out to be a fast way to learn what a good piece of work requires
In this hypothetical, the trainee reviews far more files in a week than they could ever have keyed by hand, so they meet more of the odd cases that build judgment. The volume that used to bury juniors now teaches them, provided someone experienced is there to explain what they are seeing.
Isn't it simpler to hire experienced people instead?
Sometimes it is, and the honest answer depends on whether the role ever led anywhere.
If a junior post was pure volume with no route upward, a temporary data-entry role for example, AI may rightly absorb it, and pretending otherwise helps nobody. And if your business is too small or stretched to give a junior a named person with real time, do not hire someone into a role AI has emptied. They will learn little and leave.
For most businesses of ten people or more with a trade, a craft or a profession to pass on, though, the experienced hire is a one-off fix for a recurring problem. You can buy experience once. You cannot buy it every time someone leaves, from a market where everyone else has also stopped making it.
What I Tell Business Owners
When I ask a leadership team who will be doing their most senior jobs in five years, the room usually goes quiet. Across the business teams I work with, owners have thought hard about which tasks AI can take and hardly at all about where their people will learn.
Before AI, I spent two decades helping people learn new ways of working, and I have now trained more than 150,000 people across more than 30 countries. What I tell owners is the line on my business page: "The hard part of AI at work isn't the technology. It's helping people use it well, safely and every day. That's what an educator does." That is truer of juniors than of anyone. An experienced colleague uses AI with a career of judgment behind every check. A new starter uses it with none, which makes developing them a teaching job, not a technology one.
So I give owners three pieces of advice. Decide junior hiring in the succession conversation, not the AI conversation. Make checking the new junior job, but never a lonely one. And put one question on the table before any entry-level role is cut or frozen:
Who will be doing your most experienced people's jobs in five years, and where are they learning it now?
If nobody can answer, the vacancy you were about to close is the answer.
What should you do this week?
You can do all of this without a project, a consultant or a data team.
- List every entry-level, trainee, apprentice and graduate role in the business, filled or planned.
- For each one, split its tasks into two columns: work AI can now do a first pass at, and things the person learns by doing.
- Put the five-year question on the agenda of your next leadership meeting and write down the answer, even if the answer is "we don't know."
- Pick one junior task and turn it into a paired check this week: AI drafts, the junior reviews and notes every change, and a senior colleague spends 15 minutes on Friday going through the notes.
- Name the person who teaches each junior and block that time in both calendars.
- Do not cut or freeze any junior role until step 3 has a written answer. If the hours AI saves do come out of a junior's job, take a before number first so you know what you actually gained.
Where to go next
If you want your whole team, juniors and seniors alike, using AI in a way that builds judgment instead of skipping it, that is what my AI training for business teams is built for. If you are planning a leadership offsite or a business event on what AI means for jobs and skills, I speak on exactly this in my keynotes for business audiences. And if you want to see where your business stands first, take the free Business AI Readiness Scorecard.
Sources and further reading
- "Entry-level hiring in the UK: a snapshot", Department for Science, Innovation and Technology, 8 June 2026. gov.uk
- Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence", Stanford Digital Economy Lab, revised 12 August 2026. digitaleconomy.stanford.edu
- Tom Bowen, "New Grad Hiring Report 2026", Gusto, 30 April 2026. gusto.com
- David Bharier, Ben Etheridge and Paulo Morais, "AI adoption and workforce change in SMEs", ISER Working Paper 2026-01, University of Essex, 18 March 2026. iser.essex.ac.uk
Dan Fitzpatrick is The AI Educator: a Forbes contributor, international keynote speaker and bestselling author who has trained more than 150,000 people across 30+ countries. More about Dan.
