AI Can Do the Junior Work. That’s Exactly Why You Still Need Juniors.

An owner told me recently that he had not replaced his last two junior hires and did not intend to. The work they used to do, first-draft proposals, chasing numbers, tidying reports, all of it now takes his senior people about twenty minutes with AI. The maths looked obvious to him. Two salaries saved, same output, no training overhead.

I asked him where his next senior person was coming from. Long pause.

That is the whole argument in one exchange. Cutting the bottom rung of your business is the cheapest decision you will make this year and one of the most expensive over five. Because seniors are not a thing you buy when you need one. They are a thing you grow, and the growing takes years you have to start paying for now.

The Story The Big Companies Are Telling Is Not The One The Data Tells

You have read the headlines. AI does the work of a graduate, so graduate roles are finished. It has been said loudly and mostly by very large organisations, which is worth noticing, because a company of eighty thousand people has structural reasons to want a redundancy narrative that a company of forty does not.

Look at what employers actually expect and it is a different picture. Among senior talent leaders, roughly 2.7 times as many expect AI to increase entry-level hiring this year as expect it to fall. About 41% of small business leaders think AI adoption will produce a net increase in jobs across the next two years. And small firms are on track to hire close to a million graduates in this hiring season, at better starting pay than last year. IBM, of all companies, is tripling entry-level hiring while openly saying AI has rewritten what those roles do.

So the roles are not evaporating. They are being re-scoped. Which is a much harder problem than deletion, and a much better opportunity, because most of your competitors will take the easy read and quietly stop hiring.

What Actually Broke Was The Apprenticeship

Here is the bit that deserves honesty. The people saying AI does junior work are not wrong about the tasks. They are wrong about what the tasks were for.

Nobody ever hired a graduate because the business desperately needed a fifth-rate first draft. The grunt work was tuition. You built the model twelve times and somewhere around the eighth one you started to feel when a number was off before you could explain why. You sat in on the calls, wrote the notes, watched how the partner handled the awkward question. The output was almost beside the point. The repetition was quietly turning a person into someone with judgement.

AI took the repetition. It did not take the need for judgement, it raised it, because now every output arrives fast, fluent and occasionally confidently wrong. If your business has nobody learning to spot that, you have automated your production and deleted your quality control in the same quarter.

This is what I mean when I talk about amplifying intelligence. AI multiplies whatever human capability you point it at. Point it at a team with deep judgement and you get phenomenal leverage. Point it at a team where nobody has been made to develop any, and you get very fast mediocrity. The amplifier does not care which one you hand it.

The Junior Role, Rewritten

So what does the job look like now? In one sentence, they stop producing the work and start running the thing that produces it.

The clearest workforce shift happening in businesses right now is people moving from doing a task to overseeing a system that does it. Your junior runs the AI that drafts the quote, then checks it against what your business actually charges and why. They handle the exceptions the system cannot. They notice that the tone is off for this particular client. Instead of producing three proposals a week they are reviewing fifteen, which means they are seeing five times the pattern in the same period.

That is genuinely a better apprenticeship than the one it replaced, on one condition. Somebody has to make them explain their corrections. A junior who fixes the AI draft without articulating why it was wrong has learned nothing and will plateau in eight months as a fast approver of plausible rubbish. A junior who has to say “this is wrong because we never quote fixed-price on scope like that, we got burned in 2023” is building the exact thing you cannot buy.

Human in the loop is not a compliance box. It is your training programme.

You Need Their Knowledge As Much As They Need Yours

There is a trade running the other way that older leadership teams consistently undervalue.

A twenty-three year old who came up with these tools has an intuition for them your senior people are unlikely to develop by attending a webinar. Not the tool list, the instinct. Knowing when a model is drifting. Knowing to reframe rather than repeat. Knowing that the answer is worth pushing on twice. That reverse apprenticeship is real value, and it only shows up if you actually put juniors near the people making decisions instead of parking them at the edge of the org chart.

The businesses getting this right have juniors and seniors working on the same problem, in the same room, trading in both directions. The senior supplies the context and the judgement calls. The junior supplies the fluency and the speed. Neither is complete without the other, and both are learning. That is what building capability rather than buying tools looks like in practice when you get down to who sits with whom.

Hire For Judgement And Curiosity. The Rest Expires.

If you are hiring into this, the criteria have moved.

• Ask what they have built, not what they have used. “I use ChatGPT for assignments” tells you nothing. “I automated my job application tracking and it broke twice before I fixed it” tells you almost everything. The difference between a creator and a consumer shows up in about ninety seconds of conversation.
• Test whether they will disagree with a machine. Hand them an AI output with a deliberate error and watch. Some candidates will polish the grammar. The ones you want will stop and say the underlying claim looks wrong.
• Prefer curiosity over any specific tool knowledge. Whatever software they are fluent in today will be replaced inside eighteen months. The habit of pulling things apart to see how they work will not.
• Watch how they handle not knowing. AI makes it trivially easy to sound authoritative on a subject you have never touched. A junior who can say “I do not know, here is how I would find out” is worth three who can generate a confident paragraph on anything.

The Real Question Is What Your Business Looks Like In 2031

Every argument for cutting entry-level roles is a one-year argument. Save the salary, keep the output, move on. And for one year it works fine.

Then your senior estimator retires. Your best account manager gets poached. And you go to market for a replacement at the same moment every business that made the same call is doing exactly that, into a talent pool that nobody spent the last five years growing. The price of experienced people in that market is not going to be kind to you. Meanwhile the people who have all the knowledge about how your business really works are the same three people who had it in 2026, only now they are five years closer to leaving and there is still nothing written down.

Before you make another headcount decision, do this. Take your last two junior roles and write down what those people actually did all day. Cross out everything AI can now do faster. Look at what survives, and be honest that most of it is judgement, relationships, exceptions and learning. Then ask whether you were paying for the tasks or paying for the person those tasks were producing.

If it was the second one, and it usually was, then AI has not removed the reason to hire. It has removed your excuse for a boring first year.

Frequently Asked Questions

Should I still hire junior staff now that AI can do entry-level work?

Yes, but the role has to change. AI has absorbed a lot of the routine work that used to fill a junior’s first year, so hiring a junior to do that work no longer makes sense. Hiring one to direct AI, check its output, and learn your business from the inside does. The businesses cutting entry-level roles are saving a salary this year and losing their supply of experienced people by 2030, because seniors are not something you can buy at short notice. They are something you grow.

What does a junior role look like when AI does the routine work?

It looks like supervision rather than production. A junior now runs the AI that drafts the quote, the report or the first-pass analysis, then checks it against what the business actually knows and flags what looks wrong. The volume of work they get through is much higher than a junior five years ago, and the skill they build is judgement rather than typing speed. That shift from doing the task to overseeing the system is one of the clearest workforce changes happening across organisations right now.

How do juniors build real expertise if AI does the repetitive work?

By reviewing far more work than they could ever have produced by hand, and by being made to explain their corrections. The old apprenticeship worked because repetition eventually produced pattern recognition. You can get to the same place faster if a junior reviews twenty AI-generated drafts a week with a senior checking their calls, as long as someone insists they say why an output is wrong rather than just fixing it. Without that step you get a fast approver who has learned nothing.

Is AI actually reducing entry-level hiring in small business?

The headlines and the data disagree. Large companies have been loudest about AI replacing junior staff, but among senior talent leaders roughly 2.7 times as many expect AI to increase entry-level hiring in 2026 as expect it to decrease, and around 41% of small business leaders expect AI adoption to produce a net increase in jobs over the next two years. Small businesses in particular are still hiring graduates in large numbers, partly because a young person who is genuinely fluent with AI tools is now an asset rather than an overhead.

What should I look for when hiring a junior in 2026?

Look for someone who builds rather than consumes. Ask a candidate what they have made with AI, not which tools they have used, and listen for whether they have ever wired two things together, automated something small, or pushed past the first answer a model gave them. Then look for the willingness to say an output is wrong. Judgement and curiosity are the two things you cannot install later, and they matter far more than whatever software the candidate happens to know today.