What's Left in an Entry-Level Job After AI Takes the Routine Work?

What's left is the human part: thinking on your feet, reading a room, earning a client's trust, and holding a team together when the plan breaks. AI has absorbed much of the routine drafting, summarising and data work that used to fill a graduate's first year. The role didn't disappear. It got harder, because the easy parts went first and what remains is the part that runs on instinct and connection. PowerProv builds exactly that, for graduate cohorts across Australia.

What does the research say about AI and entry-level jobs?

The Institute of Student Employers surveyed 144 UK employers over January and February. The picture it found is not the one the headlines describe.

Institute of Student Employers, Student Development Survey (144 UK employers, surveyed over January and February)

Nearly nine in ten employers expect change. Almost none expect deletion. In the ISE's words, 40% of employers don't expect to replace any entry-level roles with AI, and a further 42% expect only a small number, between 1% and 10%.

The interesting number is the middle one. Forty-three per cent said their entry-level roles had already evolved in response to AI. Not planned to. Already had.

Why "reshaped, not replaced" is the finding that matters

If entry-level roles were vanishing, the answer would be a hiring decision. Employers would simply hire fewer graduates and move on.

They aren't vanishing. They're changing composition. And that makes it a training decision, which is a much less comfortable problem, because it means the graduate program you already built is now pointed at the wrong target.

What the Australian data shows

The ISE data is British. The Australian picture, measured separately, points the same way.

Indeed Hiring Lab Australia

Indeed Hiring Lab Australia found something worth sitting with: the decline in graduate postings for AI-exposed occupations started well before AI adoption was widespread. The timing doesn't fit the story that AI is deleting Australian graduate jobs. Graduate postings in highly exposed occupations are rising again.

So the Australian question isn't whether graduates will be hired. It's whether the ones you hire can do the job that's left.

What's actually left in an entry-level role?

Take the drafting, summarising, formatting and data-pulling out of a graduate's week and what's left isn't a shorter list of tasks. It's a list of human capabilities.

None of it is knowledge a graduate can look up, and none of it is something AI can hand them. It runs on instinct and intuition built through experience, and on a layer of trust between people that no tool can install.

Here's the sting for anyone running a graduate program. The routine work AI absorbed is exactly where graduates used to serve their apprenticeship. Those small jobs were never really about the output. They were how a new person learned who to ask, how the place works, and who they could trust with a stupid question. AI took the tasks and, without anyone deciding to, took the apprenticeship with them.

“We learned to accept and embrace failure and to be receptive, and adaptive to when problems arise.”

Jeremie S. · Research Analyst★★★★★

The 43% problem: the roles changed, the programs didn't

Here's the gap that should worry anyone running a graduate program.

Forty-three per cent of employers say their entry-level roles have already evolved because of AI. The ISE describes these roles as having changed without formal redesign. Nobody sat down, mapped the new role, and rewrote the training to match. The work drifted. The program stayed put.

That means a lot of graduate programs are now spending their budget on capabilities the job no longer requires, while leaving the capabilities it does require entirely untrained.

This is the audit worth running before the next intake. What does a graduate in your organisation actually spend their days doing now? And does the program you're about to run bear any resemblance to it?

Why employers no longer trust what a graduate claims

The second finding in the ISE research is the one with the sharpest edge.

Two-thirds of employers believe graduates and apprentices are using AI to misrepresent skills. The year before, it was around half. That's a fast collapse in trust, and it lands on a specific thing: written evidence.

Cover letters, applications, take-home tasks and written assessments were never perfect signals, but they were signals. When two-thirds of employers assume the writing may not be the candidate's own, the signal stops working.

What replaces it is anything that has to happen live. How someone thinks on their feet. How they handle a question they didn't prepare for. How they behave in a group when there's no script. Those can't be outsourced to a model, which is precisely why they're becoming the currency.

This is PowerProv's read rather than an ISE finding, but the logic follows: as written evidence loses value, demonstrated live capability becomes the thing worth investing in. Both for graduates who need to show it, and for employers who need to build it.

What Australian employers are already doing about it

PwC Australia, a PowerProv client, has announced that every graduate joining its Early Careers Program will take part in a three-day immersive AI experience. The framing in its own announcement is the tell: the program is built to combine human judgement with AI, with responsible use and human oversight at the centre.

PwC's stated reasoning is that AI is reducing routine, time-intensive tasks, which creates more room for professional skills development and meaningful client work, letting graduates step into higher-value work earlier.

That's the same conclusion from the other direction. When AI clears the routine work, what's left is more demanding, more human, and needs more deliberate training, not less.

How do you train the skills AI can't take?

Not with a slide deck. Reading a room, thinking on your feet and recovering from a mistake aren't knowledge problems. You can't read your way into them any more than you can read your way into swimming.

They're practice problems. They improve when people are put in unscripted situations, repeatedly, in a setting where getting it wrong is safe and useful rather than career-limiting. That's the whole basis of PowerProv's method: improv principles applied to professional capability, delivered as structured skills training rather than performance.

Worth noting where that checklist earlier in this article came from. It isn't a description of an entry-level job. It's PowerProv's published outcomes list, the capabilities a team walks out with. The overlap with what AI left behind in a graduate role is close to total, which is the whole argument of this piece in one place.

To be clear about what that isn't. PowerProv is not a comedy workshop. Nobody is learning to be funny or perform on a stage. It's genuinely fun, and the fun is a large part of why the skills stick, but the output is capability graduates use the next morning. It's introvert-friendly by design, with no acting required.

“It's a safe, fast-paced, mind-expanding environment and gets you thinking differently about how to solve problems.”

Kamal S. · Management Consultant★★★★★

The results are measured rather than assumed. PowerProv's Personal Power Index™ tracks participants before and after each workshop, and 98% report improvement in self-confidence, collaboration, thinking on their feet, listening, and social anxiety. Those are the exact capabilities left standing in an entry-level role after AI takes the routine work.

PowerProv runs dedicated graduate cohort programs for groups of 12 or more, in person across Australia and online. It also works directly with universities through its career and employability programs, which matters when the trust problem starts before the graduate is even hired. For programs already fighting early attrition, the companion question is how to keep graduates past day 90. And for leaders asking the same question about their existing teams rather than their intake, there's a separate piece on whether AI replaces teams or makes them more valuable.

Where to start

Audit what your graduates actually do now, not what the program assumes they do. Then train the gap.

If you'd like help mapping that gap, book a discovery call with PowerProv. If you're still weighing it up, you can see how it works first. PowerProv has a 4.9-star rating, 13-plus years of work with hundreds of companies including Google, PwC, Westpac and Canva, and a money-back guarantee. You can also review participant results and reviews before committing to anything.