The work was the training

31 August 2026 | 5 minute read

Several of the major professional services firms, EY and KPMG among them, are asking junior staff back into the office, framed as a response to AI. That framing undersells what's actually going on. As AI takes over the entry-level work, it may also be taking over the way people learned to lead. This edition is about what I think disappears when the training ground goes, and why the answer isn't the responsibility of HR...leaders I'm looking at you 👀

Several of the big professional services firms are rethinking how their early career (junior in their language) people work and learn. EY's UK consulting leadership has been telling partners to remind junior staff what the office is actually for, arguing that a life set up around working from home "is just not the route to success in the world of AI." KPMG is experimenting with new in-person training formats to build soft skills. The public conversation has mostly landed on office attendance but that's not really the story.

The story is what junior people have always built their capability from: formal training, yes, but mostly routine work, observation, feedback and steadily increasing responsibility. They prepared the analysis, observed what was said and done by others in the meetings they sat in; and saw how their 1-up's handled prioritisation, disagreement, and everything in between. They saw when a technically correct recommendation was rejected anyway, and learned why. AI can now do a good share of that junior work faster. And hybrid working arrangements have thinned out the informal observation that used to fill the rest.

Put those two things together and you get a real strategic question: if you remove the work people learned from, how do they build the judgement their next role requires?

You can't assume AI will answer that. You also can't hand it off to HR as a people policy or capability challenge to solve. You need to be part of identifying the challenge and developing the solutions. Spotting talent and shaping it for what's coming is leadership work. And too many of you have gone soft on it, not from laziness in the ordinary sense, but because you've let talent development become something you endorse in principle and never examine in practice.

1. Removing the task removes the lesson

Routine work was never exciting. It was where people learned what good evidence looks like, where a pattern started to feel familiar and then used data to confirmed it, where a mistake was small enough to be overcome yet useful enough to remember. Slowly, that work built the grounding to know when an output is incomplete, misleading or just wrong.

Take the task away and you don't only lose a line item. You lose the repetition that built the judgement underneath it. I'm not arguing we maintain inefficient work, but it should make us pause; do you know what the task was building and was there other value in the process, not immediately obvious?

Try this: pick one early-career role on your team. List what AI will likely take over in the next two years. For each task, ask your people what they actually learned by doing it, and where that learning comes from once the task is gone.

Insight: removing the task removes the lesson.

2. Proximity isn't development, exposure is

People learned by watching, in real life situations. Situations like how you challenged a poor assumption without humiliating whoever raised it, or how you changed your position when you had access to new information and evidence, how you explain real uncertainty or how you took accountability out loud when an outcome went badly. None of that comes from a training session or module. It comes from repeated exposure, with reflection and feedback attached.

Ordering people back to the office won't restore this by itself. A junior person can sit five metres from you five days a week and learn almost nothing, if they're excluded from the conversations that matter or you never explain the judgement behind what just happened.

Try this: look at next month's calendar. Pick one meeting, stakeholder conversation or decision that could become a genuine development opportunity for someone in your team. Invite them for a stated reason, tell them what to watch for, and debrief them well afterwards.

Insight: proximity isn't development but exposure is.

3. AI may raise the value of judgement while cutting the road that builds it

Here's the contradiction that stuck with me and might with you. As AI gets more capable, organisations say they'll value critical thinking, influence, empathy and judgement more, not less. At the same time, some of those same organisations may be dismantling the early work, the close observation and the gradual responsibility that built those capabilities in the first place.

You don't close that gap by adding "critical thinking" to a capability framework, and you don't close it by waiting until someone reaches a senior role to discover they've never had to exercise judgement under real pressure. Your workforce plan, if it only asks what skills you'll need without asking what experiences build them, is a shortfall in planning.

Try this: for every capability your workforce plan calls future-critical, name the experience that develops it. Not a course, but something on the job that enables the person to try, test and build the capability in real time.

Insight: don't just plan for skills. Plan for the on-the-job experience that builds them.

4. This isn't HR's problem to carry alone

Your people function are the experts in designing the framework, the pathway, and the program, but it is not their accountability to notice, in real time, who's capable of more, that's for you to own. For example whose technical skill has outgrown their exposure, who's showing judgement without formal authority, who is craving challenge or where you have been consistently hiring in capability that you should have grown within. All of that 'noticing' is your leadership work.

You find time to review budgets and interrogate risk, because those are recognised as core to your role, but somehow, talent development often gets a different treatment: raised at the annual review, after someone resigns, or once succession risk is too uncomfortable to ignore. If this is you, you're not dedicating enough or the right time to strategically considering your talent.

Try this: for each person who reports to you, answer three questions honestly.

1. What could they be contributing in two years that they aren't today?

2. What experience do they need next?

3. What are you personally doing to help them get it?

Insight: you are responsible for your team's talent plan, and your personal investment in terms of interest and time should reflect that.

5. "AI will free people for higher-value work" is not a plan

It shows up in almost every AI transformation deck, and sometimes it's true. But higher-value work doesn't appear automatically once routine work disappears; and people aren't automatically ready to do it. Moving someone from preparing information to influencing a stakeholder, or holding a difficult decision, takes more than freed-up hours. It takes capability, context, trust, and the authority to use all three.

Before you claim AI will free people for higher-value work, you should be able to say what that work actually is, who will do it, what it requires, how people will build the capability, how it'll be judged, and whether you are prepared to hand over the authority that comes with it.

Try this: next time an AI proposal promises to free up capacity, ask the business case to say exactly where that capacity specifically goes: which work, which people, what new expectation, what investment is needed to get them there.

Insight: freed time only becomes improved capability if you decide it will.

A question to sit with this month

Which of your senior people built their judgement doing work you're about to automate for the next generation, and what's replacing it?

If this raised a question about how you are building leadership judgement through AI-enabled change, I'd be genuinely interested to hear what you're seeing. And if someone in your network is grappling with the same thing, send this one on.

Source: [Financial Times, 27 August 2026](https://www.ft.com/content/7fd9c234-a92b-4ab2-ba1f-969cf9a23f52); corroborated by [Personnel Today](https://www.personneltoday.com/hr/ey-backs-office-return-so-juniors-can-build-human-skills/) and [BM Magazine](https://bmmagazine.co.uk/news/ey-return-to-office-ai-skills/) on EY's and KPMG's positions.

Key Takeaways

• Removing a task removes what it taught, so know what a task is building before you automate it.

• Proximity isn't development, exposure followed by reflection is, and an office mandate won't manufacture that on its own.

• A skills plan without a named experience behind each skill is a wish.

• Noticing and growing talent is leadership work, not an annual people exercise.

• "AI will free people for higher-value work" needs a business case behind it, not an assumption.

Frequently Asked Questions

Isn't this just an argument against automating anything?

No. Automate the task, just don't assume the capability it was building gets automated along with it. Those are two separate decisions.

We can't keep junior staff doing inefficient work purely so they learn. What's the actual alternative? You don't need to preserve the task, you need to preserve what it was teaching. That might mean a shorter, deliberately designed version of the work, a shadowing structure, or a debrief built into a process that used to teach by accident. The point isn't to slow AI down. It's to stop assuming the learning survives the automation by default.

How do I know if I'm genuinely losing development value, or just being sentimental about how things used to be done?

Ask what specific judgement a task was building, not whether the task itself was enjoyable or traditional. If you can't name the judgement, you're probably being sentimental. If you can name it and can't point to where it will now be built, you've found the gap.

This sounds like more work for me personally. I don't have time to mentor everyone.

You don't need to mentor everyone. You need to notice, decide and create the conditions for people to grow, which is different from carrying their development yourself. Most of what's described here is minutes, not hours: an invitation to a meeting, a five-minute debrief.

About Louise

About Louise: Louise Zawada is a Perth-based leadership coach, change consultant and facilitator with more than twenty years helping senior leaders and executives lead through complex change.

She's the leadership expert senior leaders call when change is relentless, and you can no longer afford to look like you don't have it handled.

Her flagship program, the Confident Change Leader, is a six-month one-to-one program for senior leaders feeling the pressure of complex change.

Louise is the creator of the Leadership Friction framework and writes regularly on executive judgement, organisational legibility and the behavioural evidence that determines whether strategy becomes action.

If you're leading significant change and need a trusted thinking partner, connect with Louise or book a conversation.