AI Doesn't Automatically Reduce Your Workload. It Changes It.

A new piece of research written up in Harvard Business Review looked at what happened when a group of employees adopted generative AI tools over eight months.

The headline finding was uncomfortable:

AI didn't reduce workload. It intensified it.

Not because the AI didn't work. It worked brilliantly. People got faster, took on broader tasks, and expanded what they thought was possible.

But instead of banking that time back, they filled it with more work. Breaks disappeared. Hours crept up. Stress followed.

• AI Doesn't Reduce Work—It Intensifies It, by Aruna Ranganathan and Xingqi Maggie Ye (2026), Harvard Business Review: https://hbr.org/2026/02/ai-doesnt-reduce-work-it-intensifies-it

Most teachers aren't at this point yet. Adoption of AI in schools is still patchy: a few enthusiastic early adopters, plenty of curiosity, and a large number of teachers who haven't really started.

If that's you, this isn't a warning about something you're doing wrong. It's a heads-up about something worth thinking about before you get there.

Because I am there. And I recognise the pattern immediately.

What the research found

Researchers at UC Berkeley's Haas School of Business followed a US-based tech company of around 200 employees as generative AI tools spread through day-to-day work. Adoption was voluntary, not mandated.

Three things happened.

1. People got more productive.
   Tasks that used to take hours took minutes. People also started doing things outside their usual role, because AI made it feel achievable.

2. People took on more, without being asked.
   Instead of using time saved to rest or slow down, they expanded what they did. They filled breaks with prompting. They extended their hours. Work expanded to fill the new capacity.

3. Stress and fatigue rose.
   The initial buzz of productivity gave way to something less sustainable: cognitive strain, poorer rest, and a higher burnout risk.

There's also a readable, non-academic write-up of the broader theme here:

• Business Insider: https://www.businessinsider.com/is-ai-making-jobs-more-intense-study-2026-2

The researchers aren't anti-AI. Their point is simpler (and more useful):

Productivity gains don't automatically become wellbeing gains.
That conversion takes deliberate choices.

Why this matters for teaching, even if you're not using AI much yet

Teaching is already one of the most workload-intensive professions going. And it's not just anecdote.

In the Department for Education's Working lives of teachers and leaders survey, only 17% of teachers and leaders agreed with the statement: "I have an acceptable workload."

• DfE (2024): https://www.gov.uk/government/publications/working-lives-of-teachers-and-leaders-wave-2/working-lives-of-teachers-and-leaders-wave-2-summary-report

So when research suggests "AI can make workload worse, not better", it's worth paying attention. Not to be fearful, but to be prepared.

Here's why I think teaching is especially vulnerable to this pattern as AI adoption grows.

The work is never "done"

In many jobs, there's a finite task list. In teaching, there's always another resource you could make, another student you could support, another email you could reply to properly, another intervention you could design.

AI doesn't create a boundary. It removes one.

When marking takes less time, the mental space it frees up gets immediately colonised by everything else on the to-do list.

Teachers are driven by care, not KPIs

In the study, people took on more work partly because it felt rewarding and feasible.

Teachers do this constantly. Not because a line manager told them to, but because they genuinely want to do right by their students.

That instinct is what makes great teachers great. It's also what makes them vulnerable to workload creep when a new tool suddenly makes more things feel possible.

There's no one tracking the recovered time

If AI halves your marking load on a Friday evening, nobody's going to say: "Great. Take Saturday morning off."

You're more likely to plan Monday instead, tidy the seating plan, rewrite the starter, update the tracker, and reply to three emails you've been avoiding.

Most school leaders aren't tracking how teachers spend time saved by technology, and most teachers aren't tracking it themselves.

What I've noticed in my own work

I should be honest here. I build an AI marking platform, and I use AI tools constantly in my teaching and business work. I'm exactly the kind of person this research is describing.

I've caught myself filling every gap.

An AI tool finishes a task in ten minutes that would have taken an hour. Instead of taking fifty minutes back, I immediately start the next thing. Breaks get shorter. The to-do list doesn't shrink, it just cycles faster.

I end the day having done more than ever, but not feeling any less drained.

One of the research findings that really hit home was that break time was one of the first casualties.

I've done this: sat down with a coffee that was supposed to be a rest, and ended up refining prompts or generating resources because it felt productive and easy.

The cognitive rest never happened. The afternoon was worse for it.

That doesn't make AI "a problem". It makes it a tool that needs using with intention.

A few habits worth building early

If you're just starting to explore AI tools (or your school is beginning to think about this more seriously), the research points to a few habits that are easier to build now than to retrofit later.

Notice where the time goes

If an AI tool saves you an hour, pay attention to what happens to that hour.

• If it goes into something that energises you, like a creative lesson, a proper conversation with a student, or leaving on time, brilliant.
• If it silently gets absorbed into more admin, that's worth noticing.

Resist the scope creep

Just because AI makes it possible to generate a twenty-slide revision pack for every topic doesn't mean you should.

In teaching, the gap between "could" and "should" needs to be quite wide.

Scope creep also shows up in subtler ways:

• Feedback gets longer because it can.
• Resources multiply into three versions "just in case".
• Emails become more polished, which makes them take longer to write overall.

Protect the breaks

For teachers, this might look like spending every free period refining AI-generated materials instead of having five minutes with a cup of tea.

But your afternoon lessons are better when your brain has had a rest. That's not laziness. It's how cognition works.

Here's a simple rule I'm trying to stick to:

The "bank it" rule: if AI saves me 30 minutes, I deliberately spend at least 10 of those minutes on genuine rest (or finishing earlier), not more tasks.

It sounds almost childish written down, but without a rule like that, the time just disappears.

Have the conversation at department level

The researchers' argument is basically this: collective boundaries beat individual willpower.

In schools, that might mean a department agreeing:

• "We'll use AI for marking mocks, and we'll use the saved time for X."
• "We're not using AI to create extra resources for every lesson, because that's a one-way ticket to more workload."

Even a short "stop list" can help. Not because AI is dangerous, but because endless expansion is.

The bigger picture

AI is going to change teaching. It's already starting to.

And most of those changes will be genuinely positive: faster feedback, less repetitive admin, and more time for the parts of the job that actually matter.

But this research is a useful reminder that "more productive" and "less overworked" aren't the same thing.

In a profession that already struggles with unsustainable workload, that distinction really matters.

The teachers who get the most out of AI won't be the ones who use it to do more.

They'll be the ones who use it to do enough, and then stop.

I'm still working on that last part myself.

Over to you

Have you noticed this?

• If you're using AI tools in your teaching, has it actually reduced your workload, or just changed the shape of it?
• If you haven't started yet, does this change how you're thinking about it?

I'd genuinely love to hear, because I think this is a conversation we need to have before it becomes a problem rather than after.

---

Gary Roebuck is an A Level Economics teacher and founder of TeachEdge, an AI-powered marking and feedback platform built by teachers, for teachers.