Why AI's real promise isn't productivity - it's a human reset

The world of work was already broken

Over the past twelve months I've interviewed dozens of HR and Learning leaders across sectors, and the word that keeps coming up when people describe AI adoption in their organisations is chaotic. People can't get licences signed off. Often the models they have access to at work are noticeably weaker than what they're using at home. Governance approval takes months. And even when people are adopting tools, they're not changing the way they actually work - workflows remain stagnant, just with a chatbot bolted on.

But here's what I think we need to be honest about. Even before AI arrived on the scene, we had a lot of challenges going on. The world of work was pretty broken.

Manager engagement has collapsed too - down from 30% to 27%, with the sharpest drops among young managers and female managers. And seventy percent of team engagement comes down to the manager. So the people who are supposed to be leading this transformation are themselves checked out.

We've spent decades and billions on leadership programmes. We've run culture change initiatives until the phrase itself has become a punchline. And yet here we are - four in five people showing up to work not fully engaged, many of them burned out, most of them operating like machines just trying to get through the day so they can get home.

The toll on people

And it's not just an organisational problem - it's a deeply personal one. Workplace stress has been climbing steadily since 2012. A longitudinal study across 149 countries found that the proportion of people experiencing emotional stress rose from around 28% in 2008-2011 to 36% between 2012 and 2019 - and then accelerated sharply through the pandemic years. In the UK alone, the Health and Safety Executive reports that 964,000 workers suffered from work-related stress, depression, or anxiety in 2024/25, resulting in 22.1 million working days lost. That's more than double the rate from 2001. In Australia, mental health claims have risen 97% over the past decade, with workplaces losing an estimated 20 million sick days per year to stress.

The burnout numbers are staggering. Depending on which measure you use, between 66% and 82% of workers are now reporting symptoms of burnout - with the sharpest rises among Gen Z, where 74% experience at least moderate burnout. Microsoft's 2025 Work Trend Index found that 68% of employees are struggling with the pace and volume of work, and 80% of the global workforce reports lacking the time or energy to do their job properly.

People are paying for this with their health, their relationships, and increasingly, their willingness to stay.

The working wounded

Here's what makes it worse. Many organisations aren't just failing to fix these problems - they're not even acknowledging them.

The Academy of Management Annals published a landmark review in 2022 examining over 1,500 articles on work-related psychological trauma, and found something striking: despite four decades of recognition in psychology and medicine, organisational trauma remains dramatically underexplored in management literature. We have a blind spot.

Organisational trauma doesn't just happen in dramatic events. It accumulates - through repeated restructuring, through toxic cultures left unaddressed, through layoffs that leave 53% of remaining employees reporting damaged workplace culture and psychological safety. Research shows that trauma in organisations is contagious - it spreads laterally and vertically through what psychologists call "parallel process," perpetuating cycles of negativity that organisations struggle to break free from.

Amy Edmondson's research at Harvard tells us that 43% of workers feel tense or stressed daily - but that figure jumps to 61% among those with low psychological safety. Chronic fear doesn't just make people unhappy. It siphons off the brain's resources, hindering working memory and problem-solving at the exact moment we need people thinking clearly.

We have a workforce full of what I call the "working wounded" - present, but emotionally and psychologically carrying damage that nobody is addressing. And now we're asking those same people to lead the most significant workplace transformation in a generation.

Layering technology onto a broken system

And this is where the story gets dangerous. Because what most organisations are doing right now is taking the most powerful technology we've ever had and layering it directly onto these broken human systems.

It's not a new pattern. Seventy percent of digital transformations fail. The global cost of those failures? An estimated $2.3 trillion. And the research is clear about why: the failures are overwhelmingly cultural, not technical. McKinsey has consistently found that culture - not IT budgets, not digital savviness, not the technology itself - is the single biggest barrier to successful transformation. Organisations that invest heavily in culture change see 5.3 times higher success rates than those taking a technology-only approach.

We saw it with ERP implementations, where failure rates ran between 40% and 90%. We saw it with digital transformation programmes. And we're seeing it again with AI. As one commentator put it: we're automating dysfunction. Simply adding automation into a business process without understanding why it's broken - or worse, automating a process that's already failing - amplifies the problem at organisational scale. AI didn't break this system. It turned the lights on.

The inflection point is real

Without a doubt though, we are at a technology inflection point. And it's not what most people think it is.

Most organisations are still treating AI like a faster search engine - a tool you type a question into and get an answer back. But something fundamentally different started happening in late 2025 and became mainstream in early 2026. We moved from chatbots to agents. And the difference matters.

An agent doesn't just give you information. It does the work. You instruct it, it plans and executes against the goal - researching and writing the report, processing the applications, analysing the data. It's not a conversation. It's a teammate.

Microsoft, in their Frontier Firm report, said intelligence will be "on tap" - abundant, affordable, and available on demand. Eighty-two percent of leaders say they plan to use digital labour to expand workforce capacity in the next 12-18 months. This is not a hypothetical future. This is happening now.

When AI designs something genuinely new

There's a story that illustrates what becomes possible when we stop using AI to optimise what already exists and start using it to create something entirely new.

In Birmingham, a collaboration between a precision metal parts manufacturer, an AI company, and the University of Birmingham faced what seemed like an impossible challenge: designing a wind turbine that could work in a city where average wind speeds are just 3.6 metres per second - well below the 10 m/s minimum that traditional turbines need, complicated further by the turbulence created by buildings.

Rather than tweaking an existing turbine design, they let AI start from scratch. Over just a few weeks, AI generated more than 2,000 design concepts using evolutionary design technology - mimicking natural selection to find solutions no human engineer would have reached. The result was the Birmingham Blade: a vertical-axis turbine with curved blades that is up to seven times more efficient than previous urban turbine designs.

This is a powerful metaphor for what's possible in organisations. The old models of work were designed for a different environment - like conventional turbines designed for open plains. Our workplaces have changed. The conditions are different. The turbulence is real. And the opportunity isn't to bolt AI onto the old design. It's to let AI help us design something genuinely new - work that fits the actual conditions people are operating in. But here's where I need to be direct about what's going wrong.

We are repeating the same mistake

The early evidence on what happens when people gain time back from AI is - and I don't say this lightly - alarming.

Researchers from UC Berkeley spent eight months embedded inside a 200-person tech company. Nobody was being pressured to hit new targets. But what they found was that as AI freed up people's time, their to-do lists expanded to fill every hour - and then kept going. People didn't work less. They worked faster and took on more. The boundary between work and non-work blurred. Harvard Business Review published a piece in February 2026 with a title that should give every CEO pause: "AI Doesn't Reduce Work - It Intensifies It."

We promised people that technology would free up their time and they'd have to do less. What's actually happening is that the technology truly does free up time - and then we fill it with more work, more speed, more output. We're using the most transformative technology of our generation to accelerate the very patterns that were already breaking us.

This is exactly what we should have been afraid of. And I think it's happening because we're missing the real opportunity entirely.

From doing to being

There's a framework most of us will be familiar with - Be, Do, Have. We need to be in a certain way as we do certain tasks in order to have or create certain outcomes. For most of our working lives, the emphasis has been on the Do. Our identity, our confidence, our sense of progress - all built around what we produce.

Here's what changes with AI agents: the Do is increasingly going to be done by our AI teammates.

Which means the human opportunity - the only opportunity that matters - is to focus on the Be. Orchestration. Evaluation. Strategic direction. Judgement. The wisdom that comes from slowing down rather than speeding up.

And she's right. If we're competing against machines on intelligence - on information processing, pattern recognition, speed of output - we've already lost. But machines will not become wise. Wisdom requires something different. It requires stillness. It requires emotional depth. It requires the kind of knowing that only comes from being fully human.

And here's the uncomfortable truth. Many of us struggle with the Be. We've built our entire professional identities around being busy, being productive, being the person who delivers. I was listening to a software developer recently who said he could never have imagined a time when he wouldn't be coding. He used to tell people "always learn to code and you'll always have a job." He hasn't written a line of code in three months. Some organisations have been created with the explicit rule that no human will write or check code.

These identity shifts are going to happen to a lot of us. And when they do, people will look for certainty outside of themselves - in their job title, their output, their busyness. But the certainty that will actually hold is the relationship a person has with themselves. Stillness and slowness are what's needed as the machines bring the speed. People need to bring the deep thought, the judgement, the wisdom.

Progress will no longer look like being productive yourself. That's counterintuitive and it's going to take some unpicking.

Emotions, not mindset

I've spent 23 years in the space of behaviour change, and one of the things I've learned - the hard way, over a lot of programmes that didn't land the way we hoped - is that mindset alone doesn't cut it. (And I know a thing or two about what doesn't work!)

We've thrown mindset at every organisational challenge for twenty years. Growth mindset. Agile mindset. Digital mindset. And yet mindset programmes often fail because they try to override what people are emotionally feeling. If you're using a mindset to try and override fear, or shame, or the deep uncertainty of not knowing what your role is anymore - it will be short-lived. It may even lead to burnout.

What we need to be paying attention to is the emotional landscape. Fear, guilt, shame, embarrassment - these are the emotions that will dim people's light as this transformation happens. And they're real. We need to get them onto the table where they can be discussed and worked through, not left hidden where they silently undermine every AI rollout we attempt.

Curiosity is where the big wins are going to lie. But curiosity doesn't survive under time pressure. When people are stressed, panicked, operating from a sense of shortage - they just want the quickest answer. Curiosity needs space. It needs the kind of slowing down that most organisations are currently doing the opposite of.

And here's the thing about culture. If an AI agent followed your organisation's unwritten cultural rules - the real ones, not the ones on the poster in the lobby - would you be happy with the output? There have been some appalling stories of agents behaving unethically. But often those behaviours were learned from us. We need to be intentional not only about the behaviour of our agents, but about the culture we're creating for our humans. Are people thriving, or are they showing up as machines just to get through the day?

From wellness to wellbeing - what "good" actually looks like

Josh Bersin's research offers a framework for understanding what organisations actually need to do differently. In 2021, the Josh Bersin Company published a landmark study spanning 25 dimensions and 91 practices across hundreds of organisations. What they found was a clear maturity model with four levels.

At Level 1, organisations focus on basic safety. At Level 2 - where most companies sit - wellbeing is treated as a programme, run out of the benefits department: gym memberships, employee assistance lines, mental health apps. Well-intentioned, but fundamentally disconnected from how work actually gets done.

The shift happens at Level 3, where companies move from "making you healthier" to "helping you perform better." This means flexible schedules, adequate tools, manageable workloads - wellbeing embedded in the design of work itself, not bolted on as a perk. And at Level 4 - "The Healthy Organisation" - only 15% of employers reach this: high psychological safety, employees connected to mission, a culture where people step in to help each other even outside their job description.

This is the lens through which we should be thinking about AI adoption. The question isn't whether you have a wellness programme. It's whether you've designed work - and the systems, culture, and management practices around it - so that people can actually thrive. Because if you haven't done that, AI won't save you. It'll accelerate the damage.

The agency reset

The endgame here isn't AI adoption. It's human agency.

Agency means people playing to their strengths, trusted to make decisions, connected to what lights them up. Bruce Daisley, in his books The Joy of Work and Fortitude, tells stories of organisations where customer service reps can solve a problem on the spot without escalating to a manager. That's agency. Now imagine that across every role, amplified by AI teammates handling the routine.

There are practical, tangible things organisations can do right now. The four-day work week trials in the UK have been striking - after a six-month trial across 17 companies, every single one kept the shorter week. Some moved to a nine-day fortnight instead. Productivity didn't drop. Wellbeing went up. Burnout reduced by 71%. And critically, people started thinking differently about how they use their time - not just at work, but in their lives.

Now connect that to AI. If your agents are genuinely freeing up 20-30% of someone's time, the question isn't "what more can we have them do?" The question is "what does this person need to be?" Protected thinking time. Space for curiosity. Emotional check-ins that aren't tick-box exercises. Leaders who convene rather than command.

Imagine you could increase your workforce capacity by 25% without a single additional hire. If you knew these AI teammates understood the context, pulled in the same direction, and delivered quality work - what would you have your organisation create? How much more value could you give to your customers? And what would your human employees be freed up to actually do?

That's the invitation. Not more productivity. A reset.

Tend to your lamp

I want to come back to those two travellers on the road in 1910.

The one who kept walking didn't have the full picture. She couldn't see the whole road. She had five metres of light and a willingness to move forward anyway. And that was enough.

We don't have the full picture either. We're somewhere between doom and boom, and the honest answer is that nobody knows exactly where this is heading. But waiting for sunrise - waiting for the full picture to become clear before we start moving - means other people and organisations will have crossed the town before we've stood up.

The lamp is your light. Your wisdom, your judgement, your humanity. Tend to it. If it starts to lose its brightness - if the fear or the busyness or the uncertainty is dimming it - pay attention to that. Because everything else flows from the state of your light. The quality of your decisions. The culture you create. The agency you model for the people around you.

As Arianna Huffington says - the possibility that excites her most isn't how AI can become more human. It's how AI can help humans become more human. The agents are here. The question is whether we'll use them to find our agency.

Five metres of light is enough to start walking.