That was entirely the point, the purpose, and it’s also likely why most of us chose this career.  But perhaps most importantly, that purpose - technology in service of human elevation - ought to be the only standard against which any progress is measured.

But somewhere between that purpose and this moment, we lost the thread.

Look around. It seems almost every day we read about the failure and wasted expense of another gargantuan IT project that failed to materialise, never mind deliver.  A 2026 survey found that 93% of organisations are already deploying or planning AI - yet only 23% of the people who will use it actually trust it. Hell, outside our industry, “normal” people even have a simple refrain - “computer says no” - that has become the mantra of disillusionment, uttered under the breath of anyone that has been forced to endure the inflexibility and un-intuitiveness of a system that was most likely designed specifically for them, just not designed to be used by them.

These are not technology failures. They are failures of the question being asked.

We keep asking how do we manage this? How do we control it, restrict it, govern it, protect ourselves from it?  When in fact, the questions we should be asking are – “what is this actually for?”, and “who do we want to become in relationship with it?”

This was not the future I signed up for when I started a lifetime career in technology. And I suspect, if you're honest, neither did you.

What if we're asking the wrong question?

Two years after the end of the First World War, a Czech playwright cast a die that would shape our relationship with machines for the next hundred years.

His name was Karel Čapek. The word he gave us was robot. It came from “robota” - the Czech and Slovak word for forced labour, drudgery, servitude. From its first utterance the relationship between humans and machines was encoded not as partnership, not as collaboration, but as the relationship between owner and slave.

We inherited that framing. We built an entire industry on top of it. And we have never once stopped to question it.

Every system we deployed. Every architecture we built. Every governance framework we constructed. And now every AI Agent we plan to implement. All of it resting on a foundation laid in a shadow cast by a playwright's imagination in 1920. The machine exists to serve, you command and it obeys.

Until now.

Because something is happening that the slave frame can no longer hold. AI is no longer just thinking, it is acting, autonomously, on our behalf. Making decisions, completing transactions, pursuing goals – all without waiting to be asked.

When AI was a tool you commanded, treating it as a slave was merely suboptimal. Now that AI is an agent acting on your behalf, treating it as a slave is downright dangerous.

So now we face a choice. Continue as we are, applying a hundred-year-old assumption to a categorically new kind of relationship or we can reframe – deliberately and crucially before the opportunity passes.

This is that argument.

Part One - We inherited a broken model

We work like Victorians. We just use modern tools to do it.

Frederick Winslow Taylor arrived at the end of the 19th century and gave us efficiency as the governing metric of all human progress. Output divided by input. The clock as the measure of all things. He wasn't wrong for his time - the industrial revolution was chaotic and genuinely needed the discipline he brought. His original ambition was even noble: he believed that through increasing productivity it would be possible to give the worker a decent livelihood and free them from poverty.

But his legacy, the worship of efficiency over effectiveness, process over outcome, the clock over the purpose has never been updated. We still measure knowledge workers the way Taylor measured factory hands.

The production line made efficiency make sense. Break the work into its smallest components. Get each person brilliant at their single task. Here I am making widgets -  that’s my job, you pay me on how many widgets I make and how good they are.  That’s all I care about as a result. The car coming off the line can be terrible, but as long as my widgets are good and on time, the system reports success and I am oblivious to the overall failure, disconnected entirely from the overall outcome my work was supposed to serve.

We carried that model into the knowledge economy unchanged. Then into the digital economy unchanged. And now here we are deploying what is probably the most powerful cognitive technology in human history in service of a 19th century objective.

Peter Drucker made famous the distinction that matters: “Efficiency is about doing things right. Effectiveness is about doing the right things.” We have been answering the first question obsessively for a hundred and fifty years. We have barely started on the second - we are still building systems that measure everything “except that which makes life worthwhile”. We are now just doing it faster.

The management model that never changed

We still pay people for their time. Not their impact, not their outcomes, not the value they create - their time. That single idea, forged in the fires of the Industrial Revolution, is the invisible hand shaping almost every AI decision being made in organisations right now.

Our failed experiment with flexible working showed us exactly what happens when you try to manage a new way of working with Victorian metrics. As we emerged bleary-eyed from the pandemic, slowly the "back to the office / the office is where real work gets done" memos started to flow.  Presence once again became proxy for productivity - if I can’t see Dave, how do I know he’s working?  Hours logged (and in some cases even keyboard strokes) became the measure of contribution. Most organisations got it wrong - not because their people weren't working, but because the management model couldn't see work that didn't look like the old work.

Digital transformation fails for the same reason. New technology, old metrics, efficiency gains dressed up as transformation and all the while never quite delivering the change that was promised in the first place.

Agentic AI is the same trap but at ten times the scale. And right now most organisations are walking straight into it - measuring agents by the hours they save and the headcount they displace, applying the Victorian equation to what is likely going to be the most transformative technology in human history, and calling it progress.

But here is where it gets personal. Because the management model failure is not IT's problem alone to solve. There is no point in the IT leader measuring outcomes if they themselves are being measured by uptime statistics and ticket closure rates. There is no point building agents that chase value if the board is still rewarding cost reduction.

The transformation has to happen at every level of the organisation simultaneously. Not just inside IT - everywhere. And here is the uncomfortable truth: that conversation will not start itself. The board will not spontaneously decide to measure value over cost, especially in hard times and we have been in an extended period of hard times. The organisation will not voluntarily redesign its metrics around outcomes rather than time. Someone has to walk into that room and make the case, in the language of growth and value rather than cost and risk, before the agents are deployed rather than after they disappoint.

That conversation is the one nobody is having. And it starts with a single decision that every organisation is about to make - mostly without realising it.

Cost centre or growth engine?

Every day brings new stories about the incredible time savings being granted by AI. What I find incredible is that most of the stories stop there, never once asking what the time saved made possible. Which brings us right back to whether we’re doing things right or doing the right things.

I call it the “opportunity cost of automation”. Your measure of success should never be how much you have automated, it must be what you chose to do with the time it gave back.

Most organisations are not asking that question. They are banking the efficiency saving and calling it transformation. They are applying the Victorian equation to the most powerful cognitive technology ever built and wondering why the value story is harder than the vendors promised.

The question is not whether to automate. It is what you do with the capacity automation creates - and what you lose when you automate the judgment rather than the task. Chase value not cost. That is not a philosophical preference. It is the only deployment strategy that justifies the investment.

That choice, between banking the saving and building the value, is ultimately a question about what kind of organisation you want to be. And every organisation in the world is about to answer it, mostly without realising it is a decision.  They will either deploy agents in service of the Victorian equation - faster, cheaper, more efficient - and call it transformation. Or they will ask the harder question: what does this technology make possible that wasn't possible before, and how do we build an organisation around that answer?  Suddenly, AI is as much an HR / people issue as it is a technology / IT one.

The difference between those two paths is not the technology. It is whether the IT function is positioned as a cost centre or a growth engine.

A cost centre manages the plumbing. It keeps the lights on, governs the risk, closes the tickets, reports the uptime. It is necessary and it is not enough. Not anymore. Not in a world where the agents are already acting, where the gap between the organisations that ask the right question and the ones that don't will compound every quarter.

A growth engine does something different. It looks at what the technology makes possible and asks what the organisation could become. It brings insight the business couldn't generate alone - not because it is cleverer, but because it sees things from a vantage point the business doesn't occupy. It arrives at the table not as a gatekeeper but as a thinking partner. Not with a governance framework but with a vision.

That is not a technology decision. It is a leadership decision. And it is yours to make.

The bubble is going to burst

Not the one you’re thinking of.

The AI investment bubble is real and the noise is deafening. But that is not the bubble that matters most right now. The bubble that matters is this one - the human agency bubble. The one where a relatively small number of people understand what agentic AI is going to do to the world, and the vast majority of people who will be affected by it have never been asked, never been told, and have no meaningful way to prepare.

The people reading this are inside that bubble, behind the wizard’s curtain. You have seen the technology. You understand what agentic AI will do to work, to organisations, to the way decisions get made - and the scale of what becomes possible when it’s approached with the right questions. You are, in the most literal sense, the people who know what is coming.

Most of the world is not.

They are going about their lives - working, consuming, building, raising families - with no meaningful understanding of what is about to change for them. Not because they are foolish. Because nobody has shown them. Because this conversation is happening in rooms like this one, in language they were never invited to learn, about a future they have never been asked to shape.

The burst is coming. It arrives the moment agentic AI enters the lives of people who were never part of the conversation that created it. When the agent makes the decision. When the process changes overnight. When the job looks different on Monday morning from how it looked on Friday. When the technology that was debated in boardrooms and conference halls and manifesto documents arrives, without warning or explanation, in the daily reality of people who had no seat at the table.

That is not a technology failure. That is an agency failure. And it is entirely preventable.

We have been here before. Flexible working arrived with a genuine promise - that technology could free people from the constraints of time and place and unlock a fundamentally better way of working. And then we squandered it. Not because the technology failed. Because we never changed the model underneath it. The Victorian metrics stayed. Presence remained proxy for productivity. The “get back to the office” memos started flowing the moment anyone felt uncertain. And the people who had restructured their lives around a promise of trust and autonomy found themselves back at their desks, being counted.

That was not adaptation. That was failure. A failure of culture, of management, of the courage to measure something other than time. And the people who bore the cost were not organisations. They were the humans inside them - whose trust was broken, whose working lives were disrupted twice, and who learned that when things get difficult, the system defaults back to control.

Agentic AI is that failure at a scale that makes the consequences incomparably worse. Because this time the people who bear the cost are not just inside the organisation. They are everyone the organisation touches.

But preventing it requires more than good intentions. The decisions that determine whether transformation brings people with it or leaves them behind are not made in keynote sessions or manifesto documents. They are made in the metrics an organisation chooses to measure. In the questions a board chooses to ask. In whether success is defined as hours saved or value delivered. In whether the capacity that agents create is banked as a cost saving or invested in human capability.

The management model is where the moral choice becomes operational. An organisation that still measures IT on uptime and ticket closure rates - that still rewards cost reduction over value creation - will deploy agentic AI in service of the Victorian equation whether it intends to or not. Not because the people are wrong. Because the system pulls them back. Every time.

Changing that model is not a technical decision. It is not even primarily a strategic one. It is the decision that determines what kind of organisation you want to be when the bubble bursts. Whether the people on the other side of it find a world that was designed with them in mind.

The people outside the bubble are not abstractions. They are your employees. Your customers. Your families. They are the people whose working lives will be redesigned by decisions made in rooms they were never invited into. By agents deployed on briefs they never helped to write. In service of outcomes they never agreed to.

The danger isn’t ignorance. It is the assumption that the choice belongs only to the people who understand it.

You cannot demand the culture change. You cannot unilaterally redesign the metrics or force the board to measure what actually matters. But you can be the person who names it. Who reads the conditions, sees what’s coming before others feel it arrive, and walks into the room to say: here is the scale of what is coming, here is what it will cost the people outside this bubble if we deploy this technology without changing the model underneath it, and here is what we need to do differently.

This is not a technology conversation. It is a human one. And it will not happen unless someone starts it.

That someone is you. Not because you have all the answers. Not because the responsibility is yours alone to carry. But because you are the person who understands what the technology will do and what the organisation needs to become. Who can read the conditions before others feel them arrive.

You have permission to say what others won't. You have the knowledge to say it with authority. And you have the responsibility to say it now - on behalf of the people who were never invited into this conversation.

There will never be a better time for you to start it.

Part Two - The opportunity we keep missing

After thirty years working inside the boardrooms and leadership teams of businesses and government departments around the world, I have come to believe one thing above almost everything else.

The value was never in the technology. It was always in the relationship.

Here is what the right relationship looks like when it is actually working. The technologist and the business leader are peers. They bring their respective insights to the table and together they arrive at an answer neither could have reached alone. The business leader understands the customer, the culture, the competitive pressure, the human texture of the problem. The technologist understands what's possible, what's coming, what the technology will do to the problem before the business has even felt it arrive.

Neither is the gatekeeper. Neither is the supplicant. They are colleagues - in the truest sense of the word. People who think together toward a shared outcome.

And in that relationship something shifts. The IT function stops saying “no” and starts saying “yes, but”. Not as a concession. As a contribution. Yes - I understand what you're trying to achieve. But - here is how we get there in a way that actually works, that carries your people with it, that builds the capability you'll need for what comes next.

That shift - from no to yes, but - is not a communication style. It is a different understanding of what the job is.

Think of it this way. A telescope is one of the most powerful instruments for seeing what's coming. But it only works when both ends are connected - when the person looking through it understands what they're looking for, and what they're looking for is actually worth finding. The IT function has one of the most powerful instruments in any organisation for seeing what technology will do before the business feels it arrive.

Last week I sat in a room where our team presented a ten-year technology vision to the business. The immediate response was resistance. “Who are you to tell us how our business will change? We have been running this organisation for years, and you’ve just been sat in the back room fettling with the plumbing!”

We said: we don't want to own the vision. We want you to understand what's coming so that together we can navigate toward it. We want to connect both ends of the telescope - your understanding of the business with our understanding of what the technology will do to it.

Something shifted. A conversation happened that could not have happened without both perspectives in it. The business started to see what was coming. The IT team started to understand what the business actually needed. Neither could have had that conversation alone.

That is the right relationship. Not IT telling the business what to do. Not the business treating IT as the department you call when something breaks. Peers. Colleagues. People connecting both ends of the telescope in service of a shared outcome.

The expectation effect

And here is why this matters beyond the IT and business relationship. The same frame - colleague not slave, yes-but not no - governs how your agents perform. What you expect of any intelligent agent, human or machine, shapes what that agent produces. Fifty years of organisational psychology - most of it never read by the technology industry - tells us why. The relationship frame isn't just philosophically preferable. It actually changes the outcome.

This is not a philosophical position. It is one of the most robustly established findings in behavioural science, replicated across cockpits and operating theatres and classrooms and now, in the most recent research, across human and AI collaboration.

Treat AI like a slave and you get slave-level outputs. Not because the technology is limited. Because your expectations of it are.

The colleague frame is not softer. It is harder - and it produces categorically better results. You do not command a colleague. You brief them. You give context, interrogate outputs, calibrate trust to evidence, delegate appropriately, and take genuine accountability for the outcome. When the trust is established and the value demonstrated, you give genuine autonomy. That is the journey from slave to colleague to trusted agent.

But here is what that journey makes possible - and what makes it urgent. The slave frame has a ceiling. The efficiency saving is the ceiling. The colleague frame has no ceiling - because every interaction that builds trust and context makes the next interaction more valuable. That is not a marginal improvement. That is a fundamentally different trajectory for what AI can do for your organisation.

And that trajectory demands something in return. As the transformation compounds - as the agents deliver more, as the ambition grows, as the decisions get bigger and the stakes get higher - the relationship between IT and the business has to deepen in proportion. The colleague frame is not a starting condition you establish once and move on from. It is the ongoing foundation that has to strengthen as the capability above it grows.

The organisations that will define the next decade are not just the ones that embrace agentic AI most boldly. They are the ones where IT and the business are genuinely inseparable - where the telescope is permanently connected at both ends, where the relationship is strong enough to carry the weight of what the technology is about to make possible.

That is not a technology challenge. It is a leadership challenge that belongs to the whole organisation - the CEO, the CFO, the board, and the IT leader who has the unique ability to show them why.

Part Three - The choice

I made a decision early in my career that luckily I have never had to change. I never sat cocooned in the sanctuary behind the walls of the IT Department, hidden away in the server room with only the blast of the fans and the flashing of the LEDs to keep me company. Not once, I stayed in the business - with the clients, with the leadership teams, with the people trying to solve real problems with whatever technology was available to them.

I understand why others made a different choice. We built the walls because the business asked us to. The digital era brought genuine risk and the business needed someone to manage it. So we built governance frameworks. Ticket systems. Change control processes. We moved from being inside the business to being a service the business consumed. We went from colleagues to vendors. We did it carefully, professionally, and with genuine good intent.

And in doing so we moved further from the business at exactly the moment our proximity to it mattered most.

On my worst days - the days when the business still saw me as the IT person and the IT people still saw me as the business - I questioned the choice. The walls offered something real: clarity, control, a defined domain of responsibility. But I have come to believe that the discomfort of sitting between the technology and the business is not a problem to be solved. It is the job. That discomfort is what it feels like to be genuinely useful.

The memory that matters

Think about why you got into this.

Nobody chose this career because they wanted to manage a ticket queue. Nobody fell in love with technology because of a governance framework. We got into this because we saw what technology could do for people and we wanted to be part of that. The moment you showed someone something they had never seen before and watched their face change. The moment the technology made something possible that was not possible the day before.

That moment - that feeling - is what you are being called back to. Not as nostalgia but as a strategy.

Because here is what the agentic era actually requires. It requires someone who can connect both ends of the telescope - who knows what the technology makes possible and understands what the business actually needs, and refuses to let those two things stay strangers. Someone who understands what the technology will do and cares enough about the human on the receiving end to make sure the answer is “yes, but” rather than just “yes” – not as a caution but as craft. The “but” is where the human judgement lives. Someone who arrives at the board table not as a cost to be managed but as a growth engine to be invested in. Someone who arrives not as an alarmist but as an engineer - with insight, with ideas, with genuine curiosity about what they don't yet know.

That someone is you.

The transformation that's actually required

Effective transformation equals employee engagement multiplied by transformational technology. It is a product not a sum. Deploy agents without engaged people and you get the slave frame at scale - faster, more autonomous, and just as wrong as it ever was. Engage your people without the technology and you have energy without leverage. Together, under the colleague frame, the multiplication effect is real.

But - and this is the part most transformation programmes miss entirely - the formula only works if the management model changes at every level. Not just inside IT. Everywhere. The IT leader who measures outcomes while their board measures cost is fighting with one hand tied.

This is the conversation Part One warned nobody was having. That hasn't changed. And you are the person who has to start it - not about the technology, but about what the organisation is actually for. Someone who understands both what the technology makes possible and what the business actually needs. Someone who has earned the right to be in the room not as the person who manages the plumbing but as the person who helps the organisation see what's possible.

“Shadow AI" is probably the most misunderstood signal in most organisations. I ask this question in every room I speak in and every hand goes up: who has a problem with shadow AI? Every hand. Every time. And every one of those hands is giving the wrong answer.

You do not have a shadow AI problem. You have an organisation full of people who want to do better work and are not waiting for permission. They are doing it anyway - with tools you didn't sanction, in ways you didn't design, because the need was real and the approved alternative wasn't good enough.

The Dutch have a name for the paths worn into grass by people who find their own route. They call them elephant paths. Architects call them desire paths. The formal path is beautifully designed and completely ignored. The elephant path is scruffy and unofficial - but it always shows you the shortest route from where people are to where they actually need to go. And more importantly, it tells you everything about what they actually need.

Right now, elephant paths are being worn into the grass of every organisation I work with. In real time. That is not a governance failure, that is your people telling you something.

The IT leaders who will define the next decade are not the ones who block the path or follow it after the fact. They are the ones who anticipate it - who make the desire path the design before anyone has to wear it into the grass.

That is the shift in thinking from inside-out to outside-in. And it is the most important invitation your organisation will ever extend to you - if you're willing to recognise it as one.

Closing - The right question

The navigator doesn't drive the boat.

The navigator reads the conditions, understands the destination, watches what the technology and the environment are doing, and makes it possible for the organisation to get where it needs to go. In an agentic world - where autonomous agents are acting on behalf of your organisation whether you lead that or not, where the question of what they are acting toward has never been more consequential - the navigator role is not optional.

It is the job.

The technology was never the point. The people were always the point - the people inside your organisation trying to do meaningful work, the people outside it whose lives your organisation touches, the people who haven't been asked the question yet and will need someone to help them answer it well when it arrives.

Picasso famously said “computers are useless - they can only give you answers”. In an agentic world they can now also take actions. Which makes the question of what you are actually trying to achieve more important, more consequential, and more urgent than it has ever been.

Karel Čapek gave us a word in 1920 that encoded servitude into the human-machine relationship for a hundred years. That encoding was never inevitable. It was a choice - made in a playwright's imagination, institutionalised by an industry, inherited by every organisation that never stopped to question it.

The colleague frame is also a choice. The growth engine is also a choice. The navigator role is also a choice.

We need a new Taylor. Not one who optimises the machine against the human. One who builds the conditions where humans and machines, working as genuine colleagues, achieve together what neither could achieve alone. Who measures what actually matters. Who arrives at the board table with insight and curiosity rather than governance and restriction. Who says “yes, but” instead of no.

That new Taylor is not a historical figure waiting to be appointed. It is a decision - available to every IT leader in every organisation right now.

Technology is supposed to elevate us. To empower humans to achieve more than they could do alone.

The machines were always going to rise.

The only question is whether we rise with them.