The Great Remaking (Video)

Transcript

I first proposed The Great Remaking in early 2026, and I proposed it because of what I’d seen, not what I had read. In the organisations getting the biggest returns from AI, the ones everyone else was trying to copy, nobody was bolting a tool onto the business and hoping for the best. They were taking processes they had relied on for years, sometimes decades, pulling them apart and rebuilding them around what the machine could do now. Not faster versions of the old work, different work. I called it The Great Remaking because that is the scale of it. I’ve worked through five technology revolutions and none of them asked this of a business. AI does. Every organisation will have to remake itself, and the ones that start now will take the largest share of what is on offer. What follows is what is being remade, why the gap between those who move and those who wait keeps widening, and what a Board should be asking about its own position. Five technology revolutions have reshaped how organisations operate in the past half century. The desktop computer transformed individual productivity. The internet transformed connectivity. The web transformed distribution and commerce. Mobile transformed access and ubiquity. And the cloud transformed how organisations provision, build and scale. I’ve lived through all five and work at the heart of the last three. And I can tell you that each felt transformative at the time because each was. They were significant, and they genuinely changed how businesses operated. So it would be the most natural thing in the world to read AI as the sixth, the next wave in a familiar progression to be met with the same playbook. It’s not. The five waves changed the tools around the work. How we computed, connected, distributed, accessed and scaled. The work itself, what a business actually does all day, came through every one of them largely intact. AI is different in kind. It doesn’t sit at the end of the sequence. It runs beneath all five, and it is remaking the work itself. Which raises the question the next slide answers. What precisely is that work?

Strip any organisation back far enough, whatever it sells and wherever it operates, and you find four kinds of work, and only four. I call them the essence of work. Thinking, making sense of the world, the analysis, research, synthesis and pattern recognition that tells you what is going on. Deciding, the committing to a course of action, the resources needed to support the action, and bearing the consequences. Creating, designing and building something new, a product, a campaign, a line of code, a contract. And delivering, producing outcomes in the physical world where atoms move and not just bits. These are not departments. They are the irreducible dimensions of what a business does, and AI is reaching every one of them.

But what is AI doing to each essence? And what we find when we look at that is that the pattern is asymmetrical. Because AI is not touching the four in the same way or at the same speed. In thinking, the machine already outreads any team you can hire. What matters is that the first phase, AI as a faster research assistant, is already commoditised. And the organisations whose analysis rests on public information are the most exposed. In deciding, the change is quieter and deeper. A credit manager who approves or overrides a recommendation is deciding differently from one who worked the case from first principles. Even though the signature looks the same, and accountability, as we will see, stays exactly where it was. In creating the most visible shift, a marketing function that has genuinely restructured, is not writing copy faster. It runs a different ratio of human direction to machine execution. More variance, faster tests, people concentrated on tone, values, and brand. And the people building the technology are quite open that where the output is digital, the whole process is a target. In delivering the story is cost as much as capability. Humanoid unit costs falling 40% a year. And the gap between this is a manufacturing story and that is our story narrower than most operating plans assume.

The third row is the moat. And the thing to notice is that none of them is the technology. In thinking, it is not the data you hold, but your proven ability to make it useful to your own systems. Plenty of organisations sit on decades of data and can’t reach it. In deciding, the moat is speed with governance attached. And because a decision made at machine speed without oversight is liability, it’s not an advantage. In creating, execution capacity is no longer a moat at all. Originality and taste are. In delivering, physical complexity still protects you, but it is shrinking. And how fast depends on your sector. The last row is what stays human. And it’s where the investment goes. Judgement. Knowing when the analysis is technically correct and strategically wrong. The willingness to commit, which no model carries. Knowing what is worth making and what reflects your values. Adaptability and trust, where the environment will not hold still. And running through all four, not on the grid at all, but it’s the thing AI replicates least well, is relationships. Because trust is relational, not functional. And the machine can’t yet be trusted as a counterparty. Four essences, all being remade, none of them at the same pace.

The four essences tell you what the work is. They don’t tell you how it gets done. And for that, we need a second idea. Because every piece of work, whether a person does it, a machine does it, or both, runs on the same eight components. Knowledge. What you know and how you reason about a problem. Perception. How you gather and understand what is happening around you. Planning. How you break an objective into steps and decide what to do in what order. Action. The doing itself. Memory. What you retain and build on rather than starting fresh each time. Adaptation. How you improve from what worked and what did not. Orchestration. How you coordinate when many things must happen at once. And governance. How you keep the whole thing safe, compliant, and aligned with your values and ethics. Those eight run across the top of this grid. And the four essences run down the side. So every cell is one component of one kind of work. How a decision perceives. How creating remembers. How delivery orchestrates. 32 cells. And between them, they describe how an organisation actually works. The components never change. What does change, and this is the entire story, is who or what performs each one. And that’s what the colours show. Grey means unchanged. People hold both the agency and the accountability. And they always did. Pale yellow means assisted. The machine informs, drafts, recommends. People still do the work. Yellow means delegated. People have handed a meaningful part of the work to AI to perform within boundaries. And their job has moved from doing it to directing and overseeing it. Red is redesigned. The work has been reorganised around distributed agency. People and machines together with human accountability built into the design. Because it can’t be delegated away. Read the colour in any cell. And it tells you where authority has moved. This is today. And the scores are mine. The warmth sits in the thinking and creating. Where the machine reads and drafts better than teams of people already. Deciding is cooler. Agency is being handed over more carefully there. And for good reason. Delivering the physical world is only just beginning to change colour. Notice before we move on how much of the grid is still grey or pale. Most organisations are assisted, not remade. But here is where it’s heading. This is the same grid. The same key. A few years out. Almost everything has turned red. The difference between this picture and the last one is the remaking. That distance is what every organisation is going to travel. Whether it chooses to or not. Look first at the three columns on the right. The memory, adaptation, orchestration columns. Those are the components that make an agent an agent. A system that remembers what it has done. Improves on it. And coordinates with other systems without being asked. They are pale today. And they will be red. And that is where the next wave of remaking happens. Because once work can remember, learn and coordinate on its own. The shape of the organisation around it stops making sense. Look next at the delivering row. It warms all the way across. And it is a cost that does it. Robots getting cheap faster than operating plans get written. And look at what redesign now means for a director. It doesn’t mean that the machine runs the business. It means the work has been reorganised around people and machines together. With the accountability for it written into the design. Because in a redesigned cell, there is no one to hand it to. Which is why one cell stays cool in both pictures and stays cool on purpose. The last cell of the deciding row. The governance of decisions. AI can take agency in the decision. It can assess, recommend, and act. It cannot take the accountability. That doesn’t move with it. And that cell is the Board’s.

The governance of decisions stayed cool in both pictures for a reason. And this is it. Every remaking in any organisation has two groups in it. The people who do the work. And the people who answer for it. Agency on the left. Accountability on the right. The machine is now joining the first group. And joining it fast. Assessing, recommending, in bounded domains acting. It is not joining the second. Not yet. And not under any law a Board operates under today. An underwriter’s value was never the analysis. It was the signature. A Board’s value was never the insight. It was the commitment and the consequences that are attached to it. When a machine is wrong, the Board bears it. So a remaking only works when both sides are present and looking at the same thing. The people doing the work and the people answerable for it. Working from the same stages and the same evidence through the same window. And one rule should stand at the front door of every piece of work an organisation remakes. If either side is missing, it does not happen. No operators with the skills and authority to do it. It doesn’t start. Nobody answerable all the way up to the Board. It does not start. A sponsor who fades after kickoff is not accountability. And a change nobody answers for fails often enough that the right decision is simply not to begin one. Governance is not bolted on beside the work. It is built into how the work runs. That is the standard. Now let me show you what the evidence says happens when organisations get this right.

This is what the remaking pays. In the words of the people who measure it, nearly 9 in 10 organisations already use AI and 4 in 10 have it in production at scale. But the gains, where they arrive, arrive for a reason.

Two thirds of people get an hour a day back when they work in redesigned processes around the machine. Drop a tool into the old process instead and only half of them do. Productivity rises a third when people guide the AI rather than hand over to it. The organisations that redesign are banking three and a half times the shareholder return on the laggards, nearly double the revenue and three times the revenue per employee, and a premium in the labour market for the people who can work this way. And the part the headlines get wrong is the people. On the evidence, the workforce is being redeployed, not replaced. A quarter to two fifths of roles are retrainable. A fraction of a percentage of last year’s layoffs attributable to AI. Workers who want the hours back and the remaking now reaching the physical world as robots get cheap. That is what you get if you get this right. Growth, returns and your people doing more valuable work. But there is an alternative view here and it’s the same evidence. BCG puts 10% of AI’s value in the algorithms, 20% in the technology and 70% in redesigning how people and machines work together. If you bolt AI onto the old processes and you are buying the 30% that every competitor can buy too, and you get the bolt on numbers, half your people getting the hour back instead of two thirds, the eight in 10 organisations reporting no meaningful bottom line gain, the pilots that stall, do nothing. And you are not standing still because the organisations behind these numbers are not standing still either. Neither outcome is a verdict on AI. Both are a measure of execution and oversight capability, which is precisely what the Remake framework is built to supply. And it’s why the value is already concentrating in very few hands.

Two numbers and together they say that the prize is not being shared out. It is being claimed. Three quarters of all economic value AI has created so far is sitting with a fifth of organisations. And inside that fifth, the ones who have turned it into substantial financial gains, real revenue, real cash flow, real margin, are around one in 20. Everyone else is somewhere between a promising pilot and a line in the annual report. What makes that a race rather than a league table is what the 5% have in common. It is not that they are the largest or the best funded or they bought the better model. The models are available to everyone. They redesigned the work early and every month since then, the redesign has been paying them in data, in capability and in speed that the others do not have yet. So the gap you are looking at is not a technology gap. It is a systems gap. And systems gaps do not sit still. They compound.

And this is where they compound. Starts with a single decision to restructure one piece of work around AI rather than bolt AI onto it. That decision sets three loops running and they link, which is why the gap compounds instead of closing. The first is data. And the starting point is an asset you already own. Every established organisation sits on years of proprietary history. Transactions, decisions, customers, operations. The closed institutional knowledge no model has ever been trained on and no competitor can reach. For most, it is sitting in systems that AI cannot get at. The redesign is what moves it into the models. And the moment it lands there, it starts to compound. The models get better on your history. The redesigned work produces data the old process never did. A credit function that has run AI-integrated decisions for two years holds two years of decisions with outcomes in a form its models can learn from. And each cycle makes the next one sharper. Nobody outside can buy any of it. Not the history because it is yours. Not what the work now produces because it only exists where the work has been changed. The second is talent. And it is the loop most often mistaken for a training programme or an enablement programme. People doing redesigned work develop a judgement that no course conveys. When to trust the system. When to override it. And why, in the context of the business, it is important. It is tacit. It accumulates and it is visible from the outside. So an organisation known for it attracts more of the people who have it. The moat is not only what your people have learned. It is who chooses to work for you next. The third is redesign itself. Each cycle teaches the organisation how to do the next one. What it takes across the functions. Where the value is. What governance keeps it honest. The second redesign is faster than the first. The fifth faster still. And at some point it stops being a programme and becomes simply how the place works. And the three feed each other. Your own data inside the models makes better AI. Better AI develops more capable people. More capable people design better work. Better work produces richer data. And on top of what you already had. Round again and a little faster each time. And it does not stop. Which is exactly the problem for anyone planning to catch up later.

The instinct of most Boards is to follow fast. Let the early mover take the risk. Watch what works. Then buy the equivalent. And close the gap. In every previous wave that was perfectly good strategy. Because what the early mover had adopted was a product. It was observable. It was replicable. And by the time cloud arrived it was rentable by the hour. A company that had spent three years building data centres. Could be matched by one with three months and a credit card. The advantage was real. And it was purchasable. So the follower caught up. This time the advantage is not a product. It’s the three loops. And the centre of them is the one thing a follower can never acquire. The leader’s own data. Years of it. Already inside the leader’s models and compounding. Compounding. Plus everything the redesigned work has added since. There is nothing to study and nothing to copy. Because the asset is proprietary. By definition. And it only grows where the work has been done differently. And the technology itself is going the other way. The gap between the best model and the tenth best has narrowed to a few percentage points. And the costs of frontier level inference fell 280 fold in two years. The models are commoditising. Which means the advantage has moved decisively away from which model you can access. And towards what you’ve put into it and built around it. You can buy the technology. You can’t buy what the leader has put into it and built around it. Because that is theirs. And it only grows where the work has been done differently for long enough to accumulate. So a late mover starting today is not 18 months behind. It is structurally behind. And every month the gap widens rather than closes. Which is why the questions a Board asks have to change.

So the questions change. The ones most Boards ask. How many pilots? How much spend? Do we have a strategy? Do we need a chief AI officer? are activity measures. They tell you money is moving and projects are running. Which matters. And which is not the same as knowing whether the loops are turning. Three questions, one per loop. And they are hard to answer well with a prepared slide. On data. The probe is whether the management can name a workflow that has been redesigned around AI. Not augmented in the last 18 months. And say what data it now produces that the old one never did. A credible answer has a name and a measurable asset. Our data strategy is not the answer. It is the sound of the loop not running. On talent. The probe is what would actually walk out of the door. If your most AI capable people left tomorrow. If the honest answer is tool skills you can hire back. The loop is not producing advantage. If it is hard-won institutional knowledge. Of where to trust the system. And where to override it. And why it is. On redesign. The probe is what the organisation has learned about redesign itself. Not about any particular technology. And what it would do differently next cycle. A specific operational answer. Means a practice is forming. Generalities around change management. Mean a project happened. And the organisation went back to how it was. None of this is a criticism of management. Overestimation is structural. Every organisation inflates its own AI position. And these questions are the corrective. The teams that answer them well. Are building the loops. The teams that reach for the dashboard. Are where most organisations are today. And that is the window.

None of this waits for a transformation programme to finish. It begins with a decision. And the decision is small enough to take this quarter. Take one piece of work. And restructure it around what the machine can do. Rather than adding the machine to it. That is available to any organisation today. Its value starts compounding the moment it’s made. And it depreciates with every month it is deferred. Because the organisations already making it. Are not waiting for you. Remake is how I make that decision operational. And the four questions on the screen. Are its shape. Which work. Named by its essence. Because AI transformation is not a piece of work. And a redesigned credit decision is. Who has the skills. And the authority to remake it. And who answers for it all the way up. Both present. Or it does not begin. That is the rule from earlier. Built rather than bolted on. How one piece of work. At a time. Through the stages that earned. On evidence. Rather than scheduled on a plan. And what you get at the end. An investment case that stands up to scrutiny. A plan you can actually run. And an organisation that sustains the change. Without the programme team in the room. Each remaking. Makes the next one faster. The organisations already doing this. Are not simply ahead. They are ahead and accelerating.

The full argument. And evidence behind every number you’ve seen. And the diagnostic. Are in the Board briefing. At mariothomas.com Start with the four articles in order. The compounding argument. in the third changes how you read the fourth. Then take the three questions. Into your next AI agenda item. And listen to the answers. What comes back. Will tell you more honestly. Than any dashboard. Whether your organisation. Is remaking the work. Or bolting AI onto it. Thanks for watching.