AI and Organisational Design: When Capability Belongs to Opportunity and Purpose, Not the Org Chart

AI is dismantling the artificial boundaries between job roles inside organisations, making it possible for people to move across disciplines without waiting for specialist teams. This creates a strategic need for internal talent mobility, but the psychological fences built by years of conditioning often remain long after the technological ones have fallen.

1. What I saw in Kruger

I was walking through Kruger National Park recently when I noticed small pieces of metal sticking out of the ground, the remains of old fence posts. I asked the ranger what they were, and he explained that Kruger had not always been the enormous open ecosystem we see today. Parts of the park had once been divided by fences, animals had lived within separate sections, and because they were contained, people had to actively manage what happened inside each one. Over time, many of those internal fences came down.

I asked what difference that made. He said some people believe Kruger has too many elephants, since elephants can fundamentally change an ecosystem when their numbers concentrate in one area, and the historical response to that has included culling. But there is another argument. If animals are free to move across a sufficiently large ecosystem, perhaps our first response should not be to micromanage the population at all. The ecosystem can still be managed, sometimes quite deliberately, but the starting instinct shifts from controlling every enclosure to understanding the whole. Perhaps some of the problem was created by the boundaries we put around the animals in the first place. We built the fences, created the contained system, and then had to manage the consequences of our own containment. As I stood looking at that old fence, I realised something remarkably similar is beginning to happen inside companies because of AI.

2. We built fences around people

Most organisations are collections of artificial boundaries. Engineering sits here, operations sits there, data belongs over there, marketing has its territory, and finance and risk have theirs. We then created job descriptions, reporting structures, budgets and permissions that reinforced those boundaries, until we became so accustomed to them that we forgot they were artificial.

If operations wanted software built, they needed engineers. If an engineer wanted sophisticated analysis, they needed data. If product wanted marketing material, they needed marketing. The fences determined where people could roam, and because those fences existed, organisations needed enormous amounts of management, through projects, committees, handoffs and governance forums, simply to coordinate activity across them. Much of what we call management is really the machinery required to move work across organisational fences. Then AI arrived.

3. AI is ripping the fences down

Over the last six months I have watched something fascinating happen at Capitec. The boundaries between skills are becoming porous. An engineer can suddenly do sophisticated data analysis. A product person can prototype software. Someone in operations can interrogate logs, understand code and automate a process, all without waiting for a specialist team. None of that required anyone to change teams or wait for a reorganisation. The capability moved while the org chart did not, and I will come back to what happened when the centre noticed.

The skill itself has not disappeared, and expertise still matters enormously. What has disappeared is the absolute dependency on somebody else having that expertise before you can move, and that is a profound change. Some teams have understood it instinctively: given a problem, it becomes difficult to tell where one person’s job ends and another’s begins, because people move towards the problem, learn things outside their discipline, and help one another. The fence is gone, so they roam. But I am also seeing something else, which is that some people are standing exactly where the fence used to be.

4. The fence can disappear before the psychology does

This may be the most interesting part of the AI transformation. Removing a boundary does not necessarily make someone cross it. For years we taught people that competence meant staying within their area of expertise, and job descriptions and reporting lines told them what they were responsible for and where they belonged. Those structures provided more than clarity. They provided certainty and identity, answering questions like what am I responsible for, what am I allowed to do, and perhaps most importantly, who am I here.

Psychological safety becomes important precisely once people start crossing those boundaries rather than staying inside them. Amy Edmondson describes psychological safety as the belief that it is safe to take interpersonal risks, meaning speaking up, admitting mistakes and experimenting without fear of embarrassment, and that is exactly the belief someone needs when moving into unfamiliar territory in front of colleagues who knew it first. Research into job crafting adds another piece, showing that people can reshape the tasks and meaning of their work rather than simply occupying a static description, and AI dramatically expands the territory in which that reshaping can happen. But we have removed the technological barriers much faster than the psychological ones. AI might tell someone they can do this now, while twenty years of conditioning still tells them it isn’t their job. That is why two people given the same tools can behave completely differently. One sees an enormous open landscape, and the other sees the ghost of a fence.

5. AI is also an identity crisis

The conversation about AI at work is usually framed around skills: what will disappear, what we will need, how to train people. Those are important questions, but underneath them is something more human, which is identity.

Imagine spending fifteen years becoming the person who knows how to do something difficult. That expertise becomes part of your professional identity, determining your status and your place in the organisation. Then a machine arrives that lets somebody with a fraction of your experience attempt parts of that work, and that can feel threatening even when your expertise remains valuable. The instinctive response is to defend the boundary, through statements like this needs to sit with my team, or we need proper governance. Sometimes those statements are correct. But sometimes they are requests to rebuild the fence.

6. There is another person standing at the fence

I used to think the interesting question was why some people cross these disappearing boundaries while others stand where the boundary used to be. I now think that is only half of it, because there is another group standing beside the old fence: the people who used to own it.

When a capability that was historically centralised gets distributed, the people doing the work move closer to the problems they are solving, decisions happen faster, and teams become more autonomous, so performance can improve dramatically. But the leader of the old central function has lost people, decisions, visibility and often budget, and with them some of the things by which organisations have traditionally measured leadership: headcount and control. That creates a psychological trap, because loss of control is very easily interpreted as loss of quality, and those are not the same thing.

7. When better feels worse

A distributed team can objectively be performing better, shipping faster and operating closer to customers without the old coordination overhead. Yet from the old centre’s perspective, the organisation can look worse, because there is more variation, more decisions made without permission, and it becomes harder to see and control everything. The organisation feels less orderly, and human beings have a powerful tendency to confuse order with effectiveness. So when inconsistencies appear, the old structure suddenly looks attractive again, through calls for stronger governance, a single accountable executive, or a return to old standards.

Each of those statements may contain a legitimate concern, but the proposed solution often hides an unstated assumption, which is that to influence something, I must control it. AI is going to challenge that assumption profoundly.

I have seen this pattern play out cleanly. A capability sitting inside a central function gets pushed out into the teams using it every day, each picking up AI tools to do work that used to require a handoff back to the centre. The embedded teams start shipping faster, and quality holds up because they are closer to the problem than the centre ever was. Yet within months, the former centre starts making the case for bringing the capability back, not because anything has gone wrong, but because it can no longer see everything happening at once, and an absence of visibility gets quietly reinterpreted as an absence of control. That is the moment worth watching for, because it is the moment an organisation decides whether it trusts the outcomes it is measuring or the feeling of oversight it has lost.

8. The people who owned the fences need a new purpose

Many organisational transformations go wrong here. When a central capability is distributed, we focus almost entirely on the people who have moved and rarely spend enough time redefining the purpose of the people who remain. If your leadership identity was built around owning a hundred people now embedded across the organisation, what exactly are you leading? The easy answer is to pull them back, but that is usually the wrong problem to solve.

The central function needs to evolve from a centre of control into a genuine centre of mastery. Rather than owning the people, it owns the standards, and rather than approving the work, it builds tools that make good work easier. It develops specialists rather than allocating them, and measures outcomes rather than controlling execution. It stops being the place where expertise lives and becomes the mechanism through which expertise travels. That is not a smaller leadership role. I would argue it is a much bigger one.

9. Influence without ownership

For decades we have measured organisational importance through proxies for control: how many people report to you, how big your budget is, what decisions require your approval. AI increasingly asks us to measure something different, which is how much capability you enable.

Imagine a leader with only ten direct reports whose standards, platforms and coaching make ten thousand people better at their jobs. Is that person less important than someone with five hundred direct reports? Of course not, but many of our structures, titles and incentives still behave as though they are.

There is a simple way to state what is changing. For most of the history of modern organisations, power has effectively been measured as people, budget and decisions controlled. What this article is really arguing is that power should increasingly be measured as capability enabled. That is a considerably more radical proposition than decentralisation, because it changes what counts as a big job, how careers progress, and the incentives underneath almost every executive structure we have. This is why AI transformation is a leadership transformation as much as a technology one, and some leaders will have to learn that their influence can increase while their control decreases.

10. Standards do not require fences

None of this means abandoning standards. If twenty teams independently build twenty versions of the same component, that is wasteful, and if security or regulatory controls are ignored, that matters enormously. But the answer does not have to be putting everyone back in the same department, because organisations have historically bundled together two concepts that are actually separate: governance and containment. You can have common standards without common reporting lines, security controls without the security team writing every application, and expertise without monopoly.

That distinction matters given where the conversation is heading. I increasingly hear questions about where AI capabilities should live, who owns automation, who owns agents, who owns the platform. These sound like sensible questions, but they may belong to the world we are leaving rather than the one we are entering. If AI genuinely lets people operate across traditional skill boundaries, drawing stronger organisational boundaries around those capabilities could destroy much of the value we are trying to create, rebuilding fences around animals that have just discovered they can roam. That does not make Kruger an anarchy. It still has rules, boundaries and expert conservationists who intervene when necessary. The job of the centre changes rather than disappears, moving from containing capability to amplifying it.

11. Managing an ecosystem instead of an enclosure

When Kruger’s internal fences came down, the job of conservation did not disappear, it changed. The rangers did not become irrelevant because they stopped deciding which enclosure every animal belonged in. Their expertise arguably became more important, but the nature of their responsibility shifted from managing individual enclosures towards understanding an ecosystem.

Leadership inside companies is going through something similar. The challenge is not deploying AI tools or training people to write prompts, and it is certainly not creating another central AI department and expecting transformation to radiate outward from it. It is learning to operate an organisation when capability is no longer aligned with the org chart, which means giving people confidence to cross old boundaries, rewarding movement towards problems rather than protection of territory, and becoming comfortable with a little less control. When the fences disappear, the organisation can look messier, and some experiments will fail. But there will also be extraordinary things created by people who discover the territory available to them is far larger than they thought.

12. Who finds this genuinely uncomfortable

It is worth naming plainly who struggles most with this, because the discomfort is not evenly distributed.

Chain of command leadership struggles hardest, because its entire operating logic depends on instructions flowing down and status flowing up through a known structure. Ambiguity about who is deciding something is not a minor irritation there. It is close to an existential threat, because the chain only functions when everyone can point to exactly where they sit within it.

Teams with rigid ways of working struggle for a similar reason, particularly when their identity is built around a hardened process and a leader who enforces it without much negotiation. These teams often perform well in a stable environment precisely because the process has been tuned to remove variation, so a team whose competence comes from executing that process correctly will experience capability moving across boundaries as a threat rather than an opportunity, and a dictatorial leader has the least incentive of anyone to loosen the grip that gives them authority.

Project managers, as a discipline, sit in a particularly difficult position. Much of their traditional value comes from sequencing dependencies, fixing timelines, and giving an organisation the comfort of a plan that appears under control. That discipline was built for a world where capability sat in known places and moved along predictable paths, and it does not map cleanly onto a world where work can suddenly be attempted by someone outside the plan, or where the honest answer to how something gets done changes week by week rather than being fixed at the outset.

There is one more personality worth naming honestly, because it explains some of the loudest resistance. Narcissistic leaders, in the everyday sense rather than the clinical one, tend to experience this shift as something close to a personal injury, because so much of their sense of importance is built on visibility and credit. When capability distributes, credit distributes with it, and a leader whose self worth depends on being seen as the source of every good outcome will find that intolerable long before they find it inefficient. That leader is often the most articulate voice arguing for recentralisation, dressed up as a concern about governance, because losing the spotlight is harder to admit to than losing control ever was.

What this actually requires is leadership built around continuous dialogue rather than a document: decisions made in short cycles, in conversation with the people closest to the work, with a roadmap treated as a direction rather than a contract. None of this removes the need for planning where genuine dependencies exist. But the leaders and teams whose comfort depends on a fixed plan, a settled hierarchy or a guaranteed spotlight are the ones who will feel this transition most acutely, and recognising that early is far better than discovering it mid reorganisation.

13. Capability has become mobile, and that is the real asset

There is a practical claim underneath everything I have described, and it deserves to be stated on its own. The valuable thing now moving around inside organisations is not primarily people changing jobs. It is capability itself becoming mobile, meaning someone’s effective competence can travel across a professional boundary without that person changing teams or waiting for a formal mobility programme to notice them.

That distinction matters, because internal talent mobility in the traditional sense has always existed and always been slow, requiring a manager willing to let someone go and a great deal of individual persistence. AI does not primarily accelerate that process. It sidesteps much of the need for it. To be precise, AI does not compress years of expertise into an afternoon, and nobody should pretend it does. What it compresses is enough of the accessible knowledge around that expertise to let someone make a credible first attempt at work that used to require the specialist entirely. An operations person can now produce a meaningful piece of engineering work from exactly where they sit. The expertise stays valuable while the monopoly on attempting the work disappears. The capability moves. The org chart does not need to.

That is the asset leadership teams should be watching, and most are not managing it deliberately at all. It shows up as an engineer quietly producing analysis that used to require the data team, and it is easy to mistake for an isolated productivity win rather than evidence of a structural shift in where capability lives. Leaders who notice can act directly: give people explicit permission to attempt work outside their formal role, build the common tools and standards that make it safe for capability to travel, and start measuring how much capability a team enables elsewhere, not only what it produces inside its own fence.

This is the moment I promised to come back to. None of it is theoretical for me. Inside Capitec I have watched people reach for AI to attempt work that would never have been part of their job description five years ago, without moving teams or anyone reorganising around them, and the organisation is materially better for it. What was missing was never the talent. It was a structure that made reaching for the problem the obvious thing to do rather than the exceptional one.

14. The old fence line

I keep thinking about those pieces of metal sticking out of the ground in Kruger. The fence itself had disappeared years ago, but you could still see exactly where it used to be. Many organisations are entering that phase now, as AI removes barriers between disciplines faster than we can redraw our charts. Some people have already crossed the old lines and barely notice them anymore. Others are standing beside fence posts waiting for somebody to open a gate that no longer exists. And the people who once owned those enclosures are standing there too, some trying to understand their new purpose, some learning they can influence an ecosystem without owning everything inside it. And some are arriving with new rolls of wire. That last group worries me.

Because the great organisational opportunity created by AI is not that we can rebuild our existing organisations more efficiently. It is that we may no longer need many of the boundaries around which those organisations were designed. The job of leadership is changing. We don’t need leaders who protect territory, we need leaders who create environments in which capability can move. We don’t need centres that accumulate expertise, we need centres that distribute it. We don’t need control to create standards, and we don’t need ownership to create influence.

Perhaps the most important lesson I took from Kruger was not really about elephants at all. It was about leadership. You don’t have to own the animals to care for the ecosystem, and perhaps the job of leadership in the AI era is not to decide where to build the next fence. Perhaps it is to help everyone realise that the old one is already gone.