Most organisations are built around the specialist’s logic: hire the best person for each function, give each function clear ownership, coordinate handoffs at the boundaries. The org chart is a map of specialisms: product here, engineering there, data separate, commercial separate, each with its own hierarchy, its own vocabulary, its own career ladder.
This made a kind of sense when the dominant challenge was depth: when the complexity of each function was the primary constraint, and the cost of someone operating outside their lane was high. It makes less sense when the main challenge is integration: when the real problems live at the boundary between functions, and the handoffs are the primary source of friction, delay, and misunderstanding.
The problems run deeper than friction. Functional structures breed empire building: the budget that must be fully spent before year end, or next year’s allocation shrinks; the ownership that means every decision queues for one function’s approval, even when adjacent teams have the expertise and the bandwidth to act. Everything moves at the pace of the slowest specialism, because crossing the boundary to help — even when perfectly qualified to do so — is structurally discouraged. The incentive at every level is to protect the domain, not to serve the outcome.
The shift towards T-shape was happening before AI. AI accelerates it by making depth more abundant and integration more scarce. This is the case for the t-shaped organisation: a structure that builds breadth into the design, not just the culture.
A T-shaped organisation, as defined in the T-shaped professional pillar, is not an organisation without specialists. It is one where specialists develop genuine operational literacy in at least one adjacent domain, and where the structure actively enables that development rather than rewarding pure depth. More than that: it is one where multiskilled individuals are welcome to contribute outside their primary remit. Not just to develop breadth — to use it.
In practice, this comes down to three things — all of which are harder to implement than they sound.
The first is hiring for range alongside depth. Not depth or range — both. The question in a hiring process is not just “what do you know?” but “where have you operated outside your core domain with real stakes, and what happened?” The absence of an answer to the second question is a flag. The candidate who has never crossed a boundary is telling you something about how they have managed their career. For most leadership roles and most roles that require translating across functions, that matters.
The second is rotating people deliberately. Not randomly, and not as a “development opportunity” that looks good in a performance review but produces nothing. Deliberately: identifying the specific boundary that needs strengthening, assigning someone who has enough depth to contribute meaningfully once they get there, and giving them long enough to get past the initial uselessness phase. Six weeks is too short. Six months may be about right.
The third is measuring integration, not just output. Most performance frameworks measure what a person does inside their domain. A T-shaped organisation also measures what they do at the boundary: problems spotted before they became escalations, translation that unblocked a team on the other side of the fence, questions asked in rooms they were not technically responsible for. None of this appears naturally in a standard performance review template. You have to design for it. One step further: measure the contribution of the team, not just the individual. Individual metrics reinforce the specialist’s incentive to defend their domain; team metrics create conditions where crossing the boundary makes sense for everyone.
What this costs, in the short term, is efficiency. A specialist spending time in an adjacent domain is not fully deployed in their core one. Rotations have ramp-up costs. Hiring for range takes longer and is harder to screen for than hiring for depth. The organisation that runs this model will, at any given moment, appear to be running at slightly below the output efficiency of the one that does not.
The payoff is compounding.
Organisations with genuine T-shaped breadth in their people have fewer handoffs, because the people who need to collaborate across functions understand each other well enough to communicate directly. Boundary decisions get made by people who have earned the right to make them. And context that used to bottleneck through one or two individuals becomes more widely distributed.
These organisations are also structurally better equipped to work alongside AI agents.
Most discussions of AI transformation miss this. The question of how an organisation works with AI is not primarily a technology question. It is a question of whether the people in the organisation can evaluate what the AI produces, redirect it when it is wrong, and integrate its output with the context it does not have. That requires the T-shape at the individual level. At the organisational level, it requires enough people with genuine cross-domain literacy that the integration is a distributed capability, not something that bottlenecks through the handful of people who happen to have been exposed to multiple functions.
You Can’t Build a Centaur made the point that the best human-AI collaboration is not a workflow where people supervise AI output, but a centaur: a person integrated enough with the tool to direct it in real time. A T-shaped organisation is one where more of your people can be centaurs — not just in one domain, but at the intersections where the real decisions are made.
You do not build a T-shaped organisation by writing a strategy document or announcing a new operating model. You build it by what you hire for and what you rotate people through. The rest — the vocabulary, the culture, the meeting dynamics that feel different — follows from those two decisions, not the other way around.
The organisations best placed for the next five years are not the ones with the deepest specialists or the broadest generalists. They are the ones that have built the capacity to integrate across domains into the structure, before the pressure to do so became unavoidable.
The T-shape is not a philosophical stance. It is a competitive advantage.
Leave a Reply