First published on LinkedIn, 19 October 2026.
In short
- Succession frameworks assume a predecessor, a handover and a defined role. A growing share of critical roles have none of the three.
- Roles that are created rather than replaced fail at the specification stage, not the selection stage.
- Four questions to answer before appointing anyone into a role nobody has held.
The role had never existed. No predecessor, no handover, and no description beyond a title and four countries. Nobody could tell me what good looked like, because nobody had done it.
So the first act of the job was not a decision about the business. It was an assessment of the four markets: what each was strong at, what each was weak at, and where to put the effort. That assessment shaped the role, simply because nothing else was going to.
It worked. The business grew from a standing start to around £30 million of revenue. What I didn’t understand at the time was that it worked by accident of fit rather than design, and that a role defined by whoever happens to hold it cannot be judged by anyone else. There was no standard to measure it against, because I had written it myself.
What I would say to a board now is that the arrangement carried a risk they never priced. Had the fit been wrong, the first clear indication would have arrived a year and a half later, in the results, because there was no interim standard the role could be held to and no predecessor whose performance offered a comparison. The upside was mine. The exposure was theirs.
What succession planning actually assumes
Succession is high on every board agenda, and rightly so. In the NACD’s 2026 Governance Outlook, directors ranked CEO succession planning as the board practice most in need of improvement. A Korn Ferry survey found that only 12 per cent of boards discuss succession five or more times a year. The Conference Board and Heidrick & Struggles have tracked the demographic pressure behind it: around 11 per cent of S&P 500 chief executives are now aged between 65 and 69, up from just over 7 per cent in 2017.
All of that attention rests on three assumptions. That there is a predecessor. That there is a handover. That someone other than the person who will hold it has defined the role.
A growing proportion of the roles that now matter most have none of the three.
The roles boards are creating right now
Consider what has been added to executive teams in the past two or three years. Heads of AI. Data and analytics leadership. Transformation directors. Roles created because the board concluded the capability was missing, in an area where the board itself has limited fluency.
These are not succession cases. Nobody is being replaced. No predecessor sets a performance benchmark, no handover to inherit, and often no clear view of what the role owns versus what it advises on. The board applies a replacement framework to a creation problem, then wonders eighteen months later why the appointment hasn’t delivered what was expected.
The failure is rarely the appointee. Nobody defined the job, so it became whatever the appointee was best at.
What the board loses
An undefined role costs the board its ability to diagnose. When the appointment underdelivers, and many do, the board cannot tell whether the role was wrong, the person was wrong, or the definition was never made. Every available remedy is therefore a guess, and the most common guess is to replace the individual, which resets the clock without touching the cause.
A second cost lands later. A role shaped around its first holder becomes a role only that person can do. When they leave, the board is recruiting against a specification written backwards from an individual, which is why second appointments into invented roles fail more often than first ones.
There is a recruitment distortion too. With no internal pipeline and no benchmark, boards reach for candidates who hold the title elsewhere, on the reasonable assumption that the title means something. It rarely does. A head of AI at another company was doing that company’s version of the job, defined by its board, its data and its customers. Importing the title imports somebody else’s definition.
Two different board conversations
Replacing a chief executive and creating a chief AI officer are treated as the same category of decision, and they are not. The first is a matching problem: the role exists, the requirements are understood, and the board is assessing fit against a specification the organisation already knows how to apply. The second is a design problem, and design problems fail at the specification stage rather than at the selection stage.
Boards are well equipped for the first and largely unpractised at the second. Nomination committees are built to assess candidates. Very few are built to define a role nobody in the room has ever held, in a domain where the board is, by its own admission, still learning.
Four questions before appointing to a role that has never existed
- What decision does this role own that nobody currently owns? If the answer is that it advises, influences or supports, the role has responsibility without authority, and the first difficult decision will go around it rather than through it.
- What is it allowed to stop? A new function that can only add activity will add activity. The authority to stop something is what separates a leadership role from a specialist one.
- Who tells the board it is not working, and when? Set the review point before the appointment, not after. A role with no predecessor has no natural performance comparison, so you have to build the comparison deliberately.
- What does success look like in twelve months, written down before anyone is hired? If the specification is written after the appointment, it will be written around the person, and the board will have lost the ability to tell whether the role or the individual is the problem.
The part that transfers
My own version of this worked out. That is not a defence of how it was done. Being handed four markets and left to define the job produced a result because the fit happened to be right, and the board had no way of knowing in advance whether it would be. They were not running a succession process. They were running a hope with a title attached.
The remedy is not complicated. Write the specification before the search. Review it at six months against what the person is actually doing, and treat any large divergence as information about the specification rather than about the individual. That single review turns an invented role into a defined one, and a defined role is the only kind that can ever be succeeded.
A succession plan that only covers roles which already exist is a plan for the part of the organisation that is not changing.
The roles a board should be most anxious about are the ones with nobody to compare the incumbent to.
Terry Irwin is a Non-Executive Director and Board Adviser to PE-backed, growth-stage and mid-cap businesses in the UK and Europe. More on his board work: NED and board advisory.