Someone builds the succession plan in a workshop. Names go on a slide. Everyone in the room agrees. The deck gets saved somewhere sensible.
Then things move. The named successor takes a sideways role. Another person quietly picks up the exact skill the plan says is missing. A third stops putting their hand up for anything. None of it gets written down, because writing it down is nobody's job in particular.
Nine months later the seat actually opens. What HR has is not a plan. It's a document about a company that no longer exists — and the search starts from the beginning, under time pressure, with the person who knew the role already halfway out the door.
We've read the Gartner AI Use-Case Assessment for HR, in which ClayHR is named. We won't summarize it here. The research is licensed, and a summary would be both a licensing problem and a poor substitute for reading it properly. If you have access, the link is at the bottom.
What follows is our own argument, sharpened by the read.
Most HR teams can name successors. Ask a leadership team who would step into the operations director role and you'll get answers — maybe not confident ones, but answers.
The frameworks aren't the bottleneck either. Nine-box grids, readiness tiers, critical-role definitions: these have been around for decades and they work fine. Nobody's succession programme is failing because the grid has the wrong number of squares.
What fails is maintenance. A succession plan is a set of claims about people, and those claims decay. It says this person will be ready for that role by roughly this date — a claim resting on what someone delivered this quarter, what training they finished, what their manager said in a review, what they've told you they actually want to do next.
Those things live in four or five different places. Reconciling them is manual work, so it happens once a year at best. The plan is accurate for about a fortnight after the annual talent review and increasingly wrong for the other fifty weeks. That cadence made sense when the data could only be gathered by hand. The constraint has gone. The cadence hasn't.
Here's the part that gets less attention. When a plan is built from memory in a room, it doesn't just go out of date — it starts out skewed.
The names that surface are the visible ones. People who present to the leadership team. People who happen to sit near the decision-makers. People who delivered something memorable recently, which is a different thing from delivering consistently. This isn't malice, and the people in the room would be genuinely surprised to hear it described this way. It's how recall works.
The same twelve names then circulate for years, because nothing arrives to disturb them. Someone two levels down who has quietly become the strongest candidate never enters the conversation, because entering the conversation requires someone to remember them at the right moment.
Current, evidence-based data doesn't eliminate bias. It removes one of its biggest inputs — the reliance on who happened to come to mind. A record built from completed work, finished training and structured feedback surfaces people the room would have skipped. The decision still belongs to humans, with all their judgement and all their blind spots. But they're choosing from a wider and more honest starting list.
The third problem is the direction of the question. Succession planning asks who replaces this person. The strategic question is different: will this role still exist in three years, and in what shape?
Plans get built against today's org chart, so they optimise for continuity of positions rather than continuity of capability. You end up with a well-developed successor for a job the business is about to redesign — and nobody ready for the capabilities you'll actually be short of.
Connecting the two means treating succession as a skills question rather than a headcount question. If you can see what your workforce can do today, and what the business needs it to do next, the bench you're building starts to reflect where you're going instead of where you've been.
Not choosing the successor. That decision involves judgement, politics, and things that aren't in any database. It should stay with the people accountable for it.
The useful work is narrower and duller:
That's an unglamorous list, and we think it's the honest one. A system that quietly keeps the picture current is worth more than one that produces a confident recommendation from stale inputs.
Skill levels update from evidence, not from data entry. ClayHR updates skill levels automatically based on AI assessments, performance reviews, completed learning programmes and skill-linked feedback. This is the mechanism everything above depends on.
Critical roles and gaps come first. You identify the roles the business can't afford to leave empty, then see which have no qualified successor behind them. The gap is the output, not the name.
Employees assess themselves. Self-assessments sit alongside career plans, so a person's own view of their skills and ambitions enters the record directly — including people no one in the workshop would have named.
Matrices you can configure and filter. Nine-box and four-box performance–potential matrices, filterable by role and organisational unit, so you can examine one business area properly rather than circling the same executive layer.
A capability view, not just a headcount view. A colour-coded skill matrix maps competencies across roles, teams and projects, and gaps feed targeted learning — so development work is pointed at what the organisation is short of.
Knowledge doesn't leave with the person. People retiring or moving on can upload documents and knowledge artefacts through Ask HR, so successors can find them later.
Software does not fix bias. It narrows one source of it. If your leadership team is determined to promote the people it already likes, better data will not stop them — it will just be ignored more politely. This is worth being clear about, because the opposite claim is common and it isn't true.
Automatic updates need something to update from. No performance reviews, no learning completion tracking, no structured feedback — nothing arrives for the system to act on, and you're back to manual entry with extra steps. The mechanism is only as good as the habits underneath it.
Succession Planning sits in a higher package tier, not the entry one. Worth checking before you get attached to the idea.
And if your succession process is twelve executives and one board conversation a year, don't buy software for it. A spreadsheet and a good facilitator will do the job. This becomes worth it somewhere around the point where you're planning across a few hundred roles and multiple business units — or if your central question is about the external labour market, in which case a specialist tool serves you better than we will.
If you're thinking through how your succession programme should work, we're happy to talk it through — including the cases where the answer isn't us.
.webp)
