Every HR team has watched this happen. Someone good leaves for a job they could have done internally, and nobody knew they could do it. The skills were there. The system didn't know.
We've just finished this year's Gartner Hype Cycle for AI in Human Resources — 94 minutes of it, by Gartner's own estimate. Partly because ClayHR is named in it, mostly because it's a useful annual check on which AI ideas in HR are real and which are still promises. 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.
Career pathing software has existed for years. Show an employee their current role, show them roles above and beside it, draw the arrows. Add AI and the arrows get smarter — the system suggests moves you wouldn't have thought of, based on people with similar profiles.
The technology works. The problem is what it's working from.
Ask most HR systems what an employee can do and you'll get some combination of their job title, a self-selected list of skills they picked from a dropdown two years ago, and whatever their last manager wrote in a review. None of that is a capability record. It's a rough guess, assembled from convenience, and the confident career recommendation built on top of it inherits every gap.
This is why career pathing pilots disappoint in a specific way. They don't produce obviously wrong answers. They produce plausible ones — suggestions that look reasonable and aren't grounded in anything, which is harder to catch and worse for trust. An employee gets a recommendation, senses it doesn't fit, and stops opening the tool.
The traditional answer to career development was time. Spend three years doing the work, absorb the expertise, move up. The path was implicit in the sequence of jobs.
That sequence is thinning out. A lot of the entry-level work that used to build expertise — first-draft research, routine analysis, basic production — is now assisted or automated. The output still happens. The learning that used to come with producing it doesn't, or at least not the same way.
Which leaves organizations with a gap: people need to advance, but the rungs that used to carry them are less reliable. Someone has to work out what capability a person actually has, and what specifically is missing between that and the next role, because it can no longer be assumed from years served.
That's a measurement problem before it's a recommendation problem.
Four things, in this order:
Most tools do the third thing well and assume the first two. That's the wrong way round.
We built our talent capabilities around the measurement problem rather than the recommendation problem.
Skill levels update themselves from evidence. Rather than relying on what people declare, ClayHR updates skill levels automatically from AI assessments, performance reviews, completed learning programmes and skill-linked feedback. Self-assessment still exists — employees can rate themselves and see where they stand — but it's an input, not the record.
The Skill Matrix makes capability visible. A colour-coded view mapping competencies across roles, teams and projects, so strengths and gaps show at organisation level rather than one employee at a time.
Gaps connect to learning. Identifying a gap is only useful if something happens next. Skill gaps link directly to targeted training, so a career path becomes a set of actions rather than a diagnosis.
Performance and potential together. Configurable 9-Box and 4-Box matrices segment people by performance and potential, filterable by role and business unit — the bridge between individual career pathing and organisational succession planning.
Knowledge doesn't leave with people. Employees approaching a transition can document what they know through AskHR, so successors inherit context rather than starting cold.
Skills-based career pathing is only as good as the effort put into the job architecture underneath it, and that work is unglamorous. Someone has to define what roles actually require, and keep doing it as roles change. Organizations that skip this and go straight to the AI layer get exactly the plausible-but-ungrounded recommendations described above.
There's also an honesty cost. A system that names capability gaps precisely will sometimes tell people they aren't ready for something they want. That's more useful than vague encouragement, but it needs managers who can have the conversation.
And if your organization is small enough that leadership genuinely knows who can do what, you may not need any of this yet.
If you're working through how to make career development real rather than aspirational, we're happy to talk it through.
.webp){kind=small}
