Every organisation I work with wants better visibility of skills.
That’s the right call – workforce planners can’t plan against roles that no longer describe the work. Reward teams can’t price capability they can’t see. L&D can’t target investment against a workforce it has never properly mapped.
Something has changed in how work gets done, and the old proxies of job titles, grades and competency frameworks written five years ago often no longer provide enough information to act on. So organisations reach for skills: granular, portable and supposedly more accurate than a job title.
That’s a reasonable diagnosis, but the work then drifts.
The drift doesn’t start immediately – a skills project starts with a real question:
Who can do what, and where are the gaps that matter?
Within a few months, it becomes a taxonomy project:
- Which skill belongs under which category?
- How many proficiency levels should there be?
- Is stakeholder management a skill, competency or behaviour?
- Should the taxonomy map to a public standard or be built from scratch?
Don’t get me wrong – none of this work is inherently illegitimate. Some form of taxonomy is usually necessary if an organisation wants to aggregate and compare skills data at scale. The problem is when building the taxonomy becomes a substitute for deciding what the organisation needs the data to do.
The warning signs are usually visible in the project plan when it contains milestones for the framework: skills defined, levels validated, and taxonomy approved. What it misses are the decisions that should change once the framework exists.
Simply put, no-one says what the organisation will do differently.
This isn’t a data-quality problem but a sequencing one. Taxonomy work expands because it is bounded, measurable and relatively comfortable. You can tell whether a taxonomy has been completed but it’s way harder to determine whether a workforce decision has improved. So the harder question gets deferred, and the framework begins to consume the programme.
The test I use with clients is often considered blunt:
Does this change a real workforce, service or performance decision?
Not whether it could, in principle, after enough additional work. Which decision will change because this information exists?
Three questions usually expose the difference.
Which decision is this for?
“Workforce planning” is not a decision. A decision might be:
- whether to redeploy people into a particular service
- whether to build, buy or borrow a specific capability
- who could succeed someone in a critical role
- where development investment should be concentrated
- whether work should be redesigned rather than additional people recruited
If you can’t name the decision, you are not yet doing useful skills work. You are classifying information and hoping a use for it appears later.
Who makes that decision now, and what do they use?
The answer may be manager judgement, historical precedent, job titles, personal networks or incomplete information. Often the decision is not made systematically at all. That is really useful to know. It tells you what the skills data must improve upon. But if nobody understands the current decision-making process, a taxonomy will not solve the problem. It may simply document the absence of a process in greater detail.
What must be true of the data for someone to act differently?
This is the question many taxonomy projects skip. It requires the organisation to work backwards from the decision to the information required, rather than forwards from a comprehensive framework towards whatever uses might eventually emerge. The minimum viable data is often smaller, coarser and quicker to assemble than the project team expects. A decision about redeploying people into one service may not require an enterprise-wide taxonomy covering every role. A succession decision may require evidence of capability in a few areas, rather than a complete organisational skills inventory. The necessary taxonomy should follow the decision, not determine the programme. If a skills initiative can’t answer these three questions for at least one live decision, it is not yet a skills-first strategy. It is a classification exercise. Classification may improve how an organisation describes its workforce. That is useful, but it is a smaller achievement than changing how the organisation deploys, develops or rewards people.
This is so much more urgent now – AI-supported skills tools make this drift easier.
Vendors can infer potential skills from job descriptions, CVs, learning records and other organisational data at a scale that would previously have required extensive manual work. That might reduce the time needed to establish a starting point. It can also make a comprehensive and authoritative-looking taxonomy appear to be progress in itself. The organisation receives thousands of skills, neatly categorised and mapped but it still may not know which decision any of them should change.
There are also limits to what can be inferred from the available data. A job description may indicate what someone is expected to do, not what they can do. A CV records selected past experience, not necessarily current capability. System activity may show use, but not judgement, quality or performance.
More data does not remove the need for decisions about evidence, confidence and context. Skills-first is a genuine reframe of workforce strategy. It is not just another term for better categorisation.
The more credible approach is to begin with a decision the organisation is making badly, identify the capability information that would improve it, and build only as much taxonomy and infrastructure as that decision requires. Otherwise, the organisation may end up with an impressively structured account of what people might be able to do, and no evidence that it has changed what anyone does next.
If you are scoping a skills initiative, get in touch, and we can talk through worth pressure-testing it against a live decision before commissioning the taxonomy.