For more than two decades, strategic workforce planning has been built on a fundamental assumption: organisations need a comprehensive skills taxonomy or ontology to understand their workforce.

As a result, companies have invested heavily in defining, categorising, and maintaining thousands of skills across their organisation. Entire platforms have emerged around skills inventories, skills graphs, and skills-based workforce planning.

But what if that assumption no longer holds?

The rise of large language models (LLMs) forces us to ask an uncomfortable question:

Do organisations still need a dedicated skills ontology at all?

Why Skills Taxonomies Existed

Historically, skills taxonomies solved a technology problem.

Traditional software systems could not understand that:

  • Python Development and Python Programming referred to the same capability.
  • TensorFlow was related to Machine Learning.
  • Machine Learning was a subset of Data Science.
  • Project Management and Program Coordination shared overlapping competencies.

To make workforce systems useful, organisations had to manually create and maintain these relationships.

Skills ontologies became the translation layer between people and software.

Without them, workforce data was fragmented, inconsistent, and difficult to analyse.

The Problem with Skills-Based Workforce Planning

The challenge is that skills taxonomies were designed for a world that no longer exists.

Today's organisations operate in environments where work changes rapidly, technology evolves continuously, and entirely new roles emerge every year.

A skills framework created today begins aging almost immediately.

Consider the pace of change in AI.

Five years ago, few organisations tracked skills such as prompt engineering, retrieval-augmented generation, AI workflow design, or synthetic data generation. Today they are commonplace discussions within workforce planning.

The more detailed the taxonomy becomes, the greater the maintenance burden required to keep it relevant.

Many organisations find themselves managing thousands—or even tens of thousands—of skill definitions, often with limited confidence that the data accurately reflects reality.

The result is a significant investment in maintaining a model of work that is constantly becoming outdated.

Skills Are Not the Same as Capability

Perhaps the larger issue is that skills are often mistaken for capability.

Skills are specific.

Capabilities are enduring.

For example, an organisation may require the capability to generate customer insights.

Over time, the skills used to deliver that capability will evolve:

  • Survey design
  • Statistical analysis
  • Business intelligence
  • Data visualisation
  • Machine learning
  • Generative AI

The underlying organisational capability remains relatively stable. The methods used to achieve it change. This distinction matters because strategic workforce planning is fundamentally concerned with future organisational performance, not merely current technical skills.

Leaders are typically asking:

  • What capabilities will we need in three years?
  • Which capabilities are becoming more important?
  • Which capabilities are at risk?
  • How should the workforce evolve?

These are capability questions rather than skills questions.

The LLM Changes Everything

Large language models introduce a new possibility.

For the first time, software can understand semantic relationships between skills without requiring every relationship to be explicitly defined.

An LLM already understands that:

  • Kubernetes is related to cloud engineering.
  • Cloud engineering is related to infrastructure architecture.
  • Infrastructure architecture contributes to digital platform capability.
  • Digital platform capability enables software product delivery.

The model does not need a manually curated ontology to infer these connections. In effect, much of the knowledge previously stored within skills ontologies has been embedded into the model itself. This changes the economics of workforce intelligence.

Instead of asking:

"How do we build and maintain a taxonomy that captures every possible relationship?"

Organisations can increasingly ask:

"How do we leverage AI to interpret workforce data dynamically?"

The intelligence moves from the taxonomy into the model.

From Skills-Based Planning to Capability-Based Planning

This shift suggests a different future for workforce planning.

Rather than organising workforce data around static skill inventories, organisations may increasingly focus on:

  • Work activities
  • Tasks
  • Outcomes
  • Experiences
  • Organisational capabilities

Employees become understood through the work they perform and the capabilities they demonstrate, rather than through self-reported lists of skills.

For example, an employee who has:

  • Led a digital transformation program
  • Implemented AI-enabled workflows
  • Managed cross-functional teams
  • Designed new operating processes

Demonstrates capabilities such as:

  • Change leadership
  • Process innovation
  • Technology adoption
  • Organisational transformation

An LLM can infer these relationships directly from evidence and experience.

The focus shifts from maintaining skills records to understanding organisational capability.

Does This Mean Skills Ontologies Are Dead?

Not entirely.

There are still situations where formal ontologies remain valuable. Highly regulated industries often require standardised definitions and classifications. Workforce reporting, compensation frameworks, compliance obligations, and external benchmarking may still depend on structured taxonomies.

However, their role is changing. Rather than serving as the primary intelligence layer, taxonomies increasingly become governance artefacts that provide consistency and reporting standards. The intelligence itself is increasingly generated dynamically through AI.

A New Question for Workforce Leaders

The workforce planning conversation has historically centred on understanding:

"What skills do we have, what skills do we need, and what are the gaps?"

That question made sense when skills were relatively stable and software required structured definitions to understand them. Today, a more important question is emerging:

"What capabilities does our organisation possess, and how must they evolve to succeed in the future?"

Capabilities are more durable than skills. Work is more informative than job titles. And AI increasingly reduces the need for organisations to manually encode every relationship between them.

The future of workforce planning may not be built on larger skills taxonomies. It may be built on a deeper understanding of work itself.