The Ontology Trap
There is a pattern that repeats across industries, geographies, and company sizes. An organization decides to become “skills-based.” A project team is formed. A technology platform is selected. Months of effort go into building a comprehensive skills ontology: defining skills, organizing them into hierarchies, mapping them to roles, and populating them with employee data.
And then nothing happens.
The ontology exists. It is thorough. It may even be accurate. But the business decisions it was supposed to inform continue to be made the same way they were made before: through gut instinct, personal networks, and the same managers advocating for the same people they have always advocated for.
Chen Wei, who led a skills transformation at a 60,000-person technology company, described the realization that changed his approach: “We spent eighteen months building what I still believe was an excellent skills ontology. But when the CEO asked us to identify 200 people who could staff a new AI business unit, we could not answer. We knew what skills people had. We did not know what that meant for any specific business decision.”
This is the ontology trap: confusing the raw material with the finished product. A skills ontology is a data asset. It describes what exists. Workforce intelligence is an analytical capability. It prescribes what to do. The gap between the two is where most skills initiatives stall.
Skills Data Is an Input
To understand why the ontology alone is insufficient, consider an analogy from a different domain. A hospital maintains detailed records of every patient’s vital signs, lab results, imaging scans, and medical history. This data is essential. Without it, no diagnosis or treatment plan is possible.
But the data itself is not a diagnosis. A diagnosis requires a physician (or an AI system) to interpret the data in context, weigh multiple signals against each other, consider the patient’s specific circumstances, and arrive at a recommendation for action. The raw data is an input to that process, not the output.
Skills data works the same way. Knowing that an organization has 340 people with cloud architecture skills, 1,200 with project management skills, and 85 with regulatory compliance expertise is information. It becomes intelligence only when it is combined with context:
- Which of those 340 cloud architects are in roles where they are underutilized?
- How many of the 1,200 project managers have experience specifically with technology implementations vs. construction projects vs. marketing campaigns?
- Are the 85 regulatory compliance experts distributed across the geographies where new regulations are being introduced, or concentrated in a single region?
- How do these numbers compare to what the external labor market can supply, at what cost, and on what timeline?
Each of these questions requires skills data plus something else: organizational data, project data, geographic data, labor market data, financial data, or strategic planning data. The skills ontology provides one essential ingredient. The intelligence engine combines it with all the others.
The Transformation Layer
Between raw skills data and actionable workforce decisions, there must be a transformation layer. This layer takes the inputs (skills, organizational structure, strategic priorities, labor market conditions, financial constraints) and produces outputs that business leaders can act on.
The table below illustrates what this transformation looks like in practice:
| Raw Skills Data (Input) | Additional Context | Workforce Intelligence (Output) |
|---|---|---|
| 412 employees have data engineering skills | New data platform initiative requires 60 data engineers by Q3; current project allocations show 38 are available | Recommendation: redeploy 28 available data engineers immediately, initiate accelerated upskilling for 15 adjacent-skill employees, and begin external hiring for remaining 17 with estimated 90-day fill time |
| Skills gap identified in cybersecurity across 3 business units | Regulatory deadline in 8 months; external market shows 40% salary premium for certified professionals; internal L&D has a 12-week certification program | Risk assessment: 2 of 3 business units can close the gap through internal development if started within 30 days; the third requires 6 external hires at premium compensation, budget impact of $840K |
| Post-merger skills overlap analysis shows 70% redundancy in finance function | Integration timeline, retention risk scores, performance data, location consolidation plan | Integration plan: retain top-quartile performers across both organizations, consolidate 3 locations to 2, redeploy 45 finance professionals into business partner roles where skills gap exists, estimated annual savings of $6.2M |
In each row, the skills data is necessary but not sufficient. The intelligence comes from combining it with business context and producing a specific, actionable recommendation with quantified trade-offs.
Five Capabilities of a Workforce Intelligence Engine
Based on how the most advanced organizations are using skills data today, a workforce intelligence engine needs five core capabilities that go beyond what a skills ontology alone can provide:
1. Contextual matching, not label matching. When a business leader needs to staff a project, the intelligence engine should match people to work based on demonstrated capability and contextual fit, not based on whether they have the right skill labels in their profile. Amara Osei, VP of Talent at a global logistics firm, described the shift: “We stopped asking ‘who has this skill’ and started asking ‘who has done work like this before.’ The results were dramatically better because we were matching on actual experience, not self-reported tags.”
2. Forward-looking forecasting, not just current-state reporting. A skills ontology tells you what exists today. An intelligence engine should project what will be needed in six, twelve, and twenty-four months based on strategic plans, market trends, technology adoption curves, and attrition patterns. Static snapshots of current capability are useful for compliance but inadequate for planning.
3. External market integration. Internal skills data in isolation is like a company’s financial statements without any market benchmarks. Workforce intelligence requires continuous integration of external signals: what skills are growing in demand, what compensation premiums are emerging, where geographic talent pools are deepening or thinning, and how competitor hiring patterns signal strategic moves.
4. Scenario modeling and trade-off analysis. Business leaders rarely face binary decisions. They face trade-offs: build vs. buy, speed vs. cost, risk vs. reward. The intelligence engine should enable scenario modeling that quantifies these trade-offs. “If we invest $2M in upskilling, we close the gap in 9 months. If we hire externally, we close it in 4 months at $5M. If we do both, we close it in 6 months at $4M. Here are the risk profiles for each option.”
5. Decision surfaces, not dashboards. The final output of a workforce intelligence engine should not be a dashboard full of charts that require interpretation. It should be a decision surface: an interface that presents specific options, quantified trade-offs, and recommended actions. Dashboards inform. Decision surfaces enable.
The Architecture of Intelligence
Building this transformation layer requires a specific architectural approach. The intelligence engine sits between the data layer (which includes the skills ontology, HRIS data, financial data, and external market feeds) and the decision layer (where business leaders and HR partners interact with recommendations).
| Layer | Components | Function |
|---|---|---|
| Data Layer | Skills ontology, HRIS, ATS, LMS, project management systems, external labor market feeds, financial planning data | Collects and structures raw workforce signals from all available sources |
| Intelligence Layer | Inference engine, matching algorithms, forecasting models, scenario simulator, market comparator | Transforms raw data into contextual, actionable insights |
| Decision Layer | Recommendations engine, trade-off visualizer, action workflows, integration with planning tools | Presents intelligence in formats that enable specific business decisions |
The intelligence layer must be a distinct, purpose-built capability. It cannot be an afterthought bolted onto an HRIS. The data integration complexity and modeling sophistication demand a system designed for this purpose.
The Organizational Shift
Technology alone does not complete the transformation. Organizations that successfully move from skills ontology to workforce intelligence also make an organizational shift: they embed workforce intelligence into business planning processes rather than keeping it confined to HR.
Tomoko Hayashi, CHRO of a global manufacturing company, described what this looks like in practice: “Our workforce intelligence team does not sit in HR anymore. They sit in corporate strategy. They participate in every capital allocation discussion, every M&A evaluation, every market entry decision.”
This positioning determines the questions the system answers. When workforce intelligence lives in HR, it answers HR questions: how many people have this skill, what is our training completion rate. When it lives in the business, it answers business questions: can we execute this strategy, what is the talent risk to this initiative, and where should we invest next.
The Path Forward
The progression from skills ontology to workforce intelligence is not a technology upgrade. It is a reconception of what skills data is for. The ontology is the foundation. The intelligence engine is the building. And the decisions it enables are the rooms where people actually live and work.
Organizations still building their ontology should continue, but with clear eyes about what comes next. The ontology is step one. The transformation layer is step two. And organizational integration that puts workforce intelligence at the center of business decisions is step three.
Getting the ontology right matters. But it is not the destination. It is the starting line.
A skills ontology tells you what your people know. Workforce intelligence tells you what to do about it.
Key terms
Skills data is the raw material. Workforce intelligence is the refined product. The organizations that generate real business value from skills are the ones that build the transformation layer between the two, connecting capability data to business context, labor market signals, and decision frameworks that leaders actually use.