As organizations move from the promise to the practice of skills-based HR, a massive effort is underway to build, maintain, and operationalize enterprise skills taxonomies. This is not a one-time project but an ongoing data infrastructure challenge that requires continuous investment, cross-functional coordination, and sophisticated technology to manage.
Two years after skills-based hiring began as a mainstream HR trend, companies are discovering that the real work — and the real value — lies in the messy, granular work of mapping what their workforce can actually do.
## Why Skills Taxonomies Matter Now
The urgency around skills taxonomies has been driven by three converging business needs:
**Internal mobility:** As the internal talent marketplace market has exploded (now over 60% of Fortune 500 companies have deployed platforms), the quality of skills data has become the single most important factor in match quality. Poor skills data leads to poor recommendations, which leads to low adoption, which leads to platform underutilization.
**Workforce planning:** As organizations face restructuring, restructuring, and strategic pivots, knowing the actual skills of your workforce — not just the job titles on your org chart — is critical for making informed decisions about where to invest, where to retrain, and where to hire externally.
**AI-driven HR:** Modern AI talent platforms require structured skills data to function effectively. An AI-powered recruiting tool that can predict candidate success or a performance management system that can recommend development pathways both depend on a foundation of accurate, comprehensive skills information.
## Approaches to Skills Taxonomy: Four Models
Companies are taking four distinct approaches to building their skills taxonomies:
**1. Buy a commercial ontology.** Many organizations start by purchasing a pre-built skills ontology from a vendor likeESCO (European), O*NET (U.S. Department of Labor), or OneModel (a commercial skills intelligence platform). These ontologies provide a comprehensive starting point with tens of thousands of skills mapped to standard categories. The downside is that they are designed for broad applicability, not for the specific vocabulary and competencies of your organization. [Source: O*NET Resource Center, “Skills Framework 2025 Update”](https://www.onetcenter.org/skills.html)
**2. Build in-house.** Some organizations, particularly in technology and professional services, build their own skills taxonomies from scratch. This approach captures organization-specific vocabulary and competencies but requires significant investment in subject matter expertise, data collection, and maintenance. Microsoft’s internal skills taxonomy, for example, was built over two years by a team of 15 HR analysts and 200 subject matter experts across the company.
**3. Hybrid approach (most common).** Most organizations adopt a hybrid model: starting with a commercial ontology as the base and enriching it with organization-specific skills through surveys, manager input, and analysis of job descriptions and performance data. This approach balances comprehensiveness with relevance. According to a 2026 SHRM survey, 68% of large organizations use a hybrid approach to skills taxonomy development. [Source: SHRM, “Skills-Based HR Survey 2026,” May 2026](https://www.shrm.org/resourcesandtools/hr-topics/talent-acquisition/pages/skills-based-hr-survey.aspx)
**4. AI-generated ontologies.** The newest approach leverages large language models to generate skills taxonomies from unstructured data sources — job descriptions, performance reviews, LinkedIn profiles, learning records. Companies like Eightfold AI, Gloat, and Paradox all offer AI-powered skills extraction capabilities. Early results are promising but imperfect: AI-generated ontologies capture emerging and non-traditional skills but may misclassify or duplicate entries. Human validation remains essential. [Source: Eightfold AI Research, “AI-Generated Skills Taxonomies: A Practical Guide,” April 2026](https://www.eightfold.ai/blog/ai-skills-taxonomies)
## Key Challenges in Skills Taxonomy Management
**Skills proliferation.** As organizations add skills to their taxonomies, the number of skills grows rapidly. A well-maintained enterprise skills taxonomy typically contains 3,000 to 8,000 individual skills. Managing the metadata (synonyms, related skills, hierarchy, proficiency levels) at this scale is a significant data management challenge.
**Skills inflation.** There is a natural tendency for job descriptions to include more and more skills over time, leading to “skills creep” where the baseline expectations for a role become unrealistic. Regular auditing of skills requirements against actual performance data helps mitigate this.
**Skills decay.** In a fast-changing workplace, skills become obsolete at an accelerating rate. A 2025 study by the World Economic Forum found that the half-life of a professional skill is now approximately 5 years, down from 10 years a decade ago. Skills taxonomies must be dynamic, not static.
**Skills data quality.** Self-reported skills data has known accuracy problems: employees tend to overstate their skills, and skills that are rarely used tend to be forgotten. The most accurate skills taxonomies combine self-reporting with assessment data, learning records, and manager observations. [Source: World Economic Forum, “Future of Jobs Report 2025,” October 2025](https://www.weforum.org/reports/the-future-of-jobs-report-2025/)
## The Technology Stack for Skills Data
A mature skills data infrastructure includes:
– **Skills ontology database:** A structured repository of skills with hierarchical relationships, synonyms, and metadata
– **Skills extraction engine:** AI-powered tools that analyze job descriptions, performance data, and learning records to identify and classify skills
– **Skills assessment platform:** Tools that validate skills through tests, work samples, or peer validation
– **Skills reporting dashboard:** Analytics and visualization tools that make skills data actionable for HR and business leaders
– **Skills integration layer:** APIs that connect the skills infrastructure to HRIS, ATS, LMS, and talent marketplace systems
## Looking Ahead
The skills taxonomy landscape is moving from a data management exercise to a strategic competitive advantage. Organizations with high-quality, real-time skills data can make better hiring decisions, more accurate workforce plans, and more effective development investments than those without. The companies that treat skills data as a strategic asset — with dedicated ownership, ongoing investment, and clear business use cases — will be the ones that genuinely realize the promise of skills-based HR.