Authored by: Steve Goldberg — HR Tech Industry Analyst, Advisor, Influencer
Abstract: As AI becomes embedded in almost all HR technology offerings, competitive advantage will come less from the AI model itself and more from the quality of the ontology and human capability data beneath it.
If you spend time in HR or HR technology circles today, it will not take long before you hear the word ontology. Vendors describe ontologies as the foundation for matching talent, powering recommendations, enabling agents, orchestrating workflows, and creating intelligence layers across the enterprise. The excitement is justified. Ontologies matter, but there is a question that deserves far more attention: What makes one ontology better than another?
As AI becomes embedded in every aspect of talent management and optimizing human capital, organizations will discover that two AI systems can have access to similar enterprise data yet produce quite different recommendations. The difference will not be the AI itself, but the quality of the knowledge and capability architecture behind it.
What is an ontology?
An ontology is simply a structured representation of skills and knowledge and the relationships between concepts. In the HCM or talent management domain, an ontology might connect skills, knowledge, roles, experiences, behaviors, competencies, performance indicators, proficiency levels, etc.
The promise is compelling. When AI understands not only these individual elements but also how they relate to one another, it can generate more relevant, impactful and in-context recommendations, more accurate talent insights, and more personalized development guidance.
But all ontologies are not created equal.
An oft-cited point of view on ontologies — one I happen to agree with — is that the best ontologies are comprised of multiple layers of data from multiple sources. They may include external labor data or internal organizational data, as well as workforce insights. The latter includes evolving talent management insights resulting from employees and contingent workers using corporate systems to find suitable projects and opportunities.
Organizations are then able to use the ontology to identify and prioritize the most relevant and critical skills, update employee and job profiles, rank candidates and determine candidate sourcing strategies, and optimally deploy talent. As these applications continue to expand (plus adapt and learn), the underlying ontology increasingly becomes the engine powering enterprise talent decisions.
Different ontologies evolve to support different purposes. For example, talent acquisition-focused ontologies emphasize labor market data, candidate matching, and the prioritization of needed skills. L&D-centric ontologies focus on the skills most required to support different business functions and how employees can develop or acquire these skills. Employee-centric ontologies often support internal mobility, career planning, scheduling, and workforce planning.
Regardless of their purpose, the aforementioned question remains. How should we judge the quality of an ontology?
To distinguish one ontology from another start with these three questions:
1. Where did the knowledge come from?
Every ontology reflects the underlying sources of knowledge/intelligence and expert, continuously tested perspectives. How does it continually evolve in meaningful ways? How is it effectively governed? What is the source of its construction, and how is it updated? The quality of AI recommendations can never exceed the quality of the knowledge held by its sources. Fine-tuning models will not fix a lack of inherent quality.
2. How deep is the underlying domain expertise?
Depth often matters more than breadth. For context, imagine two AI systems with access to similar employee data and comparable language models. One consistently produces more actionable and trustworthy recommendations. Why? The answer lies in the underlying domain expertise powering all critical intelligence within the ontology itself.
Does the ontology simply catalog skills and job titles, or does it also represent the relationships between those factors and human performance, potential, motivation, development, team dynamics, etc. One is simply better than the other because it better understands the relationships between a myriad of factors and nodes.
3. Can it explain it recommendations?
Explainability is becoming essential. Managers, employees, and HR professionals increasingly expect to understand not only what a recommended or prescribed course of action is (such as a development plan, a candidate selection recommendation, or high-potential identification), but also why it was made in the first place.
Recently, Deloitte Insights suggested that organizations should design ontology frameworks beginning with desired business outcomes rather than simply organizing around skills. It is a thoughtful perspective. Starting with outcomes such as organizational agility, workforce productivity, innovation, or becoming an employer of choice helps ensure that ontology design remains connected to strategic priorities rather than becoming an academic exercise.
Ultimately, outcomes still depend on accurately representing the human factors that drive performance. Organizations achieve agility because people possess the capabilities, motivations, relationships, judgment, and leadership required to adapt successfully. Human dimensions are the foundation of any meaningful HCM or talent ontology.
Conclusion
Employees do not arrive at work as a collection of skills. They bring their knowledge, skills, attributes, motivations, experiences, aspirations, and personalities. These human dimensions then operate inside complex organizational systems where leadership, relationships, team dynamics, pressure, development opportunities, and performance expectations continuously interact. An ontology that cannot represent this complexity will struggle to generate meaningful intelligence and guidance about people; and in the age of AI, intelligence is increasingly determined by the quality of the knowledge layer beneath the interface … so all ontologies are simply not created equal.
About the Author:
Steve Goldberg has operated in senior roles on all sides of HR, HCM, Talent Management and HR Tech for over three decades and on three continents. After 15 years as a practitioner exec in Fortune 500’s, Steve led HCM product strategy at PeopleSoft, co-founded Recruiting Tech and Change Management firms, and directed HCM research practices at Bersin and Ventana Research. Now serving as an independent industry analyst and advisor, Steve has been recognized as a Top 100 HR Tech Influencer multiple years and has been engaged by 60+ solution vendors. He holds an MBA in HR.
