Why Ontology Is No Longer a Technical Issue – Exploring Enterprise AI’s Non‑Technical Challenges

On June 15, 2026, DataFunTalk hosted a round‑table titled “Crossing the Enterprise AI ‘Implementation Gap’ – Ontology‑Driven Agents and Knowledge Governance in Practice.” The discussion, moderated by Zheng Yan (Huawei) and joined by Zhang Sensen (Ping An) and Ma Jinlong (Qiming), focused on ontology, knowledge governance, AI transformation, and organizational change.

1. Ontology: From Technical to Management Issue

Zhang argued that ontology is essentially a re‑packaging of traditional hierarchical data modeling, now enriched with a semantic layer thanks to large models. He emphasized that the real difficulty lies not in building classes or relationships, but in deciding who has the authority to define, maintain, and arbitrate them, making ontology a management challenge.

2. The Modeling Paradox

While business teams feel the pain of modeling the most, IT teams experience less distress. Early on, ad‑hoc solutions such as LangChain‑style skill stacks were used. Over time, large‑model‑driven construction improved efficiency, yet fundamental data‑engineering problems—graph retrieval, materialized views, and unstructured data ingestion—remain hard and resource‑intensive.

3. Two “Diseases” of Enterprise AI

Zhang identified two typical company profiles: (1) firms seeking AI as a cure for business bottlenecks, and (2) firms adopting AI merely because peers are doing so. Both lead to “wish‑fulfilment AI” that often results in layoffs without genuine workflow redesign.

4. Hidden Debt Behind Highlight Projects

Rapid AI pilots create flashy front‑ends but generate massive technical debt on the back‑end, especially in governing large volumes of unstructured data. The cost of cleaning, deduplicating, and extracting information from multimodal sources outweighs the perceived benefits.

5. Steam Engine vs. Electric Motor Analogy

Zheng compared swapping a steam engine for an electric motor without changing the surrounding factory to replacing legacy tools with AI without reshaping the organization. He argued that organizational inertia is a far greater barrier than technology itself.

6. Building an AI‑Native Organization from the “Baby” Stage

Ma described his spin‑off team’s experiment: adopting AI‑first documentation (Markdown), recognizing immutable domains such as finance and legal, and keeping the architecture flexible for future model upgrades. He illustrated that a single person, empowered by AI, can handle tasks across front‑end, back‑end, testing, and algorithm development, yet the hardest problem is identifying the right business pain to solve.

7. No Silver Bullet, but Action Is Mandatory

The three speakers, coming from Huawei, Ping An, and a startup, held different definitions of ontology and AI pathways, but converged on three points: management outweighs technology, existing structures resist change, and enterprises must start moving forward despite uncertainty.