Why Companies Must Learn Ontology When Building AI Operations: It’s Not Just an IT Department Matter

Originally published in Chinese on HK01 on 2026-01-06 07:00 | By Michael C.S. So | AiX Society

AI is not an efficiency tool — it is a fundamental rewrite of enterprise collaboration logic. From OpenAI’s The State of Enterprise AI 2025 to the on-the-ground gaps in Hong Kong businesses.

With ChatGPT now serving over 800 million users per week, the significance of that number has long surpassed “tech adoption” itself. What it truly represents is an irreversible flywheel effect: when consumers already treat AI as a daily tool, companies that still regard AI as an experimental project face a problem that is no longer about whether employees are willing to use it, but whether the organization itself is ready to be reshaped.

In recent years, companies have been eagerly investing in artificial intelligence (AI) technology, hoping to boost efficiency, optimize decision-making, and create new value. However, many companies make a fatal mistake: treating AI implementation as a purely IT project, handing it entirely to the IT department. This approach often results in AI projects becoming pilot-stage showcase initiatives that never truly integrate into daily operations, let alone translate into tangible business outcomes.

To avoid this “high-tech, low-implementation” dilemma, business leaders must confront one of the most important foundational elements behind AI — Ontology.

What Is Ontology in AI?

Simply put, ontology is a systematic knowledge model used to define concepts, attributes, and relationships within a specific domain. Think of it as a machine-readable knowledge map that helps AI understand our language, logic, and decision-making structures.

Take the retail industry as an example. “Customer,” “Order,” and “Product” are all core concepts. Ontology doesn’t just list these terms — it goes further to describe their relationships (for example: a customer places an order, an order contains products) and the attributes of each concept (a product has a price, inventory levels, etc.).

This isn’t merely a database structure or a glossary. It is a method for enabling AI to understand “business logic.” It gives semantic meaning to the data behind the scenes, transforming AI from a simple data analyzer into an intelligent system capable of reasoning, understanding context, and making judgments aligned with business reality.

Why Does AI Need Ontology?

Avoiding misunderstandings and erroneous judgments: Without clear definitions, “customer” could be misinterpreted by AI to include test accounts or inactive users, leading to distorted reports or misguided marketing strategies.

Promoting data consistency and integration: Business units, finance, and customer service may each define “revenue” or “valid order” differently. Ontology provides a unified standard, enabling AI to understand and integrate data across departments.

Enhancing reasoning and intelligent decision-making: If AI understands that “high-value customers” are also linked to “customer service tickets,” the system can automatically flag those tickets as priority items, improving service efficiency and customer satisfaction.

Breaking down data silos and fostering cross-departmental collaboration: Ontology gives different departments a common language for AI projects, eliminating the disconnect where IT speaks in terms that business teams can’t understand.

Improving AI explainability and transparency: When AI decision-making logic is grounded in a visualizable ontology model, companies can more easily track, audit, and optimize AI behavior, reducing risk.

What Do You Miss When You Hand AI to the IT Department?

Many companies leave AI projects entirely to IT leadership, overlooking the fact that AI actually involves business process reengineering, personnel training, and organizational culture change. The result is often one of the following scenarios:

Business doesn’t participate, so AI can’t address the real problems: The IT department doesn’t understand operational details, and often builds models that are “technically correct but operationally useless.”

Processes remain unreformed, limiting AI’s impact: If legacy processes and job designs are used to interface with AI, many automation and intelligent analytics capabilities can’t be implemented — and may even cause process chaos.

Employee transformation and training are neglected: AI implementation changes job responsibilities, but without simultaneous HR restructuring and employee training, the result is resistance, inefficiency, and attrition.

Cultural disconnect causes AI rejection: When AI is led by IT without cross-departmental communication, other departments tend to view AI as an “external threat,” lacking buy-in and cooperation.

Underestimating AI deployment costs and timelines: Relying solely on IT to drive AI tends to underestimate the process redesign and workforce adjustment costs required for true AI implementation, causing project timelines and expected benefits to fall short.

How Should Companies Build AI Operations?

AI is not an IT project — it is an enterprise operations reengineering initiative. Here is a practical framework to help companies establish a more strategically oriented AI operations model:

1. Build Ontology as the Foundation for AI

• Collaborate with domain experts from each department to jointly define key concepts and business logic.
• Convert this knowledge into machine-readable ontology models that serve as the foundation for AI learning and reasoning.
• Regularly update the ontology to reflect business changes, maintaining the accuracy and relevance of AI decision-making.

2. Redesign Business Processes and Responsibilities

• Don’t just “automate” old processes — rethink how to leverage AI to design more efficient and responsive new processes.
• For example, customer service departments can use AI for predictive support, transforming into proactive service teams.
• Sales teams can restructure their customer engagement workflows alongside AI recommendation systems to improve conversion rates.

3. Implement HR and Change Management Strategies

• Establish AI training programs to enhance employees’ AI literacy and tool proficiency.
• HR departments need to redesign job roles, performance evaluation mechanisms, and career development paths, making AI an enabler for employees rather than a competitor.
• Cultivate cross-functional talent — roles like “AI Strategists” and “Data Translators” who understand both AI technology and operations — to strengthen departmental collaboration and decision-making speed.

4. Establish a Cross-Departmental AI Operations Team

• Include representatives from IT, business, operations, HR, and other departments to jointly advance AI projects, ensuring alignment between technology and practice.
• Regularly host AI application workshops that encourage internal innovation and AI idea-sharing, giving frontline personnel a voice in the AI design process.

5. Strengthen AI Governance and Ethics Frameworks

• As AI permeates operations, companies should establish AI ethics guidelines to ensure algorithms are fair, non-discriminatory, and transparent.
• Develop review processes for sensitive operations (such as HR screening and financial risk management) to prevent black-box AI decision-making.

Stop Handing AI to Just One Department

AI is not a tool — it is a complete reconstruction of how a business operates. The key to successful AI implementation lies not in how powerful the algorithms are, but in whether your company is prepared: from building semantically clear ontologies, to restructuring processes and responsibilities, to transforming employee mindsets and capabilities.

When you understand that AI is not just technology but a convergence of knowledge, processes, and people, AI can truly become your operational asset and competitive advantage.

Remember: what companies need is “AI with operational value,” not “AI that’s only good for demos.”

This is precisely the greatest value that ontology and cross-departmental collaboration can deliver. The future competitiveness of enterprises will depend on whether they possess AI that can be operationalized — no longer just technological capability, but the ability to integrate knowledge, organization, and culture.