Original Title: Ontology: A Repricing
Although they all use the term "ontology," they are not necessarily selling the same thing; rather, it represents a commercial repricing of existing capabilities in the AI era. Over the past six months, almost every vendor dealing with AI and data has been talking about the same word: Ontology. Data platform and BI (Business Intelligence) vendors are talking about it, knowledge graph vendors are talking about it, Agent application vendors are talking about it, industry intelligence vendors are talking about it, and some model vendors are also talking about it. However, comparing their claims reveals a phenomenon: no two vendors' ontologies are the same thing. Some call it a semantic layer, some call it an AI data foundation, some call it a dynamic ontology, and some call it AI FDE... Although all called "ontology," they may be selling completely different things. This does not look like a market forming a consensus, but rather a spontaneous collective action—everyone is using the term, not to unify understanding, but because it makes their existing capabilities more valuable and easier to sell.
01. Old Capabilities, New Pricing
Over the past few months, iFenxi has conducted in-depth interviews with numerous vendors that have adopted ontology as their strategic direction. When these vendors' roadmaps are placed side by side, a clear migration path from "old capabilities to new pricing" emerges. The vendor types differ, but the operational logic is consistent.
Figure 1: New Terminology and Pricing Changes for Ontologies Across Different Vendor Types
Data platform and BI vendors used to make a living by helping enterprises manage data well—aggregating, cleaning, building data warehouses, and generating reports. But in the AI era, the consumers of data have shifted from humans to Agents. Vendors have redefined their capabilities as a semantic layer—no longer just making the data middle platform more user-friendly, but turning it into a standard interface for Agents to consume data.
The pricing anchor has changed accordingly: previously, license fees were charged based on functional modules and user counts; now, pricing can be based on the completion results of AI tasks. The same set of governance capabilities generates different value, creating entirely different pricing potential. Agent application vendors used to make a living by doing projects. Each project required Forward Deployed Engineers (FDEs) to be on-site to understand the business, integrate data, configure models, and develop applications. The ceiling for human efficiency was bottlenecked by human resources. In the AI era, they have consolidated solution templates and business rules from projects into Skills, SOPs, or SDKs, continuously reusing and iterating them in new projects.
With the same number of people, output has climbed from hundreds of thousands to millions. Previously, they were selling headcount; now, they are selling reusable AI FDE digital capabilities. Ontology provides implementation- and delivery-heavy Agent application vendors with a completely new valuation logic. The traditional business of knowledge graph vendors was turning unstructured documents into usable data. Through self-developed models, they performed document parsing, table extraction, knowledge extraction, and graph construction, ultimately delivering knowledge base projects. These projects had thin profit margins and were highly substitutable. Now, they have upgraded to AI data infrastructure. While it is still about turning documents into knowledge, the consumer of that knowledge has become AI. The pricing anchor has shifted from knowledge base projects to an AI-oriented data foundation.
The capabilities remain unchanged, but they have repositioned themselves from knowledge governance vendors to AI infrastructure suppliers. The accumulation of industry intelligence vendors lies in industry chain data, including enterprises, products, technologies, patents, talents, and upstream/downstream relationships. Previously, they mainly sold to governments and industrial parks for analytical reports; now, it is called a dynamic ontology. The same data assets: originally priced by how much a report costs, now priced by how much an AI-callable industry decision-making platform costs.
Model vendors, in the past, focused on post-training models in the enterprise service sector, improving Agent application effects by building enterprise-specific models. Now, model vendors have moved a step closer to the core business of enterprises, upgrading specific models to business ontologies and directly building Agent applications for enterprises. Ultimately, they also charge based on the actual business value of the applications. Six types of vendors, six types of businesses, using the same word. Therefore, the real function accomplished by "ontology" in the Chinese market is more like repricing. Each vendor uses it to tell customers one thing: What I sold you in the past is now subject to a new set of commercial pricing logic in the AI era.
02. Why the Word "Ontology"?
It is no coincidence that a single word can be simultaneously co-opted by six completely different types of vendors.
First, the Palantir endorsement. This U.S. company, with a market capitalization of 300 billion U.S. dollars, has Ontology as its core concept, realizing enterprise data objects, business logic, and actions through it. "We benchmark against Palantir" and "We are doing the localization of Palantir" have become sales pitches that carry a natural premium.
Second, it is abstract enough. Ontology can refer to the semantic layer, the action layer controlling device operations, or AI modeling based on ontology. The concept is so abstract that no single vendor can monopolize its definition, allowing everyone to use it according to their own understanding. Precisely because there is no standard definition, it is hard for anyone to question, "This is not an ontology."
Third, real demand provides a foundation. After AI enters enterprises, the biggest threshold is understanding the real business, and ontology is precisely the most suitable solution. The need to understand the business is real, and ontology happens to be the most appropriate word to describe it. Therefore, the "ontology" hype is not a bubble because the demand is real. However, it is also not a technical consensus, because it is just vague enough to be adopted by everyone to wrap their business in a completely new commercial story.
Figure 2: Why the Word "Ontology"?
03. The Real Threshold: Not Telling a Good Story, but Delivering Well
If ontology is essentially a rhetoric for commercial repricing, then the true watershed lies not in who tells the story best, but in who can get customers to renew. This points not to storytelling and sales capabilities, but to delivery and organizational capabilities. The most critical variable is the fate of FDEs in the Chinese market.
FDEs are essentially the last mile from the ontology concept to a runnable system at the customer site. No matter how beautifully the model is designed, if it still requires a massive amount of human effort for data cleaning, semantic mapping, and business logic configuration at the customer site, then ontology is not much more advanced than an ordinary implementation project. Chinese vendors have taken two distinctly different paths on this issue.
One path is using AI to replace FDEs. The logic of these vendors is: anything that requires on-site human work must be gradually encapsulated into AI-executable capabilities. Pre-sales solutions are generated by AI, data governance is automatically completed by models, and project management is taken over by digital employees. The number of FDE personnel does not increase but decreases, and the organization is compressed into a minimalist form of a few people leading a group of AI employees. This is an aggressive path: betting that FDE costs can be reduced to zero by AI.
The other path is using AI to amplify FDEs. The logic of these vendors is the exact opposite: FDEs are not cost centers, but barrier centers. They arm frontline personnel with AI, enabling one person to leverage the transformation and implementation of multiple backend business systems, doubling human efficiency without downsizing. FDEs are not objects to be replaced, but core competencies amplified by AI. They believe that only those who understand the industry and the customers can truly make the ontology generate value at the customer site. AI merely makes these people more efficient.
Two paths, the same destination: Can ontology transform from a one-time project into a renewable service?
Figure 3: Two Paths for FDEs
iFenxi judges that the true winners in this movement are not those who tell the best ontology stories, but the vendors that first solve three engineering problems: minimizing the reliance on humans for delivery, transforming one-time projects into renewable services, and getting customers willing to pay again in the second year.
04. Back to the Origin
The Chinese enterprise service market has never lacked new concepts. From data middle platforms to lakehouse architectures, from enterprise general models to industry vertical models, every era has its own buzzword. They all ultimately face the same test: whether customers actually renew. Ontology currently stands at the same crossroads. The window of opportunity will not be very long.
When everyone uses the same word, it is no longer a differentiating advantage, but a ticket to enter the game. What truly sets vendors apart is not who adopts the word "ontology" first, but who first turns it from a concept in a PPT into renewal revenue on the customers' books.