When you ask AI, “How much did revenue grow last month?”, which figure should it use? The answer changes depending on whether revenue means total order value or net revenue after cancellations and refunds. When data sits across several systems, even a metric with the same name may be calculated differently. Even if AI finds the right number, the answer is hard to use unless it understands what that number means in the business.
NVIDIA recently released Auto Ontology as open source. NVIDIA explains that enterprise AI needs to understand what company data means in a business context before it can use that data at work. Auto Ontology organizes data terms, relationships, and sources across systems so AI can query information using verified business definitions. Its release shows how much attention enterprises are giving to the meaning behind their data.
The same figure can lead an employee to different conclusions depending on the customer, operating conditions, previous actions, and the task at hand. Enhans works with teams to connect the data AI retrieves with the conditions they use to interpret it and make decisions.
How NVIDIA Auto Ontology Defines the Meaning of Data
Auto Ontology uses enterprise databases and metadata to map where data resides, how it connects, and which definitions apply. NVIDIA describes a process in which business experts review, refine, and approve the generated ontology. When AI creates a query, it can refer to those verified definitions and data sources and check whether the query follows them.
Consider a request for “revenue” that involves several tables and calculation methods. AI first needs to know which metric to use. Does it include canceled orders? Which date determines the reporting period? Where did the figure come from? Answering those questions gives the number a basis the business can check.
NVIDIA has also made this context available to AI through interfaces including MCP. When a company connects AI to its data, it needs a way to check the meaning and source of the results. That allows AI to explain a figure in terms the business uses.
Expert Judgment Is Not Always in the Records
Just as a revenue figure depends on cancellation rules and reporting dates, an equipment record depends on the conditions under which it was created. Two machines may show the same abnormal signal, yet different operating conditions or maintenance histories may call for different checks. A maintenance record may list the part replaced and the result, while leaving out why an experienced technician suspected that part first or ruled out other causes.
When AI finds a similar case in past records, the employee handling the issue still needs to decide whether it applies to the equipment in front of them. That requires the original symptoms and action taken, along with the conditions behind the decision. Which signals did the expert check first? What led them to change their assessment? When did the situation call for another inspection or a colleague's review?
To capture those criteria, it helps to examine a real case and ask what the expert observed, when their assessment changed, and how they chose the final action. The recorded facts, the expert's judgment, and any exceptions need to be identified separately.
Enhans Examines Where Expert Judgment Changes
Enhans Forward Deployed Engineers (FDEs) review operational data alongside the signals and conditions employees use to make decisions. If there is no clear answer to “Do you always replace the part when this signal appears?”, they ask which operating conditions and inspection results the employee checks before deciding. They also check whether employees mean the same thing when they use the same term.
In the field, teams need to define whom to consult when an exception does not fit the documented criteria. They also need to check whether the knowledge they have organized is reflected in data queries and answers, and how to reassess earlier answers when the criteria change. That is why business experts remain involved after the initial terms and relationships have been defined in an ontology.
Enhans connects equipment conditions and maintenance histories with the operating conditions and decision criteria identified by experienced engineers. In AgentOS, these criteria can support a workflow in which AI retrieves past records and the responsible employee reviews the results. When AI presents a similar case and its original conditions, the employee can decide whether it applies to the current equipment and what to inspect or do next.
Check Whether Business Context Shapes the Answer
Organizing business knowledge does not guarantee that AI will apply it every time. If a data description conflicts with the underlying values, or a verified criterion is missing from a query, an answer may sound convincing while drawing the wrong conclusion. Teams should also check whether answers still follow old criteria after those criteria have changed.
When using an ontology in business operations, a company should be able to check:
- Data meaning: Do the metrics and records AI used match the definitions employees use?
- Decision conditions: Do the criteria and exceptions behind the answer fit the current situation?
- Changes to criteria: When an employee updates a criterion, can the team check the resulting queries and answers again?
These checks help the responsible employee decide whether to use an AI response, look for more information, or make the judgment directly. To use an answer at work, they need to see which data and business criteria informed it.
Enterprise AI Needs Business Context Maintained with the People Who Use It
With tools such as NVIDIA Auto Ontology now available to organize the meaning of enterprise data, companies need to decide who reviews the resulting criteria and where they apply. When AI presents a past equipment issue, the responsible employee should be able to check the operating conditions and reasoning behind the earlier action before deciding what to inspect next.
Teams also need to decide who updates the criteria when a new exception appears or a business rule changes, and who checks the AI's answers afterward. This is why Enhans connects employee judgment to data and checks whether that context is used in actual work. Even after an ontology is built, employees need to be able to review and update its criteria for AI responses to be useful in their work.
in solving your problems with Enhans!
We'll contact you shortly!