Understands how data relates
The model maps customers, vendors, products, accounts, and other business entities across the systems where they appear.
News · August 2026
Our new ontology model helped Permute rank first on UC Berkeley’s Data Agent Benchmark. It gives AI agents a structured map of business data, so they can understand how records, definitions, and systems relate before answering a question.

The benchmark
UC Berkeley’s Data Agent Benchmark evaluates agents on complex, real-world data tasks. The questions require agents to work across different databases, reconcile inconsistent references, interpret unstructured text, and apply domain knowledge.
Those are the same conditions businesses face when customer, product, operational, and financial data are spread across systems that were never designed to work together.
The ontology model
An ontology is a structured description of the entities, relationships, definitions, and rules inside a business. It can identify that two differently named records represent the same customer, explain how a subsidiary relates to a parent company, or define which version of revenue a report should use.
Our model uses that map to give an agent the right context for each task. Instead of handing the agent a collection of tables and asking it to infer the business, Permute provides the relationships and reviewed definitions it needs to reason over the data.
The model maps customers, vendors, products, accounts, and other business entities across the systems where they appear.
Business rules and metric definitions are applied the same way across questions, reports, dashboards, and agents.
Answers can retain the source records, definitions, and transformations used to produce them, making review easier.
Teams can connect fragmented operational and financial systems without replacing the tools that already run the business.
See it on your data
Connect the systems behind one reporting or analytical workflow and see how governed business context changes the quality of the answers.