Cloud data platform Snowflake has introduced Business Ontology, a new capability designed to improve how artificial intelligence agents interpret and use enterprise data. The tool will allow organisations to establish consistent business definitions across applications and departments, addressing concerns around the reliability of AI-generated responses.
The capability forms part of Snowflake's Horizon Catalog, which supports data discovery, management and governance. It builds on Horizon Context, introduced earlier this year to help enterprises organise metadata and provide AI systems with information about business data and its relationships.
Business Ontology will enable organisations to define business concepts, assign ownership and connect approved definitions to underlying data tables and views.
For instance, departments may use different calculations for metrics such as net revenue or customer value. Without standardised definitions, AI agents accessing the same enterprise data could generate inconsistent interpretations.
Snowflake said the new capability is intended to provide agents with business-approved definitions rather than relying entirely on automatically inferred meanings.
Previously, enterprises could use Snowflake's Cortex Sense to identify business context from metadata and usage patterns. However, these interpretations may not always reflect definitions formally accepted by an organisation.
With Business Ontology, employees will be able to review and approve these meanings before they are used by AI applications.
Industry analysts said the approach could simplify AI agent development by reducing the need to incorporate separate business logic into individual applications.
Ashish Chaturvedi, Executive Research Leader at HFS Research, indicated that centrally maintained definitions could allow developers to reuse business context across multiple AI agents.
Michael Ni, Principal Analyst at Constellation Research, also highlighted the potential to reduce inconsistencies between agents interpreting the same information.
However, maintaining accurate definitions could remain challenging as business processes, pricing structures and reporting requirements change.
Different departments may also require separate interpretations of common terms, making governance and coordination important.
Alongside Business Ontology, Snowflake is expanding its Cortex AI Gateway capabilities to provide greater oversight of AI agent activity.
The additions include Cortex Sandboxes, which allow agents to execute custom code within isolated environments, and Agent Trace Intelligence, which enables developers to examine how agents perform tasks using enterprise data.
Snowflake is also introducing Agent Abuse Detection to identify potentially inappropriate agent behaviour, including attempts to bypass approval procedures or access sensitive information.
While sandboxing can limit the impact of faulty code, analysts cautioned that it does not automatically guarantee the accuracy of AI-generated results or prevent misuse of authorised access.
Business Ontology is expected to enter public preview soon. Cortex Sandboxes, Agent Trace Intelligence and Agent Abuse Detection are currently in private preview.
The developments come as enterprises examine how to deploy AI agents across business functions while maintaining consistent data interpretation, operational oversight and governance.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.