Key Takeaways

  • Zoho Analytics now gives AI agents semantic definitions and mapped business relationships rather than access to raw data alone.
  • Ask Zia coordinates 6 specialist sub-agents, while administrators gain lineage, activity monitoring, sandbox testing and rollback controls.
  • Pricing remains unchanged, with plans now including LLM credits and adopting Apache Ossie to support portable semantic definitions.

Zoho Corporation has expanded Zoho Analytics into an agentic analytics environment, adding an ontology layer, semantic modeling and a multi-agent version of Ask Zia. Announced on October 28, 2026, the release addresses a persistent enterprise AI problem: language models can query data, but they often lack the business context needed to interpret it correctly.

A semantic layer defines terms and metrics such as net revenue, active customer or gross margin. An ontology maps relationships among customers, accounts, contracts, orders and service cases. Together, they give an AI agent a governed representation of how a business operates.

When a sales executive asks for high-value customers, a model working directly from tables might rank accounts by revenue alone. With an approved semantic definition, it can instead calculate margin after discounts. Once an ontology is added, the system recognizes that one customer owns several accounts, those accounts have contracts, and the contracts may connect to unresolved support escalations. The agent's output becomes highly contextual and more actionable.

According to Deloitte, 30% of organizations were exploring agentic AI and 38% were piloting it in 2025, but only 14% had deployable solutions and 11% were using them in production. Governance, data quality and operational controls dictate much of that gap between exploration and production readiness.

Zoho integrates these controls into an analytics product already used daily by 100,000 businesses. Business analysts can build and edit semantic models without code, and several models can operate side by side. Finance and operations teams can preserve definitions suited to their respective workflows while sharing an underlying data environment.

Ask Zia acts as the coordinator, planning requests and delegating tasks to 6 specialist sub-agents covering analysis, visualization, data preparation, modeling, coding and actions. All 6 use the same semantic and ontology layers. Customers can use the platform's native model or bring a frontier model, reducing the degree to which business meaning relies on a single language model.

Developers receive local and remote Model Context Protocol servers, a command-line interface and a plug-in for Anthropic’s Claude. Custom agents can call external systems through the Model Context Protocol, extending Zoho Analytics beyond its own interface.

To mitigate the risk of agents acting on flawed definitions, administrative features include a data catalog with lineage, records showing which agents ran and what information they accessed, and a sandbox with rollback capabilities. These controls make testing and investigation more practical during enterprise deployment.

Data availability remains a primary requirement. TechTarget reported that Deloitte found 48% of organizations cited data searchability and 47% cited data reusability as challenges to AI automation strategies. To address this, the company added more than 30 connectors, including SAP and Adobe Commerce, to an existing catalog of more than 500. An ontology is only effective when the underlying data can be reached and governed.

The release adopts Apache Ossie, formerly Open Semantic Interchange, for describing datasets, metrics, fields and relationships in YAML or JSON. The dbt Core framework supports Ossie files beginning with version 1.12. The current specification is version 0.1.1, with version 0.2.0 in development, requiring enterprise buyers to test exports, imports and security extensions carefully.

Cloud customers receive the core updates immediately, while previews of the ontology and semantic model are scheduled to reach all regions within 4 to 8 weeks. The on-premises release is planned for the first quarter of 2027. Pricing remains unchanged: the Basic plan costs $24 per month for 2 users and includes 10,000 Zoho LLM credits.

When piloting agentic analytics, data teams can compare agent output before and after semantic definitions are applied, monitor credit consumption, inspect lineage and test an Ossie round trip. While the software provides the necessary infrastructure, the system's accuracy ultimately depends on establishing internal agreement over what the business terminology actually means.