A fifty-person distributor can model how its customers, contracts, orders and service tickets connect inside the same tool it uses for dashboards, and finish that work before an AI agent answers a single question about any of them. Zoho announced today in Austin that Zoho Analytics is becoming an agentic analytics platform. The release adds a semantic layer that defines your terms and metrics, an ontology layer that models your business entities and their relationships, and a multi-agent version of the Ask Zia assistant that hands work to specialist agents grounded in both layers.

Zoho kept pricing the same and now includes large language model credits in each plan. Cloud customers get the updates today, previews of the ontology and semantic model reach all regions over the next four to eight weeks, and on-premises customers receive the release in the first quarter of 2027.

Give your agents an ontology before you give them questions

I consider ontology one of the most important parts of getting value from AI, and businesses skip it often, because entity modeling feels like housekeeping and gets pushed behind the chatbot demo. My Infotech colleague Igor Ikonikov always tells companies to start with a good data foundation. The ontology supplies context: when someone asks a question, the system can match the intent behind it to the right information, because it knows how the pieces of the business relate.

Picture a sales leader who asks an agent for her best customers this quarter. An agent working from raw tables ranks accounts by revenue in whatever table it finds first. Give it a semantic layer, and "best" means margin after discounts, the definition finance approved. Add an ontology, and the agent also knows that a customer owns several accounts, each account carries contracts, and open service tickets attach to those contracts, so it can flag a high-margin customer who sits three escalations away from leaving.

The same language model produced every one of those answers, and the foundation underneath it made the difference.

Your business teams own that foundation as much as your data team does, which makes the ontology a business project with a technical component.

Zoho moved the glossary into the analytics tool your teams already open

In many large enterprises, data engineers maintain semantic definitions in a separate layer next to a cloud data warehouse. Zoho ships both layers inside Zoho Analytics, with no separate data platform required, and businesses can build and edit semantic models without code. You can run several semantic models side by side, so finance and operations each keep the definitions that fit their work. For me, this placement is the most exciting part of the release, because it puts the glossary where business users already work.

Everything else in the release sits on those two layers:

• Ask Zia as a coordinator. A central agent plans the work and hands it to six sub-agents for analysis, visualization, data preparation, modeling, coding and actions, all reading the same semantic model and ontology (Zoho, 2026). You can use Zoho's own model or bring a frontier model.
• Custom agents. Teams can build their own data agents grounded in the same definitions and connect them to outside tools through the Model Context Protocol, the open standard AI agents use to call external systems.
• Headless business intelligence. Developers get a Model Context Protocol server in remote and local versions, a new command-line interface, and a plug-in for Anthropic's Claude, so people can query Zoho Analytics from outside its own screens.
• Oversight for administrators. Administrators get a data catalog with lineage, a view of which agents ran and what data they accessed, and a sandbox with rollback for testing before anything reaches production.

A midsize company already running Zoho CRM or Zoho Books can reach grounded agents through a product it already licenses, and a business analyst can own the definitions instead of waiting for a dedicated data team. Zoho also added 30-plus connectors, including SAP and Adobe Commerce, to the 500-plus it already offered (Zoho, 2026). Those connectors matter because an ontology needs the data it describes to reach the platform first.

A business analyst can own the definitions instead of waiting for a dedicated data team.

Zoho chose an open standard so your definitions can travel with you

After your teams spend months agreeing on what "active customer" and "net revenue" mean, those definitions are worth more than any single dashboard you build from them. Zoho says compatibility with Apache Ossie keeps them portable and outside Zoho's walls, and the company commits to following the specification as it matures. I think Zoho made the right call by adopting the standard this early.

Ossie, formerly called Open Semantic Interchange, is a vendor-neutral format for describing datasets, fields, metrics and relationships in one YAML or JSON file. dbt Core reads Ossie files starting with version 1.12 (dbt Labs, 2026), and the project's latest released specification is version 0.1.1, with version 0.2.0 in development (Apache Ossie, 2026).

Build a round-trip export into your pilot so you can see how your access rules travel with your metrics. In Ossie's Power BI converter, details such as row-level security rules sit in custom extensions alongside the core specification, according to a technical review by element61, a data consultancy. I hope more vendors follow Zoho in shipping open-format definitions, because a business that writes its glossary once and reuses it across tools gets more value from the work.

Start your ontology now, because the previews land within eight weeks

Previews of the ontology and semantic model reach all regions over the next four to eight weeks, which makes November a good month to pick your first business question. Choose one with a known right answer, and you can measure how much the agents improve once the model goes in.

The entry price keeps a first experiment small. The Basic plan at $24 a month includes 10,000 Zoho LLM credits for two users (Zoho, 2026), and you can use the new usage and cost tracking to see how credits flow as more teams join, so you plan a department-wide rollout with real numbers in hand.

Zoho's roadmap builds on the ontology you start this year

• Zoho plans to grow today's hand-built context into "dynamic business context" that observes how your data changes over time. Decide now who in your organization will review the definitions it proposes.
• Zoho Analytics will anchor an upcoming Zoho data platform with a lakehouse for structured and unstructured data, and the ontology you build this year becomes your starting point for that platform.
• Regulated teams that want agents running inside their own data centers can plan around the on-premises release in the first quarter of 2027. Ask Zoho which agent features arrive with it.