Zoho’s latest analytics release restructures the product around four pillars – trusted data and business context, agentic analytics and actions, intelligence everywhere, and governed AI – with native semantic and ontology layers between the data warehouse and the agents or users querying it as the architectural change carrying it. Tony Thomas, CEO of Zoho USA, positions the platform as “the intelligence layer for businesses”, a trusted substrate that both AI agents and human users read from.
A quote from Zoho’s press release:
Unlike competitors, the semantic and ontology layers are native to Zoho Analytics, and arrive in the box, not requiring a separate data platform or cloud data warehouse.
Sailas Sundaram, Head of Product Management at Zoho Analytics, briefed analysts before the embargo lifted. He highlighted that Zoho Analytics is now a single platform for data preparation, context, analytics and actions. Another important element is that it’s composable – any single layer can be adopted on its own, with the rest of the enterprise stack kept in place.
Cloud updates are available now, but preview versions of pixel-perfect reporting, the ontology and the semantic model roll out globally over the next four to eight weeks. The two headline capabilities aren’t fully in production for every customer today.
The four pillars
One of the biggest architectural changes is around trusted data and business context. Zoho Analytics already pulled from 500-plus native connectors (another 30-plus added in this release, with named additions ranging from SAP, Notion and Intercom to the Zoho ERP, Spend, FSM, Shifts and Voice apps) and offered a zero-copy connection mode for data left in Snowflake, Databricks and others. Now there are two context layers above the raw data. An ontology layer defines the canonical business entities and the relationships between them. A semantic layer sits above the ontology and defines what entities and metrics actually mean in business terms – synonyms, descriptions, metric calculations, sample questions, and the SQL that answers them. Both can be built manually or through an AI assistant, and Zoho declares compatibility with the Open Semantic Interchange specification (OSI), the vendor-neutral semantic exchange standard co-founded by Snowflake, dbt Labs, Salesforce, BlackRock and RelationalAI, which has since moved into the Apache Software Foundation incubator as Apache Ossie.
Agentic analytics and actions turns AskZia, Zoho Analytics’ natural-language assistant since 2018, into a multi-agent system. There are now six-plus pre-built sub-agents – data analysis, visualization, data preparation, modeling, coding and actions – that AskZia can orchestrate. On top of these, a new custom agent builder lets enterprises create their own agents, grounded on a chosen dataset or semantic model, with knowledge-base files, Model Context Protocol (MCP) connections and tools for executing actions.
Intelligence everywhere covers two different things – one is embedded analytics, which Zoho has sold for years to independent software vendors who white-label it in their own products; it now carries the full agentic capability set, with branding controls so an ISV’s end users never see the Zoho name. The other is headless Business Intelligence (BI) – a Zoho Analytics MCP server (remote and local), a Claude plugin and a command-line interface, so AI agents or developers in other platforms can read data, build dashboards or trigger analytics actions from outside Zoho entirely.
Michalis Stoupakis, Head of Product (Supply), Desquared commented:
One of the leading apps we own and manage has over 15,000 merchants onboarded. With Zoho Analytics, we empower every merchant with AI-powered actionable insights by contextually embedding reports and dashboards directly within the app. We are actively exploring the new agent studio functionality to build custom agents for our specific needs.
Governed AI is a unified data catalog that covers data sources, pipelines, metrics, agents and models in one view, with lineage and impact analysis across upstream and downstream dependencies.
Where this sits next to Snowflake, Databricks and Fabric
Zoho Analytics will continue as a BI and analytics product, with a dedicated data warehousing engine optimized for structured data. In parallel, the vendor is building what it calls the Zoho Data Platform, targeted for 2027, which will handle structured and unstructured data in a single layer and act as a common foundation for the Zoho application ecosystem, external third-party data, agentic workflows and data science. That puts Zoho in the same category occupied by Snowflake, Databricks and Microsoft Fabric – the common data layer underneath enterprise applications and AI alike.
For now, the emphasis is on composability and competitive pricing. Any single piece of the Zoho Analytics stack can be adopted on its own. The business context layer can sit on top of a Snowflake or Databricks warehouse using Zoho’s zero-copy connectors, pulling only metadata. The semantic layer can be consumed by an external agent via MCP, or exposed to another BI tool via OSI. An existing enterprise semantic layer can – in principle, if it’s OSI-compatible – be imported into Zoho Analytics instead of being rebuilt from scratch. Pricing for Zoho Analytics remains unchanged in this release, Zoho LLM credits are now bundled into every plan, and a Basic plan starts at $24 per month billed annually, covering two users, 0.5 million rows and 10,000 Zoho LLM credits.
My take
Zoho Analytics has had the Unified Metrics Layer since 6.0, generative-AI-infused Ask Zia for over two years, and the agentic Ask Zia plus the Analytics MCP server beta earlier in 2026. This update consolidates those pieces under the business context layer rather than sitting as separate feature announcements across quarterly updates. The broader semantic-layer category is also consolidating around Apache Ossie faster than I’d have predicted a year ago. Zoho’s choice to adopt the spec without joining the working group keeps editorial control over its own data-platform roadmap while interoperating with the ecosystem others are building.
Extraction underneath the modelling layer is the biggest challenge for enterprise AI implementations – if the entities are inconsistent acrossits that mess. The AI assistant Zoho ships helps with the modelling but extraction takes work – whatever platform you use. The data preparation sub-agent needs to be able to handle entity resolution on a messy real-world dataset
Zoho owns and operates its data centers, has no ad-revenue model anywhere in the business, and has run this way for decades without taking venture capital. An analytics platform that will host an enterprise’s business context, metric definitions and agent-trace data lives on physical infrastructure – and whether that infrastructure is controlled by the vendor or routed through a hyperscaler is something procurement teams drill into quickly in regulated industries. Few BI vendors of equivalent scale can match that ownership model, and the unchanged pricing and bundled LLM credits fit Zoho’s longer pattern rather than looking like a response to the agentic AI cycle. You can bet we’ll be on the look-out for customer use cases in the future.
