Zoho unveils Zia LLM & AI suite with strong data privacy focus

Zoho has announced the launch of its proprietary large language model, Zia LLM, alongside a range of additional artificial intelligence offerings aimed at supporting both developers and end users with enhanced data privacy and operational efficiency.

In its latest announcement, the company outlined a suite of new AI tools and infrastructure investments that include Zia Agents, a no-code Zia Agent Studio, a model context protocol (MCP) server, and an Agent Marketplace featuring over 25 ready-to-deploy AI-powered agents. These developments are positioned to deliver improved workflows and financial efficiencies for organisations across various business functions.

Strategy and focus

Mani Vembu, Chief Executive Officer at Zoho, highlighted the company's objectives for its AI portfolio. He said, "Today's announcement emphasizes Zoho's longstanding aim to build foundational technology focused on protection of customer data, breadth and depth of capabilities, and value. Because Zoho's AI initiatives are developed internally, we are able to provide customers with cutting-edge tool sets without compromising data privacy and organizational flexibility, democratizing the latest technology on a global scale."

Zia LLM details

The newly introduced Zia LLM comprises three models with 1.3 billion, 2.6 billion, and 7 billion parameters respectively, each optimised for specific user contexts and trained using Zoho product use cases, including structured data extraction, summarisation, retrieval-augmented generation (RAG), and code generation. The company stated that the models benchmark competitively against similar open source models available on the market.

Zia LLM is based on NVIDIA's AI accelerated computing platform and is designed for deployment across Zoho's data centres in the US, India, and Europe. The company is currently testing Zia LLM internally and expects it to become available to customers in the coming months. As part of its commitment to data privacy, Zoho emphasised that Zia LLM enables users to keep their data on Zoho's own servers without transmitting it to external AI cloud providers.

Additionally, Zoho announced two new automatic speech recognition (ASR) models for English and Hindi, optimised for accuracy and low computational load. The company stated these models showed up to a 75% improvement over comparable alternatives in standard benchmarks. More language support is planned for future updates.

AI agents and marketplace

The company has made available a selection of AI agents directly integrated within its products, designed to handle a range of business tasks related to roles including sales development, customer support, and account management. Notable examples include the new version of Ask Zia, which features enhanced business intelligence (BI) skills tailored for data engineers, analysts, and scientists; and a Customer Service Agent for Zoho Desk, capable of processing and triaging customer requests.

The Zia Agent Studio has been updated to provide a prompt-based, no-code interface for users, giving access to over 700 actions across Zoho products. Agents built within this studio can be configured as digital employees with defined access permissions, and administrators have tools available for behavioural auditing and performance analysis.

Zoho's Agent Marketplace, now a dedicated section within the existing extension marketplace, allows customers to browse and deploy pre-built agents. The company listed several of these initial agents, including:

• Revenue Growth Specialist – identifies upsell and cross-sell opportunities for existing customers and suggests marketing strategies;
• Deal Analyser – evaluates business deals and provides actionable insights and recommendations;
• Candidate Screener – ranks and shortlists candidates for job openings based on experience and skills.

Zoho stated it will continue to expand the catalogue of available pre-built agents to address additional core and utility use cases.

Model context protocol and interoperability

The introduction of Zoho's MCP server is intended to enable interoperability with third-party clients, providing access to a library of actions from more than 15 Zoho applications at the early access stage. Integration with Zoho Flow extends reach to further external tools, and Zoho Analytics now supports a local MCP server as well.

Planned developments

The company detailed short-term and longer-term plans for its artificial intelligence ecosystem. Upcoming developments include scaling Zia LLM's model sizes, expanding speech-to-text support for more languages, particularly across Europe and India, and the release of a reasoning language model (RLM). Additional skills will be incorporated into Ask Zia for finance and customer support teams, and agent-to-agent (A2A) protocol support will be implemented to enable collaboration between Zia Agents and agents on other platforms.

Data privacy principles

In outlining its differentiation, Zoho reiterated that its AI models are not trained on consumer data and do not retain customer information, maintaining that utility and right-sizing are key design principles. The company noted that it does not have an ad-revenue model and operates its own data centres to ensure privacy and security for all users.

"Today's announcement emphasizes Zoho's longstanding aim to build foundational technology focused on protection of customer data, breadth and depth of capabilities, and value. Because Zoho's AI initiatives are developed internally, we are able to provide customers with cutting-edge tool sets without compromising data privacy and organizational flexibility, democratizing the latest technology on a global scale."

The new artificial intelligence products, including Zia LLM, Zia Agents, Zia Agent Studio, Agent Marketplace, and the MCP Server, are in early access for selected customers, with general availability planned for the end of 2025. Zoho intends to publish pricing at the time of full release.