Onome Amuge
Zoho, a global technology company with a growing footprint in Africa, has unveiled a suite of artificial intelligence tools including its own proprietary large language model (LLM), an agent-building platform, and a marketplace for pre-built agents.
The launch, announced on Thursday, features Zia LLM consisting a family of custom-built large language models designed for business use cases, alongside Zia Agent Studio, a no-code environment to build AI agents, more than 25 deployable agents, and a Model Context Protocol (MCP) server that allows third-party systems to interact with Zoho’s software.
“Today’s announcement emphasises Zoho’s longstanding aim to build foundational technology focused on protection of customer data, breadth and depth of capabilities because of the business context, and value. Our LLM model is trained specifically for business use cases, keeping privacy and governance at its core, which has resulted in lowering the inference cost, passing on that value to the customers, while also ensuring that they are able to utilise AI productively and efficiently,” said Kehinde Ogundare, Zoho Nigeria’s country head.
Unlike many business software vendors who rely on third-party generative AI providers such as OpenAI or Anthropic, Zoho said it has built Zia LLM entirely in-house using NVIDIA’s AI-accelerated computing platform. The company added that it has trained three models with 1.3 billion, 2.6 billion and 7 billion parameters, each optimised for different contexts such as structured data extraction, summarisation, retrieval-augmented generation (RAG), and code generation.
Zoho executives said the strategy allows the company to right-size models for different user contexts, balancing performance with efficiency. The company plans to scale its model sizes further, with new parameter increases expected by the end of 2025.
While Zoho continues to support integrations with external LLMs including ChatGPT, Llama and DeepSeek, the proprietary approach reflects its emphasis on data privacy. Unlike many cloud-hosted AI models that require customer data to leave the vendor’s servers, Zia LLM keeps all data within Zoho’s infrastructure, appealing to companies with strict compliance requirements.
The company also announced two proprietary automatic speech recognition models for English and Hindi, claiming performance gains of up to 75 per cent over comparable benchmarks. Additional language support will follow, broadening adoption across emerging markets where Zoho has sought to differentiate itself. A reasoning language model is also announced to be in development.
Central to Thursday’s launch is Zia Agent Studio, a no-code and low-code platform for building AI agents tailored to enterprise functions. The studio supports prompt-based development and provides access to more than 700 pre-built actions across Zoho’s products.
Users can create agents to automate business processes, deploy them autonomously, or integrate them into customer-facing workflows. Once deployed, agents can function as digital employees with role-based permissions, while administrators retain the ability to audit behaviour, monitor performance, and enforce governance rules.
Pre-built agents already available include a candidate screener for recruitment, a deal analyser for sales teams, and a revenue growth specialist to identify upselling opportunities. These are accessible via an Agent Marketplace, with Zoho planning to open the platform to independent developers and ISVs in the coming months.
