It has not been an integration more or a simple turn to the market rhythm: Zoho Corporation has decided to strengthen its roadmap in artificial intelligence with the development of its own language model, a battery of agents ready for deployment and a complete environment without code for its personalization. The company has made public a series of novelties that not only seek to win in functionality, but also differentiate from the nucleus: the entire ecosystem has been designed, trained and optimized internally.
The main protagonist is Zia LLM, a large -sized language model built by Zero from scratch, with three versions (1,300, 2,600 and 7,000 million parameters), and specifically trained with the cases of use of its more than 55 applications. The model will be deployed in Zoho data centers in the United States, India and Europe.
In the words of Sridhar Iyengar, general director of Zoho in Europe, “shows Zoho’s long -term objective in building fundamental technology focused on the protection of customer data.” To Iyengar, the key is that “we democratize access to the latest technology without compromising privacy or organizational flexibility”
Next to the LLM, the company launches more than 40 Native AI agents already available in the new Zia Agent Marketplace. They are designed for specific tasks in key departments such as sales, customer service or personnel selection, and can be implemented in minutes on Zoho applications. From the “Income Specialist” to the “candidate coach”, the variety of use cases is already extensive … and will grow.
The technology company redoubles its commitment to artificial intelligence with its own tools that prioritize privacy, organizational control and immediate adoption in business environments
The other great novelty is Zia Agent Studio, an environment of creation without code (although with low-code option) that allows any user to design custom agents with access to more than 700 predefined actions in the Zoho ecosystem. These agents can act as digital employees respecting the access rules, and can be audited and monitored by administrators.
Conversational intelligence, interoperability and road map
Another strategic axes is the evolution of Ask Zia, the conversational assistant that is now expanded with functions of Business Intelligence specific for technical profiles, such as analysts or data scientists. For its part, the MCP (Model Context Protocol) server of Zoho will allow third parties to connect its own agents with data and actions within the Zoho ecosystem, always maintaining the permits defined by the organization.
In the coming months, Zoho plans to expand the size of its ZIA LLM models, add more languages to voice recognition and launch a new reasoning model (RLM). In addition, it will introduce compatibility with the Agent2agent protocol (A2A), to allow agents to communicate with each other and even those of other platforms.
Early access and general availability
For now, Zia LLM and the rest of the solutions are being released for customers on the waiting list in the early access program. The general availability is scheduled by the end of 2025, at which time the final price structure will also be known. Until then, Zoho will analyze use patterns by sector, region and company size, with the objective of adjusting its value proposition to a global and diverse customer base.
