BrainBlog for Zoho by Jason Bloomberg
With the exploding frenzy around AI, then generative AI, and now agentic AI, it might seem that vendors are racing toward their AI goals in a frantic sprint, hoping to edge out their competition before all the air escapes from the AI hype bubble.
Zoho, in contrast, is running a different AI race altogether. Zoho is competing in a marathon – and they’re leading the pack.
Zoho, best known for its comprehensive suite of tightly integrated business applications aimed at companies of all sizes, is in fact an AI leader – leader in the sense of being in the right position in a race that has many more years to run.
This month’s AI news from the company shows its careful pacing in this marathon. Zoho is putting in place essential elements to help its customers succeed with AI, not only today, but well into the future.
Rightsizing the LLM
Large language models (LLMs) are at the heart of generative AI, and by extension, agentic AI. Despite this central role, however, the LLMs on the market today suffer from a few fundamental drawbacks.
LLMs, as their name suggests, are large – so large that running them on a company’s own infrastructure or in its own cloud can be prohibitively expensive.
The alternative – leaving public LLMs where they are and calling their APIs – can yield poor performance and inferior quality results, including hallucinations as well as plagiarism and privacy issues.
To address these concerns, Zoho has built its own LLM, from the ground up – in fact, it actually built three LLMs of three different sizes to address different business requirements for performance and contextual applicability.
No, Zoho didn’t start with an open-source LLM and tweak it. No, Zoho didn’t train one of its models and then pare it down to create the others.
Instead, Zoho leveraged its many years of AI expertise and its deep bench of experts to craft its LLMs starting from a blank sheet of paper – and then trained each LLM separately to yield the best results in the broadest number of situations.
Connecting the AI Dots
Zoho’s new LLMs don’t stand alone. They are a central part of the broader Zoho strategy.
This multifaceted strategy includes running its own data centers, where it controls both the hardware and software from the ground up. This infrastructure baseline enables the company to partner with companies like NVIDIA to provide cutting-edge support for AI.
The second pillar of Zoho’s strategy is its indomitable focus on customer privacy. Customers have trusted Zoho with their sensitive data since day one, and the company has always lived up to this promise.
Zoho’s focus on privacy, in fact, underlies its LLMs, which respect the permissions and other privacy and security controls in place across all customer data. And of course, Zoho never uses customer data to train its models.
The third pillar of Zoho’s strategy is how it approaches integration. Not only does its suite of business apps integrate seamlessly with each other as well as numerous third-party apps, but Zoho also provides applications like Zoho Flow that offer advanced integration capabilities.
To this breadth of integration assets Zoho is now adding support for the model context protocol (MCP), a popular an open standard and corresponding framework that standardizes how LLMs integrate with external tools and data sources.
In fact, Zoho users can now create MCP servers of their own with a single click, leveraging a diverse application library across several applications, with more applications to come.
Adding AI Agents to the Marketplace
Zoho Marketplace offers thousands of extensions and integrations for Zoho users. The company is now adding its Agent Marketplace to this broader marketplace offering.
The Agent Marketplace includes several pre-built AI agents that work seamlessly with various Zoho and third-party apps. For example, Zoho now offers a revenue growth specialist agent for analysts, a deal analyzer agent for salespeople, and for HR professionals a candidate screener – to name a few.
Remember, however, that Zoho is running a marathon. The agents it has made available today represent a small fraction of the agents it has in the works. We can expect future agents to be both more diverse and powerful as Zoho progresses.
The Intellyx Take
I’ve covered the highlights of Zoho’s news, but there is more to the story. Zoho is rolling out new automatic speech recognition models for both English and Hindi. The company is making improvements to its Ask Zia pre-built agent that handles tasks across different Zoho apps. It’s also making improvements to its low-code Agent Studio.
Given this plethora of innovations, it can be hard to get one’s head around everything Zoho is doing with AI.
In the grand scheme of things – considering the entire marathon Zoho is running – today’s news is steps in the right direction. Better models, better integration, better agents – all incremental improvements on the way to a long-term vision of AI empowerment for businesses of all sizes.
Copyright © Intellyx BV. Zoho is an Intellyx customer. Intellyx retains final editorial control of this article. No AI was used to write this article.
