Today, more businesses than ever before are relying on AI to get work done.
The rise of vibe coding has meant that entire software products are now developed through natural language prompting. Finance teams are relying on AI models to reconcile transactions, while sales and marketing workflows are increasingly automated.
“We’re doubling down on being a system of record, not a collection of AI features.”
Chandrashekar Lalapet Srinivas Prasanna
But for all AI’s usefulness, there are also risks. Few know this better than Chandrashekar Lalapet Srinivas Prasanna—known simply as LSP—Managing Director of Zoho Canada.
Zoho offers a cloud suite of over 60 software tools for businesses of all sizes, and the company has seen AI adoption surge for its customers since launching its large language model, Zia, last year.
The company received 16 billion AI API calls in the first half of 2025 alone, which represents a 50 percent increase year over year.
In recent months, Zoho has thought carefully about how to implement AI into its software to deliver maximum value to its customers with minimal risk. BetaKit spoke with LSP about the three principles guiding Zoho’s approach to AI in 2026.
Keep the risk where it belongs
According to LSP, the tradeoffs with AI for businesses aren’t immediately obvious. AI tools can write, code, summarize, and even analyze, but no model creators can guarantee their outputs will be accurate or reliable all of the time.
“Generative AI introduces a paradox,” LSP said. ”It promises efficiency, yet it often pushes technical and operational risk back onto the customer.”
Many of the largest AI companies publicly acknowledge they cannot guarantee their outputs will be error-free. It’s a departure from what businesses typically expect from traditional SaaS products, where if a product breaks, it’s on the vendor, not the customer, to make things right.
Zoho believes that the rise of AI and vibe coding has created the need for an authoritative, centralized application that acts as the source of truth for an organization. “We’re doubling down on being a system of record, not a collection of AI features,” LSP added.
When companies build AI features on top of third-party infrastructure, they can’t control what happens under the hood. Zoho avoids that route by owning every part of its Zia large language model, from the data infrastructure layer up to the application layer. This gives Zoho control over what Zia can do and the data it is trained on. It also means that if something goes wrong, Zoho won’t pass the blame to a third-party provider.
“[Our customers are] operating inside a platform where we take responsibility for reliability, auditability, and outcomes,” LSP added. “In other words, we don’t hand customers experimental tools and wish them luck. We operationalize AI inside the core system so the risk stays with us, where it belongs.”
Feed AI the full picture
Perhaps the most visible demonstration of Zoho’s approach to AI is the latest iteration of its core software suite for businesses: Zoho One. The biggest advantage of the recently refreshed platform, according to LSP, is that it is now an integrated operating system, with Zoho’s AI embedded into every single workflow.
“Instead of moving between apps, the work comes to the user,” LSP added. “Tasks, data, conversations, and agents all converge in a single operational environment. It’s a fundamental shift from using many tools to running the business from one place.”
Zoho One gives customers access to one interface, one workflow layer, one search and analytics layer, and finally, an AI layer that understands the full context of the business.
“AI is only useful when it operates inside [a] unified business context. Chatbots fail because they sit outside the system of record.”
Zoho One allows businesses to manage all of Zia’s capabilities in one place. Here, they can set permissions, control data access, audit AI actions, and configure Zia agents. This layer also handles predictions, anomaly detection, enrichment, and automation, which every app in Zoho One can draw from. This means that Zoho is turning AI into a built-in layer across the stack, rather than a feature that sits on the side.
LSP believes the value of AI increases dramatically when it has full context across tools. Rather than AI in an email system that only sees emails, or AI in an analytics tool that only sees reports, Zoho wants to embed AI across these workflows, which allows it to draw on data from all of its products and get a more useful view of a business.
“AI is only useful when it operates inside [a] unified business context. Chatbots fail because they sit outside the system of record,” he added. “They don’t understand the data, the workflows, or the permissions.”
Embrace the chaos
In many industries, AI still feels like the Wild West, but Zoho doesn’t believe that fact should push businesses away from adoption.
“Zoho has always believed that chaos is a catalyst for clarity,” LSP said. “It forces us to rethink assumptions and find new opportunities. In the AI era, that mindset is essential.”
This year, LSP is focused on empowering Zoho’s Canadian team to experiment with new AI workflows, test unconventional solutions, and explore new use cases. All of this happens under tight operational guardrails, so experimentation doesn’t disrupt the customer experience.
“We embrace chaos to spark ideas, but we execute with precision so customers experience stability, not turbulence,” LSP added. “That’s how we stay entrepreneurial while delivering the value we promise.”
PRESENTED BY
Meet Zia, Zoho’s AI assistant built to automate tasks, deliver insights, and boost productivity in your everyday business operations.
Feature image courtesy Unsplash. Photo by Steve Johnson.