Zoho Canada’s managing director on the early decisions that can leave startups with fewer options as they grow.
As Canadian startups race to integrate AI into their products and services, many are relying on third-party models and infrastructure to power their businesses.
These technologies can help companies develop products faster and with fewer upfront costs, but they also introduce risks that may become difficult to manage as those businesses grow.
“A lot of startups are wiring their entire product to one AI provider and treating that as a permanent decision.”
LSP, Zoho Canada
AI model providers can raise prices, change their terms of service, or introduce competing products of their own. For startups that have built their businesses around a particular provider, those decisions can threaten profit margins, undermine their competitive advantage, and leave them facing costly technical challenges.
Chandrashekar Lalapet Srinivas Prasanna, known as LSP, the managing director of Zoho Canada, worries that startups are becoming increasingly vulnerable to decisions made by their technology suppliers.
“You end up competing against the company that supplies a core part of your product,” LSP said. “That’s difficult if most of what you offer is an interface or a light workflow on top of their model, because they can often ship something similar with better economics.”
According to LSP, startups building products around third-party AI models risk losing control over their operating costs and product development. He believes founders need to consider those dependencies before they become too expensive to undo.
When the supplier calls the shots
A June report from RBC noted that three US tech companies now account for 85 percent of Canada’s cloud infrastructure spending. In its National Artificial Intelligence Strategy, Canada’s federal government warned that Canadian businesses building on technology largely controlled outside the country are “competing on infrastructure they do not own, govern, or meaningfully influence.”
“A lot of startups are wiring their entire product to one AI provider and treating that as a permanent decision,” LSP said. “It feels smart in the moment because you move faster and spend less on engineering. The problem shows up later.”
LSP’s concern is vendor lock-in, where a company becomes so dependent on one technology supplier that switching to another becomes difficult or expensive. As startups develop their products, their prompts, data retrieval systems, and methods for evaluating AI performance can become closely tied to the requirements of a particular model.
“If your context structures, evaluation tests, and tool abstractions are already locked to one provider’s formats, you will only discover how expensive it is to leave when you actually need to leave,” he said.
Vendor lock-in also carries financial risks. LSP warned that higher supplier costs could erode a startup’s profit margins, particularly if switching would be prohibitively expensive. He also questioned how startups can protect their competitive advantage when their core technology can be accessed by competitors.
“An interface that just calls the same model everyone else can call is hard to defend,” LSP added. “Investors notice when the core capability can be replicated by competitors or by the model provider itself.”
The price of true independence
Zoho has taken a relatively expensive route to technological independence. The company, which offers a suite of cloud software applications for businesses, operates data centres in multiple regions, including Canada, and has invested in developing its own AI models rather than relying entirely on third-party providers. The strategy gives Zoho greater control over its infrastructure and the direction of its products.
“Owning the relevant parts of the stack gives us the ability to adapt without being forced into someone else’s timeline or terms,” LSP said.
But operating data centres requires significant investment in physical infrastructure and technical staff. Zoho also accepts that competitors using external models can sometimes introduce new capabilities faster.
“The biggest trade-off is the sustained investment in capital, people, and time,” LSP explained.
What to own, what to outsource
Most startups don’t have the resources to build their own technology stacks, and LSP acknowledged that total independence is unrealistic even for established software companies. His advice is to identify which parts of a product are worth retaining control over, and which can be outsourced without creating an excessive dependence on a single supplier.
For Canadian startups, Zoho offers one way to reduce that dependence without having to build their own infrastructure. According to LSP, Zoho’s ownership of its AI models and underlying infrastructure makes customers less vulnerable to disruptions when outside providers change their pricing, policies, or access terms.
“The point isn’t to avoid external technology. It’s to make sure you still own the knowledge and the decision rights that let you adapt while the models keep changing.”
LSP, Zoho Canada
“Controlling where your data is stored is no longer enough,” LSP said. “Businesses also need to know who controls the models and infrastructure that process it. Zoho owns both layers and has never sold customer data.”
For startups developing their own products, LSP believes external AI models and infrastructure can still play an important role, provided their software is designed to accommodate another provider. Still, he encouraged founders to calculate the cost of switching while they still have the ability to do so.
LSP suggests starting by considering what would happen if their provider changed its terms tomorrow. Could they continue operating and improving their products? Could they move their data retrieval systems and AI performance tests to another provider, or would doing so require extensive engineering work?
He also encouraged startups to consider what information they’re sharing with their suppliers, including usage patterns and corrections that help improve AI systems. Retaining control over that knowledge, he argued, is essential to building a business that can adapt as the underlying technology changes.
“The point isn’t to avoid external technology,” LSP said. “It’s to make sure you still own the knowledge and the decision rights that let you adapt while the models keep changing.”
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 Immo Wegmann.
