Why a Northern BC credit union took AI sovereignty into its own hands

Northern British Columbia isn’t usually the backdrop for national conversations about digital transformation.

But in Prince George, one credit union is grappling with a challenge that’s becoming increasingly relevant across Canada: how to adopt AI without giving up control of sensitive data.

“Why can’t companies take a measured approach of training on open data, and committing to not selling your information?”

Chandrashekar Lalapet Srinivas Prasanna, Zoho Canada

Integris Credit Union serves about 28,000 members across a region where in-person banking can be either inconvenient or impossible. To keep up with demand for mobile-first services, the organization began exploring AI tools to streamline operations several years ago. Before any implementation, though, they needed clear guardrails.

“We can’t share data with anybody,” said Integris VP of Information Technology Jeff Anderson. “Our members trust us with their financial futures. We had to make sure that we do so in a secure way, that we’re protecting their assets for the long term.”

At a time when many businesses are rushing to plug AI into their workflows, more are asking who really owns the technology and what it means to rely on models built and hosted by someone else.

“The challenge that they have had is there’s no awareness as to how data flows for these AI outcomes,” said Chandrashekar Lalapet Srinivas Prasanna (LSP), Managing Director of Zoho Canada. “Where is it going? Who is it training? Where’s this information being retained?”

That lack of transparency has already led to public backlash. In recent months, several tech giants, including Meta and Adobe, have faced criticism for shifting their terms of service in ways that allow user content to be fed back into their AI training systems, sometimes without clear or upfront consent.

In sectors like finance, education, and healthcare, where personal information is tightly regulated, even unintentional data exposure can come with legal consequences. With new Canadian privacy legislation on the horizon, experts expect scrutiny around AI usage to intensify.

Terms like “AI sovereignty” or “data residency” can feel abstract until a company faces a breach, a compliance audit, or a loss of customer trust. In practice, sovereignty means knowing how data flows, where it lives, and who has access to it.

This is a recurring theme in LSP’s conversations with IT leaders. One peer at a major Canadian bank that LSP spoke with described one of the best-known AI-powered assistants as a “black box,” meaning it’s nearly impossible to understand what’s happening inside.“No enterprise or large organization wants to deal with a black box,” LSP said.

The legal implications are just as serious. “If the data is leaving your country, then your business is suddenly subject to another jurisdiction’s laws,” LSP added. 

It’s why Zoho, which offers a suite of software and web-based tools for businesses, has built every layer of its AI stack, from infrastructure to models, and keeps all Canadian customer data in Canadian data centres. It’s an approach that’s more expensive and more complex, LSP said, but it means that customers aren’t at the mercy of shifting terms of service.

Adobe’s recent backlash, LSP noted, is a sign of what’s at stake. When the company quietly claimed the right to use user-generated content to train its AI tools, the public response was swift. 

“The fundamental question is, why should a big brother be doing this at all?” LSP said. “Why can’t companies take a measured approach of training on open data, and committing to not selling your information?”

For Anderson, that risk calculus was never hypothetical. Integris had already seen how easy it was for employee experimentation with consumer-grade tools to introduce vulnerabilities, even unintentionally.

So when it came time to actually integrate AI into core workflows, Integris needed a system that could balance utility with control. According to Anderson, Zoho’s tools offered a way to do that.

Through Zoho’s pre-built AI integrations, Integris staff can access generative AI features like writing assistance and data analysis, while ensuring member information never leaves the country, or gets used to train someone else’s model. Sensitive data is tokenized before any prompts are sent. Third-party models can be used in the background, but only through a tightly controlled enterprise layer.

“This has certainly sped things up in a really positive way,” Anderson said.

That agility has already paid off. When new financial regulations were announced requiring financial institutions to notify customers more quickly about non-sufficient fund fees, Anderson’s team was able to roll out a compliant workflow in days, something that he said would’ve previously taken weeks of development.

Integris’ goal, Anderson said, isn’t to replace human staff with AI. It’s to help existing staff work more effectively, and the systems they rely on are more responsive.

“There’s cautious optimism around AI,” Anderson said, but noted that the trust Integris has built with its members doesn’t come from using new tools, but from showing they know how to use them responsibly.

PRESENTED BY

Zoho is powering secure, sovereign AI for Canadian startups and organizations. Learn more.

Feature image courtesy Unsplash. Photo by Harsh Singh.