Ottawa firm helps companies navigate industry regulations with AI

With smaller businesses increasingly relying on AI to streamline certain tasks, one Ottawa-based expert warns that the technology is currently not good at parsing Canada’s complicated regulatory standards, putting some SMEs at risk.

With smaller businesses increasingly relying on AI to streamline certain tasks, one Ottawa-based expert warns that the technology is currently not good at parsing Canada’s complicated regulatory standards, putting some SMEs at risk.  Sonia Parmar, a former federal regulator and vice-president of regulatory affairs and government relations at the Canadian Health Food Association (CHFA) — which works with 1,400 members in the highly regulated health food industry — told OBJ that only 10 per cent of businesses in the sector are considered large enterprises. The rest are small and medium-sized enterprises (SMEs). For those businesses, keeping up with regulatory changes is an important but complex and often expensive aspect of operations. 

“The biggest struggle for most companies is the interpretation of regulations,” Parmar said. “Oftentimes, on the ground, how a regulation is interpreted will cause the most headaches. And these change constantly, so it’s very difficult for companies to keep up with that kind of flux.” Parmar said while AI has become a valuable resource for SMEs operating with lean budgets and teams, when it comes to regulations, the technology is spitting out errors that could have consequences. These include hallucinations, when a generative AI model generates false, misleading or fabricated information that sounds plausible. “Typically when a business uses AI right now, they’re using it to grow. It helps you understand and optimize because you wear 100 hats,” she said. “If you’re a one- or two-person business and you’re everything from accounting to regulatory, AI has helped a lot. But you’re also worried about hallucinations. We’ve found that hallucinations often happen in the regulatory space. AI is not accurate.” Thomas Conway, founder and CEO of Ottawa AI company Paradigm Forge, said most AI models are “horizontal,” meaning they’re designed to handle a broad array of tasks. It’s for that reason most chatbots will make errors when helping businesses understand complex regulations, he said.  In response, his company has built a “vertical” AI system designed specifically to understand Canadian regulatory standards across a variety of highly regulated industries.  “It’s a completely different AI architecture that cannot hallucinate,” he said. “Any one product is not regulated only in one area, it can be regulated in several. The advantage of our platform is that it will respond to every query by looking across the entire regulatory landscape, and it will do so with accuracy. It will not give you a response that it cannot tie to a citation, to an actual regulation, to a paragraph within a regulation.” Conway said the company has created systems for several compliance standards, including in the areas of environment, food safety, natural health products, medical devices, chemicals and trade.

For SMEs in these spaces, Conway said cost is a major barrier when it comes to compliance.  “The big guys can afford to hire a team from a Big Four consulting firm,” he said. “But the small and medium-sized enterprises in Canada often struggle with the regulatory challenges. The system helps these companies understand their regulatory requirements, their permits, their licences.” While his company’s AI system can’t hallucinate, Conway said they aren’t perfect. Like any chatbot, he said, it occasionally makes errors, though the information always goes through a series of checks before getting to the user. The model always includes citations so users can verify information for themselves.  Conway added that he’s enjoying challenging the platform to improve its outcomes.  “We’re rolling this out quite cautiously,” he said. “At a meeting with another organization yesterday, I said, 'Give us your gnarliest, toughest, grizzliest questions you can think of.' What that does is tell me if I have any edge cases here. Is there anything our platform can’t answer with a great deal of certainty? If I find something, I go, 'Yes, perfect.' Because then I can fill that gap.” The platform, which Conway said will be offered at an affordable price point, is currently available upon request from clients. In the coming weeks, Conway said a decision will be made on when it’s ready to be released publicly.  While he believes the platform will be an important resource for cash-strapped businesses, he said anything involving AI will always require a “human in the loop.” “We don’t replace regulatory experts,” he said. “We just help (businesses) with that enormous breadth of regulation out there so they can consider what’s right for their clients or their organization.” Parmar added that, from her perspective as a former regulator, the tools could be useful in the future to identify inefficiencies in the regulatory system, which she said is made up of multiple layers at the federal and provincial levels that don’t talk to each other.  “Our set of regulations is outdated and it can add extra costs unnecessarily,” she said. “I hope, one day, that regulators can adapt the program to streamline the system. They don’t always go back and clean their old stock and that can accumulate, even with red tape reductions. It can get very expensive.” While some businesses need the support more than others, Conway said regulatory compliance is a challenge all organizations must contend with.  “There’s no such thing as an unregulated business,” he said. “It doesn’t matter what sector you’re in; you are subject to regulation.”