Toronto Metropolitan University released a report titled Banking on AI, which looked at data from Anthropic’s Claude and Microsoft’s Copilot to analyze the value and risk of Canada’s financial sector in implementing generative AI to improve productivity while safeguarding privacy and compliance.
The report found that 98 per cent of Canada’s financial sector occupations are exposed to AI technologies, much higher than the national average where 56 per cent of the total workforce is exposed to the technology. The report examines which roles in the sector are likely to be augmented by AI versus those at risk of replacement. It finds senior management positions are more likely to be supported by the technology, while a much larger share of jobs, particularly in administration and customer service, face a higher likelihood of being replaced.
While generative AI in the financial sector poses opportunities for augmentation and overall productivity, the report emphasizes that the technology is not well suited for core financial tasks due to risks of inaccuracy. For example, conducting financial analyses, ensuring regulatory compliance, tax calculations, examining financial records, and more, where precision is required and small errors can have significant consequences. AI is not currently well suited to support in these tasks due to reliability concerns.
AI is found to be most effective in areas such as customer service, communications, and routine business support. Based on real-world usage data, AI is increasingly being used for answering customer questions, drafting materials, and internal support rather than for financial analysis.
The report recommends a cautious, targeted approach to adoption, prioritizing low-risk tasks while maintaining strong human oversight, data privacy protections, and regulatory compliance. Adoption should focus on areas where companies can safely improve productivity without introducing significant risk, such as front-line customer service interactions, drafting communications, and summarizing information, rather than core financial activities that require complete accuracy and numeracy.
As companies undergo AI adoption, strong data privacy protections must be upheld, ensuring compliance with regulatory standards and establishing clear accountability for errors or misleading outputs. It also calls for ongoing human oversight, including quality monitoring and safeguards to step in when AI systems fail. Financial institutions are encouraged to avoid overreliance on AI by continuing to develop employee skills and judgment.
Ultimately, generative AI is expected to deliver incremental productivity gains rather than transformative change, with success depending on responsible implementation and workforce adaptation. Read the full Banking on AI full report the Toronto Metropolitan University here.