This month, the Ontario Securities Commission (OSC) in collaboration with Ernst & Young LLP published its first report (the Report) on the adoption and use of artificial intelligence (AI) in Ontario’s capital markets. The Report canvasses the ways AI is currently being utilized to support capital market processes including asset allocation, price and liquidity forecasting, hedging, trade order execution and surveillance, high-frequency trading, futures market analysis, and sales and marketing.

The Report notes that the current AI landscape in Ontario is already strong and internationally competitive, citing Vector Institute Reports which noted significant spending on AI research and development in Ontario as well as over $2.8 billion in venture capital investment in Ontario’s AI ecosystem from 2021-2022. The influx of spending on AI related applications, the uptick in new AI companies founded in Ontario and the corresponding increase in AI-related jobs all suggest that supporting the responsible development of AI systems can be a significant contributor to economic growth in Ontario.

Through interviews conducted with Ontario capital market participants, the Report identified three overarching purposes for which market participants are adopting AI systems – efficiency improvements, revenue generation, and risk mitigation. The Report found AI could contribute to improvements in efficiencies for firms through things like improved pre- and post-trade processes automation, execution quality improvement through better liquidity forecasting, and improved customer services through the use of customer support chatbots which can interface with customers and help address common questions.

With respect to revenue generation, the Report found that firms were leveraging AI’s ability to draw insights from previously untapped datasets, to provide deeper insights into client relationships and clients themselves, and for the development of sales strategies and bespoke marketing materials. The Report also found AI was being effectively used as a risk management tool for things like AML, trade surveillance, hedging, and onboarding processes.

The Report also comes with a word of caution noting the various challenges and risks associated with the burgeoning adoption of AI applications. “Explainability,” for example, has proven to be one of the main hindrances to the broader adoption of AI models in capital markets. Given the complex workings of AI models and significant difficulties explaining how certain computational systems function, AI models may not garner the trust of parties who are uncomfortable relying on a system that cannot be cohesively explained.

There are also significant data-related concerns including difficulties for organizations in terms of data aggregation, quality, and consistency. The reliability of AI models is predicated on the quality and reliability of the data being synthesized, all of which occurs under a complex data protection and privacy framework which requires specific consent for the use of customer data. The Report also identified various risks unique to AI models which require specific governance measures beyond those traditionally employed by firms. The Report suggests that an AI-specific governance framework can be an important tool to ensure that the culture of accountability and compliance extends to the responsible use of AI. This can help address a situation where an advisor is too reliant on AI tools and is unable to recognize if flawed data is providing poor or biased recommendations that are not in a client’s best interest.

The Report characterizes the adoption of AI in Ontario’s capital markets as being at an intermediate stage with varying levels of adoption, integration, and maturity across different functions. As capital market participants continue to embrace AI innovations, the Report reminds market participants that enthusiasm should be tempered with a sensible implementation approach which prioritizes the protection of investors and the integrity of Ontario’s capital markets.

November 30, 2023