OSC publishes report on artificial intelligence in capital markets

The Ontario Securities Commission (OSC) and Ernst & Young LLP recently released a report (the Report) exploring the use of artificial intelligence (AI) in Ontario’s capital markets. AI generally refers to technology that can generate output such as decisions, recommendations or predictions for a given set of objectives. Key findings of the Report are summarized below.

Current uses

According to the Report, AI is used by capital market participants to enhance existing products and services rather than create new ones, whereby it serves three overarching purposes: efficiency improvement, revenue generation and risk management.

Efficiency Improvement 

The Report finds that the most widespread adoption of AI in Ontario’s capital markets is for improving the efficiency of operational processes. Notable examples include:

• Trade Process Automation: Businesses are increasingly using AI to automate pre- and post-trade operations. For instance, trade validation, which involves reviewing details of a proposed trade, can be simplified through using AI systems that are designed to quickly evaluate information relating to a trade and determine whether it meets the relevant risk and regulatory criteria. Such use of AI reduces manual effort and streamlines the trade lifecycle.
• Liquidity Forecasting: Liquidity forecasting refers to predicting the availability and movement of future cash flows. Traditional forecasting tends to rely on simple statistical techniques to analyze the market’s capacity to absorb and facilitate the buying and selling of assets. AI-based forecasting, on the other hand, has better predictive accuracy by virtue of its ability to learn from vast historical data. However, the Report warns that the performance of AI models may vary depending on the liquidity risk measure chosen. 
• Customer Service: Using AI to provide automated customer support functions has become more and more prevalent. This is largely attributed to the low-risk and cost-effective nature of tools such as chatbots. Despite the rapid integration of AI in this domain, however, the Report notes there are challenges associated both with training AI models with mass unstructured data and with properly storing such data.  

Revenue Generation

The Report finds that capital market participants are also adopting AI to generate higher revenue, including in the following areas:

• Sales / Marketing: There is extensive use of AI for sales and marketing purposes due to the ease of accessing internal datasets, lower risk of failure and fewer regulatory constraints. With its ability to extract insights from large volumes of transactional and consumer data, AI enables businesses to better understand and tailor their marketing strategies to the personalized needs of customers.
• Asset Allocation: Asset allocation is the process of optimizing resource distribution among different classes of assets, such as stocks, bonds and cash, with a view to maximizing investment returns. According to the Report, while there is ongoing interest among portfolio managers in applying AI’s capabilities to asset allocation, the level of adoption varies depending on the scale of the financial institution, with large funds demonstrating greater AI uptake than smaller funds.

• Trading Insights: A more common use of AI is collecting information from news sources for generating trading insights. The Report explains that recent advancements in AI, particularly natural language processing, can circumvent the limitations of traditional statistical methods employed for this purpose, which have been unable to effectively analyze text and video data. However, one subject matter expert cautioned that inadequate AI models could contribute to share price volatility by generating random trading signals.

Risk Management

The Report also finds that capital market participants are adopting AI to enhance a broad range of risk management strategies, including:

• Hedging: While still at an exploratory stage, AI has great potential for improving hedging techniques due to its ability to incorporate realistic market conditions, including transaction costs and liquidity restrictions, which factors can be overlooked in certain traditional hedging methods. Since hedging entails a perpetual need to respond to changing market conditions, AI’s capabilities in this respect may help advisors make more informed decisions.
• Futures Market Classifier: The application of AI for developing optimal futures market classifiers is also being explored. Futures market classification refers to analyzing and categorizing market conditions in futures markets, with the objective of minimizing slippage and enhancing trade cost performance. Based on the Report’s findings, AI-based classifiers are expected to outperform traditional approaches in accurately estimating market impact, particularly for illiquid securities.
• Surveillance of Market Manipulation: Market manipulation is a type of a fraudulent activity involving artificial inflation or deflation of the price or trading volume of financial assets. The Report notes that AI is extensively used in detecting such market abuse, with the most prominent trading surveillance systems (e.g., NASDAQ Trade Surveillance) all having implemented AI technology in their platforms. Compared to their predecessors, these AI-based frameworks are said to result in fewer false positives.

Challenges

Despite the many benefits of AI, the Report notes that major barriers remain for its adoption in capital markets. Consistent with the experience of AI users across all the economic sectors, market participants primarily face issues pertaining to the development or procurement of AI systems, including:

• Data Constraints: Given AI’s heavy reliance on data from diverse sources and formats, it is not easy to properly manage such data while guarding against potential privacy concerns. Other related constraints such as data availability and changes in the distribution or characteristics of data over time (i.e., data shifts) also present challenges in AI modelling.
• Investment Costs: Another significant hurdle lies in the hefty costs required for researching, developing and implementing AI solutions. Unsurprisingly the Report finds that small businesses lag behind larger players in AI adoption, likely due to difficulties in accessing technical expertise and concerns about returns on investment.
• Corporate Culture: As with any novel technology, it can be challenging to adapt entrenched processes and ways of working to the technological disruption created by AI, especially where such disruption necessitates changes in personnel and responsibilities. For this reason, market participants often grapple with issues relating to risk aversion and resistance to change from within their organizations. 

Conclusion

As the use of AI is widespread in certain areas (such as customer support) and more limited in others (such as hedging) it is generally described as being at an intermediate stage of adoption in Ontario’s capital markets. Significant and widespread advances are still to follow in the coming years.

However, as the popularity of AI tools in the capital markets grows, the risks and challenges associated with their deployment also grow and will invariably require careful monitoring and oversight by securities regulators and others.

In its concluding remarks, the OSC commits to continue studying AI’s current and future applications, value drivers and challenges, with the objective of understanding how to best support responsible innovation and practice.

The author would like to thank Maggie Shi, articling student, for her contribution to preparing this legal update.