2 Across: Changing Financial Product Recommendations

Artificial intelligence (AI) can and is already changing how firms and registered representatives recommend financial products to customers. AI has the potential to enhance recommendations that registered representatives give, by providing them with comprehensive real-time information about their customers. But AI recommendations can contain errors that firms and registered representatives need to recognize. 

12 Down: Positive Impacts of AI
Firms are already filling in the spaces for enhancements AI can provide for financial product recommendations. AI is being developed to use clues about customers from traditional sources like their assets, spending patterns, and debt balances and nontraditional sources like social media posts, browsing histories, and prior communications made through email, chat messages, and meeting notes. AI incorporates this broad range of data into its output, which can provide registered representatives with a comprehensive, real-time, and tailored view of their customers. As such, registered representatives can use this information to enhance their existing skills and knowledge to provide customers with bespoke recommendations for relevant financial products. But firms have been cautious about allowing AI to provide
customers with direct recommendations.

6 Across: Dangers of AI
Firms should be aware of potential errors in AI recommendations. AI can make flawed recommendations based on “hallucinated,” i.e., made-up, data. AI also has the potential to provide recommendations not in the best interest of customers, for example, by prioritizing lower-quality products with higher costs where lower-cost, higher-quality products are available. Further, since AI uses historical data, it is susceptible to perpetuating or amplifying existing demographic-related biases reflected in the dataset.

These problems can be amplified by the complex nature of AI models, which often makes it hard to understand or explain how the recommendations are created. In some cases, AI models can generate different recommendations based on the same input without an explanation. So, making sure recommendations are reliable and free of bias can be difficult. But in-line citations providing the source of the data for recommendations could counter this problem.

3 Down: Interested in AI Recommendations

Most importantly, firms need to understand that using AI does not relieve them of their obligations to their customers or the SEC. The SEC maintains that AI does not relieve firms, or registered representatives, from complying with all applicable securities laws, rules, and regulations. Accordingly, firms and registered representatives are responsible for any recommendations made with AI, regardless of their knowledge of how AI finished the puzzle.