Should midsized banks leverage AI?

Artificial intelligence (AI) is transforming the financial services industry at a rapid pace, changing the way financial institutions of all sizes operate and interact with their customers. While traditional machine learning (ML) and other advanced analytics methods have been used in banks for years to drive stronger insights and innovation and improve efficiency, generative AI is a newer technology that is changing the game. Generative AI is a type of deep learning algorithm that can generate content by learning and imitating the patterns and structure of the data it was trained on. This makes it capable of mimicking human behavior and thought patterns to process vast amounts of data and information instantly.

The opportunities for ML and AI in banking

With the commercialization and ease of access to AI tools, banks and financial institutions can now utilize AI in new and innovative ways. Building on the foundation of robotic process automation and traditional data analytics, AI can increase productivity and efficiency, allowing employees to focus on higher-value strategic activities.

For example, a bank could use AI to automate the process of analyzing customer data to identify potential fraud or to personalize the customer experience by providing tailored financial advice. AI can also be used to improve back office business operational efficiency, reducing human error and increasing speed and accuracy. The potential for AI to transform the banking industry is immense, providing customers with a more personalized and efficient banking experience.

Here are some common applications trending within the banking space:

• Risk assessment and fraud detection: ML algorithms can analyze vast amounts of data to identify fraud red flags and other potential risks from anomalous behavior. This helps financial institutions identify and mitigate risks, prevent fraud and enhance security measures. Generative AI could improve on this use case by taking the identified fraud red flags and creating a daily report of any outliers the model could not accurately confirm to be risk or fraud. A team could then review the report for actionable items.
• Compliance monitoring: Event-driven monitoring using AI-based rules could provide early warning of non-compliant activities around the following while enforcing model risk mitigation and responsible AI best practices:
  • Consumer protection, security and privacy
  • Unfair, Deceptive or Abusive Acts or Practices (UDAAPs)
  • Bank Secrecy Act/Anti Money Laundering (BSA/AML)
  • Office of Foreign Assets Control (OFAC) rules
• Program updates, testing and mitigation: In addition to monitoring, AI could automate much of the compliance workflow. You could use AI to:
  • Monitor regulatory agency sites and report changes to regulations, rules and guidance.
  • Perform gap analysis between compliance controls, policies and procedures, as well as published laws and regulations.
  • Generate report and alert narratives for automated compliance testing procedures.
  • Recommend action plans to mitigate non-compliance.
  • Process source documents and datasets used in subsequent compliance testing.
• Employee support: Use AI to provide employees with conversational chat and intelligent search to navigate a bank’s policies and procedures database and governance, risk and compliance applications.
• Customer acquisition: AI can help banks identify potential customers and then personalize marketing outreach, website content and other experiences for those prospects for cross-sell opportunities and increased wallet share.
• Customer service and chatbots: AI-powered chatbots and virtual assistants can provide personalized customer support, answer inquiries and assist with basic financial transactions. Generative AI makes these systems even more capable of understanding and responding to customer queries promptly with human-sounding language.
• Credit scoring and underwriting: AI algorithms can analyze credit history, income data and other relevant factors to make accurate credit assessments and underwriting decisions. This streamlines the loan approval process and improves the efficiency of credit risk analysis.
• AML compliance: AI systems can assist financial institutions in detecting and preventing money laundering activities. By analyzing transactional data, AI algorithms can identify suspicious behavior patterns and flag potentially illicit transactions.
• Regulatory compliance and reporting: AI tools can automate compliance processes, ensuring financial institutions adhere to complex regulatory requirements. AI can assist in monitoring transactions, generating reports and identifying any potential compliance issues.
• Market research and predictive analytics: AI can extract insights from vast amounts of financial data to identify market trends, predict consumer behavior and make data-driven business decisions. This helps financial organizations stay competitive and adapt to changing market conditions.

Managing AI risk

While AI offers a number of opportunities to increase efficiency and drive growth, it can also increase risk if deployed improperly or without appropriate governance. For example, if the training data used to develop AI models is inaccurate, compromised or contains errors, it can lead to flawed outcomes and compromise data integrity. In addition, AI can be subject to biases that can lead to fair lending regulation compliance risk, including:

• Data and sample selection bias, in which training data over- or underrepresents certain groups, leading to discrimination
• Algorithmic or proxy bias that emerges from the design of the algorithm or the criteria it considers (education level, zip code, etc.)
• Feedback loop bias, which can reinforce and amplify existing biases via the ongoing ML process

In addition to taking steps to avoid these biases, ensuring transparency is critically important in AI decision making in order to defend those decisions. Insufficient documentation around the inputs to the models may lead to fair lending and UDAAP issues, as it’s difficult to prove if the model is fair and ethical in its decision making.

To reduce AI-based risk, your organization must leverage enterprise risk management (ERM) to establish governance and appropriate controls. This should include:

• Proper training for employees working with AI tools as well as a general awareness campaign around the risks that come with AI
• License management for specific tools to avoid legal risk
• Policies, procedures and controls around data access
• A taxonomy for emerging risks around the fair, responsible and ethical use of AI
• Clear definitions of terms used in regulations (including the Biden executive order) such as “safe,” “responsible,” “fair” and “trustworthy” to describe the appropriate use of AI in banking and other contexts. These intangible concepts can be difficult to define and measure, but a carefully crafted risk management program will do exactly that to develop and inform AI models and use cases that are transparent, accountable, trustworthy and explainable.

AI represents a tremendous opportunity for middle market financial institutions. The right AI strategy can improve efficiency, help you reach new markets and customers, offer new products and better manage regulatory compliance. To take advantage of these opportunities while managing risk, you’ll need a comprehensive AI road map for your institution. Develop your institution’s approach now to ensure you’ll remain competitive in this quickly evolving market.