AI-Based Narrative Generation for Risk & Compliance Reporting

If we were to take a high-level view of the reporting function within a bank, it generally spans across various lines of business and support functions. Narrowing this down to the risk and compliance function, the reporting requirements in the last decade have grown significantly; at the same time, they have also increased in complexity and shrunk in periodicity to daily or to near real-time reporting for certain scenarios.

Based on industry surveys for most of the reporting life cycles, whether for regulatory or internal reporting, it has been observed that the report generation and preparation activity takes up to approximately 90 percent of the time involved, leaving aside a mere 10 percent for the review and analysis activities – hardly sufficient, as the gamut of reporting within the risk and compliance area has been on a steady rise in the last decade because of stringent regulatory requirements and due to the fact that many banks have increased efforts to gain more predictive insights from the internal reports generated.

Typically, the risk and compliance reporting function in banks and global financial institutions works in a semi-automated manner. Though there have been investments in leading third-party products and in-house development, it is very rare that one would find a bank or a financial institution achieve a solution that takes care of its end-to-end reporting needs. Preparation of high-quality reports for meeting internal and regulatory requirements has been a longstanding challenge for global banks. Typical challenges banks encounter include:

• Poor quality of data required to meet reporting expectations.
• Data integrity issues, most often seen in larger banks, which are more complex in nature, imposing a dauting task for organizations to meet standards for accuracy and completeness.
• Lack of a strong governance structure that ensures the required standards for reporting are maintained related to people, processes and data.
• Significant manual time and effort involved in analyzing numbers and interpreting them.
• Existence of firmwide data dictionary which maintains standard definitions.
• Limited time spent on report analysis and Insight generation.

While most banks have already started investing in data governance and data infrastructure for ever-expanding business needs, one area where the banks are trying to increase their investments and efforts is in the area of report analysis and predictive insights.

While many banks are hesitant to completely overhaul their existing reporting platforms, they are mostly looking at options where complementing or missing features can be embedded in their existing solutions. Banks are exploring newer concepts, like natural language generation (NLG) in combination with artificial intelligence (AI) to analyze unstructured data and uncover hidden patterns and anomalies, and they are also using the technology to write user stories or narration on the reports that have been generated. The main benefits of AI-based narratives include:

• More insightful reporting, as the solution can look at multiple dimensions and provide interpretation.
• Complementing narratives and insights for the various dashboards produced.
• Reduction in manual efforts, enabling the organization to redeploy and leverage their human capital more effectively for other value-added activities.
• Making business data more relevant by highlighting the most critical insights and improving the decision-making process.

In risk and compliance, too, there are a wide number of use cases or scenarios where the same can be applied, including:

• SAR Narrative Generation and Analytics – AML Systems churn out a huge number of alerts that need to be checked for suspicious activity. If a suspicious activity is confirmed, then a detailed report is expected to be filed with FinCEN. While most of the aspects of SAR generation, like data collection and report generation can be automated, the narrative part included in the SAR reports can be automated through the use of NLG in a language that is clearly understood by the regulators.
• BSA (Bank Secrecy Act) Annual AML Risk Assessment – An annual exercise that looks at creating reports on key dimensions such as customers, accounts, transaction and geography can also benefit through automated narrative generation. Post assessment, reports are generated from the bank’s system; following, the bank’s compliance office needs to analyze and provide a narrative commentary on the various findings, along with predictive insights. The bank’s activity in this regard is currently mostly manual. An AI-based solution leveraging NLG can help in understanding the patterns and anomalies, as well as in creating a narrative report.
• Risk and Compliance Internal Reports – The risk and compliance function apart from regulatory reports keeps generating many reports for internal consumption. The reports also need to carry a narration on the changes observed on a periodic basis and also provide a forward-looking insight based on the numbers.
• Stress-Testing Reports That Require Narratives – Global regulators have specific stress-testing requirements for their geographies. Initially, the exercise started with reporting quantitative numbers, but it has gradually incorporated qualitative aspects as well. Most of the current numbers reported and forecasts are required to be supported with commentary from the management side.

Most banks currently are at the cusp of moving away from traditional practices and are embracing newer technologies and instilling modern practices in risk and compliance.

The reporting function is the last and the most critical mile in the entire process, and automation can help to a great extent in easing the burden on most of the processes involved in the reporting life cycle in terms of data gathering, report generation and narrative generation. However, the banks using AI capabilities will also have to ensure that they maintain the required transparency, implement standardized processes and also document and explain the AI -based outputs to the internal and external stakeholders.