Artificial Intelligence (AI) for regulatory impact analysis

  • Leah Robinson, Peter Dugas and Doug Helm
  • Published: 13 March 2024


US regulators have intensified their focus on the implications of artificial intelligence (AI) in the financial services industry, taking concrete steps to understand and address the challenges and opportunities presented by AI technologies. As AI evolves, regulators continue to support innovations that uphold fundamental principles of fairness, equality, and consumer protection.

Agencies’ commitment to safeguarding consumer interests when utilizing AI is demonstrated by the Securities and Exchange Commission’s proposed rule regarding conflicts of interest associated with predictive data analytics.1 This mirrors the Federal Trade Commission’s stated concern around AI and the engineering of consumer trust, in the context of t parties unfairly or deceptively using AI to influence consumer decisions.2 More broadly, regulators across the financial services industry are emphasizing that any application of AI should not place the interests of institutions over those of consumers and investors.

Collaborative efforts – such as the joint pledge by the US Consumer Financial Protection Bureau (CFPB), Equal Employment Opportunity Commission (EEOC), FTC, and Department of Justice (DOJ) – emphasize a unified approach to confront bias, discrimination, and unlawful behavior that could be facilitated by AI.3  By coordinating their actions and sharing insights, these agencies seek to promote responsible AI deployment and mitigate risks across various spheres of activity, from financial services to employment practices and consumer protection.

Moving forward, the financial services industry will continue to see fresh regulatory steps to ensure AI innovations uphold the fundamental principles of fairness, equality, and consumer protection – and those interventions can be expected to occur more frequently as AI advances. For institutions, the complexity and pace of these regulatory developments pose unique compliance challenges, especially when rules and guidance are emerging from regulators across a multiplicity of jurisdictions and impacting products and services on an enterprise-wide basis.

Capco's Approach to AI-Driven Regulatory Insights

Capco tracks these trends in AI and financial services topics via our Center of Regulatory Intelligence (CRI) Data Feed, and one thing remains clear: regulators agree that, while dangers exist, there could also be significant benefits to leveraging AI technologies in ways that promote compliance with consumer protection laws and fairness policies. 

For instance, our CRI team has introduced proprietary AI technology to assist our clients in navigating the evolving regulatory environment, ensuring essential updates are communicated promptly and effectively, thereby trying to safeguard both consumers and the institutions themselves.

Since 2016, the CRI Data Feed has been a cornerstone of our financial institution clients’ strategic planning, providing enterprise-wide risk management frameworks that examine the comprehensive regulatory landscape. Capco’s subject matter experts have reviewed approximately 193,000 regulatory and legislative releases annually since the Data Feed’s inception, designating relevancy to about 19,000 updates per year.

Capco has developed deep learning technology in the last year to better support a tailored regulatory data feed  allowing institutions to employ a centralized strategy that enables quick and direct dissemination of critical regulatory changes. Leveraging natural language processing (NLP), advanced machine learning (ML) algorithms, and the transformer model to automate text classification and pattern recognition, we can increase the level of confidence, accuracy, and consistency that are simply unattainable via manual processes. 


  • identifies relevant regulatory and legislative updates by assigning each update with a relevancy rating and sorting as ‘keep’ or ‘drop’
  • maps each record into its primary category and issuance type, fostering a more organized and accessible data structure.

Beginning with the initial acquisition of regulatory change updates and integrating existing application programming interface (API) and robotic process automation (RPA) process components, the model is dependent on a group of mandatory data attributes, including article title, summary, rule stage, important dates, and URL. These data points are transformed and organized into a database for further processing. To optimize the organization of the data within a structured database and to support decisioning and mapping, additional data elements are added to each regulatory update.

Next, each regulatory change update is categorized to organize data into logical groups. Our system utilizes a BERT (Bidirectional Encoder Representations from Transformers) base model to determine whether to keep or drop an item and uses NLP techniques and logistic regression as the ML algorithm to determine the primary category and issuance type for each record.

To kickstart the development of our NLP model, we drew on two years of regulatory change articles – or over 300,000 records of decisioned updates – for training purposes. Each month, the prior month’s articles were added to retrain the model. Our approach of organizing data into a structured database with two tables – one for the initial data, one for the refined data – facilitates the model’s learning from both raw and curated datasets, while also preparing it to adeptly handle new data in the future.

In summary, intelligent automation has a key role to play in harvesting, organizing, and defining tens of thousands of regulatory change updates every month. Through the combined deployment of RPA, API code, NLP, and ML models – in conjunction with periodic maintenance to address data changes and model enhancements – Capco identifies opportunities to reduce cost, provide scale to include more sources, and expedite new data to our clients and end users in a manner that aligns with current regulatory trends.