Into the Future of RegTech | Automatic Summarization
“If you define the problem correctly, you almost have the solution,”
– Steve Jobs, a pioneer of technology commented on the use of technology as solutions
What are some of the challenges?
One of the most challenging hurdles financial institutions face today is the burden of regulatory compliance. Banks have been spending billions of dollars on compliance every year for the past decade, and the demand and pressure show no signs of slowing down. As a result of the constant regulatory change, compliance management remains a major expenditure. With new mandates and increasing complexity on a daily basis, banks continue to be challenged further and further (Finextra, 2017).
One of the most time-consuming tasks for compliance professionals is reading and interpreting lengthy regulatory documents such as proposed and final rules, register notices, enforcement actions, statutes, and regulations. Many of these documents can be hundreds of pages in length, are published in various formats and often use varying terminology, making them complex, time-consuming and expensive to analyze on a scalable basis. The process must break down each regulatory document into manageable pieces in order to understand the possible impacts it will have on the organization. Specific mandates might not apply to every area of a bank or to organizations under a certain asset under management threshold, and reading and subsequently summarizing the entire regulation can be a significant cost and resource drain on the organization.
Modern technologies now offer automated summarization capabilities, extracting the important aspects of regulatory documents for compliance professionals. Such solutions can provide the key attributes of a regulatory document such as: Key dates (publication, comment close, effective dates) Related rules and regulations, Changes history, the specific business or operational Topic classifications relevant to the business, and finally, summarize the content within the document into just a few paragraphs or bullet points. Such technologies can help determine what departments within the organization might be directly affected, and what action items need to be taken. The information can then be assigned to the correct individuals within the team for risk analysis, gap analysis and finally, implementation planning.
Automated content summaries can provide, with an increasing level of certainty, which key attributes, phrases or sentences within a regulatory document are most important to a specific business. Such auto-summarization can be accomplished through AI and supervised Machine Learning technologies with industry-specific algorithms, models, and training data. Often this can be decided by semantics and grammar structure. For example, Natural Language Processing (NLP), could be used to extract terms similar to “must” “have to” “should” “require to” and so forth, and give them varying levels of importance. Additionally, models can read entire documents, analyze and assign different weights to specific references within such documents. The models could be trained to look at sentence structures and emphasize sentences at the beginning of a paragraph and paragraphs at the end of the document, where you would normally find an introductory and conclusion statement, respectively. Other algorithms can be tweaked and tuned with training data, where the models could learn and become more intelligent based on human intervention, and eventually learn and improve automated decisions based on behavioral user patterns. Over time, such technologies and algorithms can become more confident in summarizing specific types of documents. In fact, a new category of summarization technologies provides summaries without the need for any additional supervision, training or user feedback.
Dejan Kusalovic at Intel believes that there are so many regulations with thousands of pages that if even a technology was built for a single regulation, it could be a helpful tool. (Finextra, 2017). The mass and complexity of an individual regulation, like Dodd-Frank or Bank Secrecy Act, requires whole departments within a bank dedicated to only the compliance demands of that regulation. The compliance demands involve reading regulatory documents in their entirety, then finding pertinent data from inside and outside the organization, researching outside content (such as comments or white papers), processing the data, making sense of it and then finally, recommending the actions that need to be taken. The ultimate goal is to automate as much of that process as possible, because of the ongoing maintenance requirements – due to frequent regulatory updates and changes. AI-driven solutions can play a crucial role in achieving this goal.
With the problem clearly defined, now a solution can come into focus. Michelle Curtin, EMEA Head of Policy, Regulatory Change & Governance BNY Mellon makes the point, “The regulatory tsunami that followed the financial crisis has significantly increased the cost of compliance for firms. Leveraging technology to help manage regulatory change and general compliance cost-effectively and efficiently has become increasingly important at a time of margin compression. With skilled compliance resources in short supply, utilizing technology to reduce certain administrative tasks or to evaluate management information, can allow compliance resource to narrow their focus to key risks.”
Are Automated Summaries the solution?
Artificial Intelligence (AI) in RegTech offers modern solutions to the various pain points faced in managing regulatory demands. Auto-summarization is a prime example of how a solution can be optimized to process vast amounts of data rapidly, then synthesize and connect disparate sources of information to provide enhanced insight and analysis that manual approaches are unable to accomplish on a scalable manner. Artificial intelligence based solutions could identify regulatory requirement patterns and create a list of actions that the organization could then use to establish a plan for compliance. As a natural next step, such a solution could identify business products or processes affected by such regulatory changes. (Crossman, 2016).
Automated approaches to regulatory change management are not an end all be all solutions. Regulatory experts and compliance professionals still need to carefully read the summaries and consider each extracted key attribute, and ultimately decide how each document relates to their business. Automated AI-driven solutions can streamline the process of getting information in the hands of the right people more quickly, consistently, with enhanced insight and thereby utilize bank resources much more efficiently than manual approaches. Technology can be used as a tool to help expedite information to the proper source and save valuable time and precious human resources. Many regulatory documents are coupled with strict, effective dates and any time saved during the regulatory change management process could help a bank met their deadlines with flying colors.
Crossman, P. (2016, October 11). ‘A Robot Could Alleviate This Drudery’: Bank Compliance Meets AI. Retrieved from American Banker: https://www.americanbanker.com/news/a-robot-could-alleviate-this-drudgery-bank-compliance-meets-ai
Finextra. (2017). FinTech Collaboration: How Banks Can Leverage the Power of RegTech. Intel.