IBM Makes Watson Smart on Financial Regulations

Over the past couple of years we’ve seen increasing interest in the use of artificial intelligence in the financial sector. From robo-advisors to banking bots, AI and intelligent assistance have been pouring into FinTech. This week IBM announced Watson Financial Services, which takes the use of AI in the financial realm to a whole new level.

In November of last year, IBM acquired Promontory Financial Group, a consulting firm specializing in financial risk management and regulatory compliance. It appears that IBM’s aim with the acquisition was primarily to leverage the vast domain expertise of Promontory’s team to train IBM Watson in specializes areas of finance.

In their press release, IBM outlined three core offerings that now comprise Watson Financial Services:

  • Watson Regulatory Compliance
  • IBM Financial Crimes
  • IBM Algo One Big Data Foundation

An Intelligent Assistant for Compliance Professionals
Out of the these three new tools, I took a deeper look at Watson Regulatory Compliance to better understand how IBM is applying cognitive computing tools to financial services. The regulatory compliance service uses natural language processing to help compliance professionals more easily access and search regulatory requirements. IBM is bringing together regulatory requirements from 200 different sources into one database.

According to an IBM video released last fall, compliance professionals can use Watson as a sophisticated intelligent assistant. Watson Regulatory Compliance aids human professionals in both understanding regulatory requirements and ensuring the company is properly positioned to apply them correctly.

Rather than reading through reams of compliance documentation, human professionals can query Watson on a specific regulatory obligation to see the cognitive computing platform’s summary of that requirement. The system applies natural language processing and machine learning to understand to identify relevant passages throughout documentation from 200 different sources. The human professional can accept or reject Watson’s summary, essentially helping to further train the system. If professionals need to customize the meaning of a requirement for their firm, they can enter their own interpretation into the Watson platform to further refine the definition.

Watson can also use intelligence to determine if passages in the documentation are providing guidance on how regulatory obligations should be followed. The Watson Regulatory Compliance system marks these passages accordingly, helping the human to identify them more quickly. The human can then label the passage as guidance and even link it to the obligation under review.

Presumably using machine learning, Watson also takes its best guess at tagging each requirement with relevant categories. For example, it may tag a requirement as being related to fiduciary duty, trade initiation, and data retention. Once again, the compliance professional can check Watson’s work and make changes as needed.

Watson also helps the professional determine how important the regulation is to different divisions in the corporation. In financial compliance lingo, that’s called defining scope and materiality. For example, one country may be doing a lot of trade initiation and therefore need to be on their toes concerning the requirements tagged with that category. Another country may not be doing trade initiation at all, so they don’t need to focus resources on that particular obligation.

In essence, Watson Regulatory Compliance seems to act as a very quick and efficient aid in helping compliance professionals wade through and make sense of huge numbers of financial regulations. Once the human professional has worked with Watson to review and store all data associated with a regulation, including customized definitions, tags, and materiality, the finished product is stored in the database for easy access. The system offers a prime example of human / virtual agent collaboration where the power of intelligent assistance greatly aids the human in executing a complex job.



Categories: Conversational Intelligence, Intelligent Assistants, Articles

Tags: , , , , , ,