We have identified three areas in finance where companies are applying AI to solve real business problems: fintech, regtech, and suptech.

Our readers might be familiar with the term “fintech,” which today loosely includes technological innovations in the finance sector, such as the use of AI in retail banking, investment, and crypto-currencies.

The Institute of International Finance (IIF) defines regtech as the use of technology to automate and optimize regulatory and compliance requirements for businesses. In our previous report, we covered the possibility space for AI applications in regtech.  

Supervisory technology, or suptech, is another area of finance that has emerged on the heels of regtech. Suptech involves applying emerging technologies, such as AI, to help supervisory agencies in finance automate and optimize supervisory tasks.

Central banks and other supervisory agencies seem to be seeking technological innovations such as data science machine learning. Effective data collection and sharing mechanisms can potentially be developed using AI to help supervisory agencies improve efficiency in their internal processes.  

We suspect the biggest driver for the use of AI in suptech is the increasing interconnectedness in global financial markets. The global financial network today also means that risks get propagated faster requiring supervisors to keep pace with the changes.

In the financial sector, it might now be possible for supervisory agencies such as central banks, capital markets, and banking authorities to utilize their internal data records and external data, such as news or social media trends entities, to better monitor financial markets using AI.

We researched the space to better understand where AI comes into play in suptech and to answer the following questions:

  • What types of AI suptech applications are currently in use in finance?
  • What tangible results has AI suptech driven in finance?
  • Are there any common trends among these innovation efforts? How could these trends affect the future of finance?

This report covers use cases and pilot projects from reputable banks and universities to help business leaders understand the possibility space of AI in suptech. Our research led us to believe the following data analysis applications in suptech are likely to emerge in the next two to five years.

  • Market Surveillance
  • Risk Prediction
  • Reducing The Cost of Fraud Detection and Anti Money Laundering (AML)

This article intends to provide business leaders in the finance space with an idea of what they can currently expect from Ai in their industry. We hope that this article allows business leaders in finance to garner insights they can confidently relay to their executive teams so they can make informed decisions when thinking about AI adoption. At the very least, this article intends to act as a method of reducing the time business leaders in finance spend researching AI companies with whom they may (or may not) be interested in working.

Market Surveillance

Bank of Italy

The Bank of Italy claims they are currently engaged in an internal pilot project aimed at deriving textual sentiment from Twitter posts and generating a score for depositors’ trust in banks at any point of time. The bank then uses this score indicator to measure the accuracy of prediction for their retail funding model. The pilot seems to be aimed at identifying any threats to financial stability due to varying levels of public distrust in the banking systems. The bank used natural language processing (NLP) techniques on Italian tweets for four major Italian banks (BMPS, UCG, ISP, UBI) and Deutsche Bank to train their model to identify sentiments.

Bank of Italy is the central banking authority in the country and claims they have developed an NLP system that can be input with tweets or web and newspaper articles through an API to identify positive or negative sentiments. The system then analyzes thousands of tweets and assigns positive or negative tags to each word in a tweet. The system then provides a report of the numbers of positive and negative tweets and the main topics of discussion taking into account all the tweets.

We could find no demonstrative video for this project.

The Bank of Italy recently announced the creation of a multidisciplinary AI team including economists, statisticians, and computer scientists from various departments in the company.

According to the post, the Bank of Italy created the AI team to enable pilot projects to help monitor non-compliance among the several entities being regulated by them. The bank claims their AI team was tasked with building both hardware and software to support their economic objectives. The team was reportedly able to analyze both structured data from the bank’s data warehouse and unstructured text data from social media such as Twitter, as a proxy for measuring inflation or understanding the depositor trust levels toward banks.

The bank’s AI software reportedly uses NLP and machine learning to extract, categorize and identify positive or negative macro customer sentiments for specific companies or their effect on stock returns, volatility, and trading volumes. In another application, the bank claims their team was able to use the software to measure economic policy uncertainty (economic risks for the country) by using Twitter posts mentioning the bank (in Italian) and other Italian news media as inputs.

The project from the Bank of Italy still appears to be in progress, and so there are no available case studies or marquee clients for the system.

Luigi Bellomarini is Deputy Head of “Applied IT Research Unit” at Bank of Italy. He holds a PhD in Computer Science from University of Studies Roma Tre. Previously, Bellomarini served as a scientific collaborator at the University of Oxford, although it is unclear if he worked on AI projects during that time.

Risk Analytics

Financial Network Analytics (FNA)

FNA is a London-based company with 17 employees. The company offers the FNA Platform, which they claim can help financial services businesses with risk and financial market analytics using machine learning.

FNA  claims users can gain real-time insights from their graph analytics engine through a dashboard. Financial services companies can potentially use the FNA platform through a Software as a Service (SaaS) agreement or as an on-premises enterprise software. The software gathers information from internal records of supervisory organizations or public databases.

Then, FNA Platform uses machine learning to analyze the data and identify potential risks to financial markets. FNA’s website states their platform has hundreds of data analysis algorithm options on their dashboard that users can choose from to identify important nodes or detect commonalities in the input data. These commonalities could range from data on equity holdings for the American financial market or investment portfolios. The system then provides a visual representation of the data to clearly show outlying data points on the software’s dashboard as shown in the figure below:

Below is a 7-minute video demonstrating how FNA Platform works:

There seems to be some evidence of FNA having worked on a project with UBS Group, a Swiss multinational investment bank and financial services company. That said, FNA does not provide any statistics reporting the details of implementation or results their software might have generated for the client.

We could not find any noteworthy companies for which the FNA Platform may have helped.

We were unable to find any C-level executives with AI experience on the company’s team, although FNA seems to employ a small number (6-7) of data scientists who are presently PhD candidates from colleges such as Columbia University or Imperial College, London.

Reducing The Cost of Fraud Detection and Anti-Money Laundering (AML)

Digital Reasoning – The Bank of England Project

Digital Reasoning is a  Franklin, Tennessee-based company with 187 employees. The company worked with the Bank of England in an AI suptech project aimed at helping the bank’s supervisory team automate information extraction from regulatory documents using NLP.

The Prudential Regulation Authority (PRA) is a financial regulatory authority in the United Kingdom. The PRA is a part of the Bank of England, and it supervises the prudential regulations for banks, credit unions, deposit-takers, insurers, and major investment firms. According to the Bank of England, the PRA supervises over 500 insurance firms, including general insurers, life insurers, and the London insurance market.

The supervisors at the PRA usually need to comb through a large amount of unstructured or semi-structured text data generated and published by the regulated companies. This was a time consuming and human resource-intensive task.

The bank was looking to develop an AI tool that enables supervisors to search through the documents of these regulated entities for information relevant to supervisory regulations. The bank also needed a way to automatically capture and flag any new information of regulatory interest so that supervisors can review these developments much faster.

The Bank claims their data science team worked alongside a team from Digital Reasoning for a Proof of Concept project which involved using the AI vendor’s cognitive system to develop an NLP algorithm capable of understanding which parts of the text in a document are of regulatory interest.  

The bank’s software development team then worked with developers from Digital Reasoning to design a dashboard.

According to the bank, for the training process, Digital Reasoning’s system ingested data in the form of samples from public documents, which included snippets of information relevant to the Bank’s insurance supervisory teams.

The Bank of England claims the tool can automatically categorize sections of text in the data and extract information specifically of regulatory interest. The tool was trained to choose the correct categories by experienced supervisors who helped with labeling and annotating excerpts of text selected by the software.

We must state, however, that the project is still only a proof of concept, and we could find no evidence of the tool being used commercially or having provided any real business results yet.

Bank of England exclusively uses the software internally, and so there are no available case studies for the software.

We were unable to clearly identify which of the leadership team members from the Bank of England was involved with the project. However, Digital Reasoning’s Senior Director of Artificial Intelligence at Digital Reasoning Brandon Carl was previously the Vice President at Bank of America and founder of Aventura Labs, a marketing automation and analytics company.  

Takeaways for Business Leaders in Finance

The same technologies that offer efficiencies and opportunities for fintech firms and banks may also have the potential to improve supervisory efficiency and effectiveness.

The benefits of AI in suptech may include (near) real-time insights derived from regulatory data access and automation of supervisory processes. However, barriers to implementation may include standardized internal IT procurement policies and cross-border data movement regulations.

A small number of supervisory agencies are currently exploring the feasibility of using AI to enhance existing supervisory functions. As with other industries and sectors, AI in suptech might help in expanding human supervisors’ capabilities by providing insights from large amounts of unstructured data. This application for AI could be used to assess financial risk, monitor or review documents submitted by regulated entities or improve a central bank’s regulatory guidance policy.

We believe that that AI in suptech is still at a highly nascent stage but is rapidly seeing the establishment of use-cases. The emergence of the terms FinTech and RegTech seem to have spawned not only the name suptech but also a lot of the same interest in using transferable lessons from Fintech and RegTech AI applications.

Five years from now it seems as though the central banking authorities in several countries might likely adopt some form of AI automation to gather insights from regulatory documents, social and news media.

In the near term, central banks and other supervisory agencies should not expect to easily automate their business processes or gain business intelligence from their data without embarking on a lengthy integration process starting with managing and organizing their data.

This might additionally require discussions with vendor support representatives and a large upfront costs. The largest enterprises may have the budget and staff to pursue the technology, but based on our research, it is as of right now accessible to companies that would be able to afford AI applications.


Header Image Credit: lfccworkforce