Compliance has always relied primarily on individuals. The regulatory burden for financial institutions over the past decade has been steadily increasing, leading to a sharp increase in the demand for compliance professionals.

To address these challenges, companies have had no choice but to recruit more and more compliance personnel to cope with increasing regulatory constraints.


The need for new technologies

However, this did not reduce the number of fines issued by supervisors. New records are being broken at each new regulatory breach identified in a Financial Institution (FI). As human limitations are being reached, the aid of technology is therefore needed.

Financial institutions and regulators have understood that by harnessing the power of technology, particularly artificial intelligence (AI) and machine learning, a considerable part of the compliance function can in fact be automated, thus reducing the regulatory constraints of compliance institutions and professionals.


AI Issues for the Financial Sector

AI’s stakes for the financial sector are vital. Above all, we must not be late. To respond to them, the CPRA launched a public consultation in 2018 (see responses to consultation) with various players (banks, insurance, fintechs, foreign central banks, etc.).

This consultation looked, in other ways, at issues for supervisors such as the governance and “explicability” of algorithms or the advantages of artificial intelligence techniques for the exercise of their own missions through the Supervisory Technology.


What communication between supervisors and financial institutions?

In this context, we can consider that the relationship between supervisors and financial institutions (FIs) is essential for the stability of the financial system. The exploitation of AI in the “Suptech” is therefore a major element of this relationship.

It can help FIs to comply with their regulatory requirements by reducing reporting rules that depend on human interpretation and that may represent certain sources of error.


The premises of an evolution: the “Digital Regulatory Reporting”

In practice, such a project has already been carried out in the United Kingdom and has been led by the Financial Conduct Authority (FCA), in collaboration with the Bank Of England and a number of financial institutions. The project is entitled “Digital Regulatory Reporting” (DRR).

This “pilot” program designed to assess the benefits of machine-readable reporting and explore how technology (recurring neuron networks and semantic web) can help financial institutions in their relationships with the regulator.


A complex context

In the United Kingdom, the manual management of reporting is a very heavy and complex burden for the supervisor, with more than 20,000 rules and 58,000 companies. This figure is constantly growing. Financial institutions must also interpret the rules and data required by their own means.

Thus, when the supervisor asks them for data, differences in interpretation often lead to confusion as to ineffectiveness, which has quite costly consequences in terms of time and money.


What are the aims?

The overall objective of this project is to reduce the time and costs required to interpret and implement the new reporting requirements. The other targeted objective is to reduce the number of individual regulatory reports that businesses have to produce.

To do this, the FCA has studied how a machine-readable regulatory framework can interact with a standardized language mapped on source data and uses semantic web technologies to identify the most appropriate approach to data specification.


An early but functional automation

The FCA has already proven that the concept is working well. The regulator has announced that it has successfully applied automatic reading technology to two different regulations. One is based on capital requirements and the other is based on mortgage criteria.

For the future, the CFA plans to expand the scope of the project in 2019 and apply the technology to more regulations. This is an interesting development for the sector.


What to remember?

Communication between supervisors and financial institutions is a complex subject. On the one hand, regulations are numerous, complex and sometimes subject to interpretation. On the other hand, taxable persons are obliged to carry out numerous and detailed reporting obligations.

It is noted that manual management is reaching its limits. The need to use RegTech/LegalTechs that use AI-based technologies and machine learning is therefore becoming vital.

The experience of the FCA, although limited, proves that it can be achieved. Of course, human intervention is essential for validation. AI avoids all repetitive tasks that have no added value. It therefore allows the human to devote himself to tasks with higher added value.