Machines can read… but only humans can execute
March 9, 2021

(Just) readable or (also) executable? The future of regulations

In our latest blog post we have introduced an essential issue for the RegTech sector - the MDMER (Model-Driven Machine Executable Regulations) - starting from the Response by EBA to EC Consultation (June 2020) and focusing on the technical aspects of this topic. In this post, instead, we will introduce some more philosophical issues related to the hypothesis of executable regulations.

Read previous post

Some numbers about financial regulations

As already mentioned, the FCA (Financial Conduct Authority) and BoE (Bank of England) RegTech Sprint (November 2017) has laid the groundwork for the creation of regulations written in a language which can be read and understood not only from humans, but also from machines. This hypothetical language would need just to be executed from machines, allowing financial institutions to automatically extract every regulatory obligation and more easily fulfill it. This would be a high-impact change in the financial world. In fact, according to a PWC analysis, only the regulatory detection and impact analysis operations represent the 15% of the total compliance costs. Besides, it must be taken into account also that, only in the 2015-18 period, the authorities have emanated 190 single-market-related new regulations to be adopted. According to PWC, the average time needed to nationally transpose a European regulation is 8,4 months. On balance, this activity is unsustainable. These figures are enough to explain why the adoption of the MDMER would be a real revolution, allowing to reduce time and costs for the transposition, interpretation and adoption of financial regulations.

The MDMER process

In brief, the MDMER would allow extracting semi-automatically regulatory obligations, exploiting semantic technologies based on open standards and changing the rules in the definition and transmission of the data from the authorities to any financial institution.

On the other hand, a difficult and delicate process would be required to implement the MDMER in practice. To sum up this process, we can consider five steps.

  1. The authority defines a taxonomy to identify univocally the terms and the application perimeter, also specifying an ontology for the relations between regulatory concepts.
  2. The authority emanates the regulation using a semantically pre-tagged text based on the defined taxonomy.
  3. Financial associations and services providers analyze this regulation, in order to give trackable feedbacks and insights exploiting knowledge-based tools.
  4. Financial institutions receive the pre-tagged regulation and analyze it according to their terms mapping, also considering their business rules to have a quick impact analysis.
  5. The mapping between the regulatory taxonomy and the business rules allows to clearly identify data and implement automations, in order to extract and build reports for the authority.

It seems very clear that all these operations are entirely hypothetical for the moment and currently long way to be possibile on a practical level. And surely we are not the first to discover that in theory many things work.

Readable VS Executable: where humans beat machines

Anyway, at Aptus.AI we believe that execute - in a specific sense - is an activity that can only be achieved by humans. Don’t get us wrong: machines are surely meant to execute, but when we talk about interpreting a regulation and making decisions based on that interpretation, we think that only a human being can do it properly. And this is not a technical issue, but - how can we say? - a philosophical one, as we believe that reality and - more specifically - any regulation are too complex and interpretable to be represented in a taxonomy and executed.

This is why we have created Daitomic, our AI solution addressed to the RegTech market and specifically thought to provide humans with a tool that allows them to make confident and quick decisions in the financial sector. Daitomic’s AI engine reads PDF - or any other document format - as if it was a human being and makes them readable by machines. But this is not all. This Saas solution creates an electronic version of any regulation, making it structured and visualizable. It also extracts and maps different contents, analyzing them and distinguishing one from another according to their type. This is what we mean when we talk about machine-readable. Which can also be expressed with accessible. Because we believe that only humans can execute - that, in this case, is intended as giving the correct interpretation of the regulation depending on a specific context.

To go deeper in this topic, you will need to learn more about Daitomic: click below and get in touch with us!