Innovate to simplify, simplify before innovating: some ideas for effective, risk-based supervision

Ladies and gentlemen,

There is now broad consensus across Europe that financial sector regulation, supervision and reporting have become overly complex, costly and burdensome. This leads to heavy operational burdens for financial institutions, which must allocate significant resources to ensuring compliance, sometimes at the expense of innovation and agility. This complexity also weighs on supervisors, who must carry out extensive controls and sometimes devote disproportionate resources in terms of the results ultimately achieved.

The remedy for this situation is well known: simplification. It is one of the key recommendations of the Draghi and Letta reports, and we have already begun work on simplification within the SSM, notably with the revision of the SREP, which we want to make more flexible and effective. 

However, we must recognise that simplifying regulation, supervision or reporting is not always so simple! As Leonardo da Vinci so aptly put it, “Simplicity is the ultimate sophistication”.

On the one hand, supervisory rules and processes do not simply result from excessive bureaucratic zeal. These rules were often introduced for good reasons; it is rather their accumulation that poses a problem. On the other hand, any simplification process inevitably raises the question of how far to go. The consensus is that we must simplify without deregulating, i.e. without dismantling the entire preventive framework that protects us from a new financial crisis.

In this context, technology can provide part of the solution. It enables us to build powerful tools – known in our supervisory jargon as “SupTech” tools – to optimise the management of our supervisory processes. However, technology cannot do everything, and it is often more effective to consider simplifying the processes themselves before embarking on the construction of SupTech tools: beware of techno-solutionism!

Therefore, we need to reflect on the best way to combine technological innovation – our SupTech approaches – with simplification measures to make our supervision more effective. To contribute to this discussion, I would like to share with you today some lessons we have learned from our SupTech experience at the ACPR in recent years (I), before discussing our current roadmap and future plans (II).

I/ In terms of sharing experiences, I would like to point out three pitfalls to avoid and, conversely, a few best practices to favour.

1/ The first pitfall, as I mentioned a few moments ago, is believing that SupTech tools alone can solve all problems. Of course, new technologies – particularly artificial intelligence (AI) – can help us manage the wealth of information inherent to our work as financial supervisors. However, innovation must not be a crutch that allows us to continue moving in the wrong direction.

For example, technology is not designed to manage the ambiguity of concepts that often results from the complexity of regulations or supervisory processes. Quite the contrary: for new technologies to be deployed effectively, it is essential to have reliable, consistent and properly structured data.

Reporting issues illustrate this point very well: for a time, some argued that banking sector reporting would disappear, replaced by supervisors having direct access to granular data from institutions. In principle, this seems like an excellent idea. However, for it to work in practice, it would require extreme standardisation of the information systems of all European banks, which currently seems beyond reach. Without such standardisation, this would in practice amount to transferring to supervisors the burden of selecting the relevant information for each statistical concept to be measured, which would not only be excessively costly but would also have the major disadvantage of leading to a loss of accountability on the part of institutions.

2/ The second pitfall to avoid is entrusting the design of SupTech tools solely to IT specialists and developers, on the pretext that their technological expertise will necessarily enable them to identify the most effective solutions. In our experience, this type of approach, which is entirely supply-driven, inevitably results in solutions in search of a problem... and therefore end up gathering dust on the shelf. Furthermore, my experience with IT projects has led me to understand that the quality of tool design is essential for effective adoption. It is therefore crucial to design our SupTech tools in close collaboration with the business lines, not only to ensure their relevance, but also to guarantee their usability, knowing – and this is an additional difficulty – that user expectations are growing alongside the progress of “everyday technologies”.

3/ The third pitfall is believing that starting with users' ideas is enough to build SupTech tools that can simplify our activities. While this often simplifies the daily work of supervisors, it does not necessarily reduce the overall complexity of processes.

The various specialists in supervisory roles often have a very clear idea of what they consider to be time-consuming or repetitive tasks. When asked, they therefore tend to first imagine SupTech tools that could perform these types of tasks for them. This is obviously a very laudable goal, and we at the ACPR have carried out several such projects in the recent years. However, this generally leads to “small victories” – limited actions that are useful for a few people, but not necessarily significant for the organisation as a whole, especially when development and maintenance costs are taken into account.

Furthermore, this type of project tends to transfer the burden of complexity to those responsible for developing and maintaining SupTech tools. A tool developed to automate a complex process is likely to be complex itself. This entails practical maintenance difficulties: after a certain amount of time – sometimes not so long, given the turnover of teams! – there is a risk that no one will fully understand how a complex tool works in depth, and that we will collectively lose control of what we are doing.

Conversely, if we want the SupTech strategy to have a strong and lasting impact, it must be used to bring about a profound transformation of core business processes, those that truly structure the activity of a supervisory authority. To achieve this, we must first combine bottom-up and top-down approaches when selecting projects: for example, mapping out cumbersome and complex processes then, after analysis, simplifying them, and only then building SupTech tools. We must also take change management issues into account at this stage. While this is not so easy to achieve in practice, keeping the “big picture” in mind, i.e. the strategic SupTech vision at the highest level, is undoubtedly the best way to ensure that the objective of simplification and efficiency is at the heart of the approach and that technology provides support exactly where and when it is needed.

II/ Based on these convictions, sometimes hard-won, about the “best practices” of a SupTech approach, what are we actually doing at the ACPR and what do we plan to do in the near future? 

1/ First, let me say a few words about how we have designed and implemented our SupTech strategy since 2018 and the role of AI within it. AI can, of course, improve productivity through increased automation – as has been seen with search and data visualisation tools. However, the ACPR's ambition is also to empower its agents with new capabilities. The aim is not to replace our agents with machines, but to create teams of “augmented supervisors”  capable of doing more and better. In this context, the various AI technologies naturally support modern, risk-based supervision by detecting weak signals and atypical behaviour, facilitating market mapping and enabling the analysis of interconnections between financial players. The “LUCIA” tool, for instance, is used by the ACPR during on-site AML/CFT inspections and enables inspectors to examine a series of atypical transactions detected by mapping of large volumes of banking transactions.

2/ At the same time, the rapid evolution (and sometimes obsolescence) of new technologies encourages the swift achievement of “quick wins” that serve as milestones on the path to more strategic objectives. Since 2018, the ACPR has therefore run an experimentation programme to equip its teams with new digital tools. This approach has enabled the ACPR to gradually develop a series of practical tools that meet various operational needs, such as the automatic translation of technical documents, the transcription of telephone conversations, the analysis of advertising content, the pre-analysis of regulatory reports, etc.

3/ The rise of generative AI, particularly “large language models” (LLMs), marked a new stage in our strategy. Recognising their potential for supervisory activities, the ACPR organised a “Tech Sprint” in early 2024, bringing together external data scientists and supervision experts. The highly promising results demonstrated the true potential of these technologies for supervisory activities. Following the Tech Sprint, and based on the prototypes and ideas developed there, the ACPR launched an initial cycle of experiments on LLMs. For instance, the ACPR has developed a tool called “Veridic”, which can extract the characteristics of life insurance products from their ”key information documents” in order to classify them according to their level of complexity, a risk factor for policyholders. This is another concrete illustration of the potential of AI for risk-based supervision.

Fundamentally, the arrival of LLMs has led us to review all the SupTech tools that have already been developed. The aim is to update, supplement, or even redesign or replace these tools, using more cost-effective and efficient solutions.

4/ To take things further and, in particular, to ensure that the SupTech approach can be used to simplify our operating methods more decisively, we must clarify our vision of how AI can and will change things for us as supervisors. I am particularly thinking of “office automation” AI with generalist “co-pilot’ assistants, as well as “agentic” AI with specialised assistance tools – the next step in the evolution. 

This is what we are currently working on. The first step is to clearly identify the main challenges to be overcome. I would like to highlight three of these here. The first concerns data confidentiality and processing. While the use of the public cloud for our projects clearly accelerates their development and adoption, it raises very sensitive issues of “sovereignty” that should lead us to prioritise technical choices that do not favour the dominant non-European providers of current cloud services. The second challenge is training our staff to effectively use the new tools at their disposal, for example by teaching them how to write prompts. At the same time, we must also make them aware of the risks associated with these tools.

Ultimately, capitalising on the potential of new technologies will necessitate addressing even more fundamental questions regarding the evolution of supervisory professions: to what extent should analysis be entrusted to machines versus humans? How will LLMs transform our relationship with information and reporting? It is here that the overall vision I mentioned earlier will be crucial.

5/ Finally, our SupTech strategy naturally incorporates a European element, with the aim of contributing to the integration of banking supervision within the SSM and capitalising on all possible synergies. To this end, we have participated in the development of several tools for the SSM, most recently a natural language query tool integrated into the Athena application. We stand ready to develop new tools, provided they align with our objectives – namely, that innovation must serve simplification and vice versa. 

The ACPR is also proud to have organised a SupTech Week at the beginning of July, featuring presentations from the ECB, the Bundesbank, the Banca d'Italia and the Banco de España. This is another aspect of European integration that is close to my heart – and I will conclude on this point: the wealth and creativity that we Europeans demonstrate when we share our experiences and ideas. 

Today is Supervision Innovators Conference, like the previous editions I have attended, perfectly illustrates this! Let us harness this creativity and collective spirit of innovation to achieve the “supreme sophistication” of simplicity!

Thank you for your attention.