In a recent panel discussion as part of the FCA’s Data Week (aimed at building the FCA’s data capabilities to become a data-led regulator), the regulator invited guest speakers to discuss the increasing use of data and AI in financial services and what the FCA can be doing to assist firms in light of such developments.

Three main topics were discussed during the session, including:

  • the regulatory opportunities that will be enabled by data in the coming years;
  • how organisations can avoid negative outcomes from bias in their data work; and
  • what data changes firms would like to see from the FCA.

Regulatory opportunities 

In terms of data-enabled regulatory opportunities, speakers highlighted that connectivity of data is an area in which little exploration has occurred.  With aspects of financial regulation still operating largely in siloes (examples included the FCA’s new consumer duty, anti-money laundering regulations etc.), there was a consensus that leveraging the vast pools of data that is currently available can help to establish connections across these multiple areas.  It was noted however that potential competition and privacy issues would require further consideration in order to do so.

There are also further opportunities for collaboration between firms and the regulator in respect of the nature of data collection.  Speakers recognised that data collection is a two-way process.  A collaborative approach that explains to firms why certain data is being requested by the regulator and how such data will be used was thought to be more beneficial than requests with little context.  With additional context, firms could better focus their resources on obtaining more relevant, quality data to respond to such requests.

With developments in technology and data collection, there is also a shift away from template-based data collection to data collection on a more granular level.  Firms are better positioned to analyse and organise their data from multiple perspectives (e.g. by region, product type etc.).  Caution is still required however as it was noted that data taken out of context can potentially be misleading where the full story with regard to the causes of trends in the data is missing.  Focus therefore needs to centre on learning how to provide the data on a granular level alongside the context sitting behind the data. 

In general, all speakers agreed that there were further opportunities for the regulator to become more forward-looking to address known problems, shifting away from analysing crises occurring in the past.  Innovation on the regulator’s part includes moving more quickly at the outset of a potential issue so that data collection occurs before it’s too late.   

In terms of how smaller firms can leverage data effectively for scale, data standardisation across existing standards was highlighted as key to enable smaller firms to leverage their systems to more efficiently gather and report on data.    

Data bias

A significant portion of the discussion focused on identifying the source of bias in analysing data.  An example given was in using data relating to a particular demographic in the context of a new demographic to inform marketing approaches, which may appear as bias after-the-fact.  It was also noted that there are underlying inherent issues of omitted variables and large data sets assumed to be representative simply due to their size. 

Discussions also focused on how firms should determine which types of data are acceptable to use in decision-making compared to those that should be protected.  This was particularly relevant in the context of pricing for risk and making lending decisions.  It was generally agreed that further consideration and discussion is required to assess the factors that should (rightly) be out of scope to prevent bias versus those that are required to adequately measure risk. 

In respect of AI, it was clear that the regulator will need to outline principles moving forwards to provide guidance on the right kinds of behaviours it expects to see.  Firms must also consider the data sets that they will use in their AI models and how they ensure data is representative, as well as ongoing governance.

What the FCA and firms can do

As highlighted, firms would generally benefit from gaining a greater understanding of the reasons behind the regulator’s request for particular data and how it will use it.  While it is understood that all fields requested are required, knowing which are most critical can help firms utilise resources in the most efficient way. 

Firms also appreciated how the regulator was engaging with industry on issues and taking a proactive, forward-looking approach, and further engagement and collaboration between the FCA and industry was encouraged.

With data and AI likely to play an increasingly pivotal role in financial services in the coming years, regulation governing how firms use data and AI in an ethical and transparent way will not lag far behind.  It is clear that firms are already considering how they can leverage their systems and technology to use their data most effectively.  However, firms should also start to consider internal frameworks and governance over data practices to both future-proof their data strategies and generate better customer outcomes.