21 Dec

5 Blockers Preventing Scaled Adoption of ML & AI in Financial Services and How to Overcome Them

Tom Jenkin

So far we have provided a series of examples as to where financial services can reap the benefits of ML & AI. In 2019, the BofE published a study that reported on the state of ML & AI adoption across financial services in the United Kingdom. As part of that research, many of the use cases discussed in this paper were of prevalent focus for financial institutions. Namely, areas such as AML and fraud detection were listed as the big-ticket items where they could unlock the most value for their businesses. Whilst additional areas of adoption had been mooted as eliminating operational waste through automating frequently executed business tasks and processes with RPA capabilities.

There are 5 BIG constraints that are hindering scaled adoption of ML & AI

There are ongoing constraints and challenges to scaled adoption that are preventative factors to accelerated maturity across financial services. The 5 key themes that the study highlighted as central constraints are:

1) Legacy systems: Legacy technology stacks were identified as a significant barrier to entry for scaling enterprise-wide adoption of ML & AI in financial services. Admittedly, this might only be prevalent for enterprise organisations that have IT infrastructure and architecture components that have been built from decades of mergers and acquisitions.

2) Insufficient access to high-quality data: As with all ML & AI implementations, firms highlighted a growing concern around the availability of high-quality data that was accessible and easily consumed in an auditable manner.

3) Internal data governance processes: In addition, organisations were concerned that their own internal data governance processes would stifle their ability to leverage ML & AI solutions during moments of significant financial opportunity.

4) Increased risk management controls, owing to model complexity: Over the course of the paper, firms also highlighted increasing concerns around the unbounded potential of ML & AI. And specifically, how the growth in use cases would ultimately result in a corresponding complexity growth in the management overhead and risk controls for governing models from concept to production.

Indeed, the view of firms appears to be that the introduction of scaled ML model usage doesn’t introduce new risks to the financial system. However, it does amplify the risk surface for known problems, should an ML model or AI-driven application begin to perform poorly and disrupt market stability.

5) Lack of explainability in model results: Firms raised concerns about being able to fully explain the results and outcomes that their ML models had generated. Whilst organisations also mooted a need to increase their maturity of model governance and lifecycle management. In order to ensure they were able to demonstrate the validation, testing and ongoing performance of their models across the route to live.

What can firms put in place to address these challenges?

Having discussed the exponential use cases for ML & AI across financial services and the provisional challenges to scaled adoption. We believe there are two immediate steps that organisations should take in order to seed foundational capabilities that can underpin their intentions to apply ML & AI-driven solutions in their business. They are;

1) Apply a data mesh approach: To address their challenges around poor quality data, legacy systems and sub-standard governance processes, we believe that by applying a data-mesh approach to their data strategy that financial services firms can unlock business value with ML & AI.

2) Implement MLOps Capabilities: In order to address their concerns around model lifecycle management, traceability, firms should adopt an MLOps approach that is fully aligned across the intersections of people, process & technology.

We will expand on the data mesh approach in our next blog. In it we will discuss the following subjects;

1) How firms can tackle big data obstacles for scaled ML & AI adoption.

2) Why high quality, clean data is the special ingredient for ML & AI success.

3) Why a data mesh approach makes sense.

4) How a data mesh can support a multi-cloud agenda for regulatory needs.

Interested in seeing our latest blogs as soon as they get released? Sign up for our newsletter using the form below, and also follow us on LinkedIn.

Latest Stories

See More
This website uses cookies to maximize your experience and help us to understand how we can improve it. By clicking 'Accept', you consent to the use of these cookies. If you would like to manage your cookie settings, you can control this in your internet browser. Find out more in our Privacy Policy