Out-Law News 2 min. read

Machine learning in financial services charted in UK study

AI


Many large banks and insurers in the UK have already deployed machine learning in their operations and use cases are expected to "more than double within the next three years", according to a study by the Bank of England (BoE) and Financial Conduct Authority (FCA).

The authorities' survey of more than 100 firms across UK financial services found that machine learning applications have "in many cases ... passed the initial development phase". More than half of the applications acknowledged by firms in the survey are already in use.

Machine learning is most commonly being used in UK financial services within "back-office functions, such as risk management and compliance", including for the purpose of addressing money laundering and detecting fraud, the report said. However, there is increasing application of the tools in "front-office areas, like customer management as well as sales and trading", it said.

Specific use cases vary across sub-sectors of the financial services market, with machine learning more prevalent in general insurance distribution and underwriting than in insurers' back-office functions currently, for example, it said.

More than half of the 106 survey respondents said they have a dedicated strategy for the research, development and deployment of machine learning, while 19% already have their own internal centre of excellence for promoting machine learning deployment across the organisation or are in the process of establishing one, the BoE and FCA report said.

Some of the main constraints that firms cited to deployment of machine learning in UK financial services include that their legacy systems "are not conducive" machine learning and that there is a "lack of access to sufficient data". A further barrier highlighted was "the difficulty of integrating [machine learning] into existing business processes", according to the report.

While the survey found that most firms (75%) do not believe UK financial regulations are "an unjustified barrier" to the deployment of machine learning, some firms do believe further guidance from regulators could serve as "an enabler" of such technology.

"Additional guidance could potentially help firms design controls, model risk management frameworks and policies for [machine learning] applications, as well as understand regulatory expectations for specific use cases," the BoE and FCA said in their report.

Luke Scanlon, an expert in financial services and technology at Pinsent Masons, the law firm behind Out-Law, welcomed the fact that the FCA and Bank of England are actively engaging with the market to determine how machine learning is currently being utilised.

"This engagement will assist the authorities in developing policy and regulation that is aligned to the practical application of the technology," Scanlon said. "It is important for firms who are looking to, or are already utilising this technology, to strategically consider how to effectively implement machine learning within their businesses, given the rate of change in this field. Firms should consider what governance procedures they need to ensure that the application of machine learning is understood by the business and risks are appropriately monitored."

In their report, the BoE and FCA characterised machine learning as a sub-category of artificial intelligence and described it as "a methodology whereby computer programmes fit a model or recognise patterns from data, without being explicitly programmed and with limited or no human intervention".

The authorities said the findings from its report constitute just a "snapshot of [machine learning] adoption" in UK financial services. They highlighted the fact that their survey sample was "skewed somewhat towards larger firms" and said it can be surmised that some of the 181 firms who elected not to participate in the survey did so because they do not apply machine learning in their organisation.

We are processing your request. \n Thank you for your patience. An error occurred. This could be due to inactivity on the page - please try again.