Five things financial institutions must know before using AI and machine learning to comply with regulations

From: RegTech Analyst

Machine learning and artificial intelligence is often hailed as the next big thing in RegTech, but what can you do to avoid them becoming empty buzz words.

Say artificial intelligence and people are prone to get mental images of Arnold Schwarzenegger in his prime impersonating a killer robot from the future. Tempting as these apocalyptic images may be, the reality is far less sinister. In fact, machine learning and AI are two of the technologies most frequently used among FinTech and RegTech companies.

From boosting regulatory compliance to estimating people’s credit scores, there is no shortage of ways financial institutions can utilise them to boost their performance.

However, a recent panel at the Global RegTech Summit revealed that there were five things companies should consider before tapping into these technologies.

Jigisha Lock, head of global products and core compliance services technology at Credit Suisse; Søren Rode Jain Andreasen, chief digital officer at Danske Bank; Kayvan Alikhani, CEO and founder of; and Travis Schwab, CEO and founder of Eventus Systems took part in the discussion.

Jekaterina Govina, executive director of supervision service at the Bank of Lithuania, moderated the panel’s discussion.

The first thing that the panel made clear was what these technologies are not. “I think it’s a bit simplistic to say that AI and machine learning is this magic pill that will immediately reduce costs,” said Lock. “I think it’s a bit more nuanced than that. And while the tools may or may not be cheap, the efficiency gains I think will take time to materialise.”

Key to getting the most out of the technology, she added, is not to focus on what it is, but the problems it can solve. “What is the key gap?” Lock asked. “What is the problem that you’re trying to solve? Are we solely trying to solve for high volumes of false positives? If so, then let’s focus on machine learning. By training the machine to identify the potential of false positives with a high degree of confidence, you’re freeing up compliance officers and redirecting them to more investigative capabilities.

“[If] you don’t have a problem of false positives, but you really have a problem of repetitive information that you need to source over and over again, then maybe investing in automation and robotics is the right way to move forward. So unless you understand your key risk and control, and then make a decision on what tools you need, you will never really truly be able to harness the potential of RegTech.”

The second thing to understand when it comes to new technologies is what they can be used for. “The main role of AI-driven solutions is really being streamlining and automating those business processes that are highly manual, very time consuming [and], frankly, error prone, non scalable, and, as a result of all that, costly,” said Alkikhani.

AI and ML can be used to “reduce the noise” and automate time-consuming processes. “Imagine a 1,000-page rulebook coming out from the FCA and you want to throw bodies at it and use a manual approach just to [do] something [that’s] very straightforward and required as obligation and assessment,” said Alkikhani. Instead of doing that and bringing out the old school Excel sheets, he advised banks and other financial firms to invest in an AI solution that could do that for them.

That brought the conversation to the third thing to consider before investing in new technology, whether to use in-house developers or to go to a vendor.

“[This is an] age old question and there is no clear cut answer to this, at least from a surveillance standpoint,” said Lock. She explained that the decision must be made out of the firm’s own situation. That being said, she seemed to lean towards enlisting outsiders to get the job done in some cases.

“If there are risks that are inherent to the industry and you have vendors with proven capabilities that have demonstrated with flexibility and adaptability to changing regulations then that is a path that we should definitely explore,” she said. “But firms will always have risks that are unique based on the business model products and sectors they serve. So there will be always an element of that custom development and custom solutions to address firm’s specific risks.”

Alikhani said had noted that it was common for companies to adopt a 80-20 split, which saw “80% bring in technology [that’s] best of breed,” which resulted in these companies “operating at a significantly faster pace than an organisation can innovate.”

Looking the other 20%, he suggested that they were “building things in house owning, proprietary or very customised personalisation that belongs to the company.”

Even though that’s the trend he’d noticed in the market, Alikhani warned that things aren’t always as easy as they may seem. “We’ve seen some lethal mistakes of trying to make rebuild something that already exists commercially,” he said, arguing that there was little point in innovating internally if several vendors had already specialised in creating those solutions and bringing them to the market.

The fourth thing to be aware of is that implementing AI and machine learning solutions will always mean more work, even if you use a vendor. “Every time we sell something to an institution, they have to do work,” said Schwab.

And that is something Eventus Systems always try to make clear to their clients. “That expectation needs to be there, [that] there is no magic bullet like this,” he explained. “I don’t care what tech you’re using, you can’t just throw ML at a set of data and magically, it works because it just does. And so I think that expectation setting so early in the process is critical.”

This led to the fifth thing to consider, which is that digitalising parts of the process could result in firms having to rethink how to use their employees. “Fundamentally, firms need to restructure their workforce,” said Lock. “And while the number of resources required will definitely drop, employees will need to be upskilled, the skill set required not only to analyse the output, but to actually manage and continue to improve the capabilities will be quite different. So there will need to be an investment in talent and I think a dynamic workforce will be necessary to truly benefit [from] RegTech.”

Copyright © 2020 FinTech Global

Enjoying the stories?

Subscribe to our daily FinTech newsletter and get the latest industry news & research


The following investor(s) were tagged in this article.