We use cookies on all our websites to gather anonymous data to improve your experience of our websites and serve relevant ads that may be of interest to you. Please refer to the cookies policy to find out more.

By continuing, scrolling the page or clicking a link, you agree to the use of cookies.

Artificial intelligence and data analytics in banking and finance

16 March, 2020Ouida Taaffe

Srikanth Velamakanni – the Co-Founder and Group Chief Executive of artificial intelligence (AI) company, Fractal Analytics – tells Ouida Taaffe about the limitations of AI, the importance of data, and how financial services can make the most of both.

Old-fashioned robot grimacingArtificial intelligence (AI) is a bit like teenage sex – everybody is talking about it, but not many firms are really doing it.

And one of the reasons for this is that AI itself is often not well defined. People can start out with a data objective and expect AI to fill the gap. AI, however, can only answer very specific questions.

“AI works well when the application is very narrowly specified,” says Srikanth Velamakanni of Fractal Analytics. “It can’t give magical answers.”

AI is, nonetheless, often a source of awe and fear – particularly as it is expected to be a potent source of automation.

“Over the last 300 years, technology has been both displacing old jobs and creating new opportunities,” says Velamakanni. “However, unpredictable physical activities and emotional intelligence cannot be automated. What we automate are tasks.”

Framing problems for AI

Velamakanni says that artificial intelligence was initially used to optimise a particular utility function, such as accuracy.

“Now it may be that optimising is the easy part,” he says. “What is difficult is knowing what the utility function is, knowing how to frame the problem to avoid silly mistakes.”

An example Fractal Analytics gives of careful framing is working out the root causes of customer friction in a digital purchase journey.

Here, advanced AI and bigdata engineering can be used to mine millions of customer interactions at the click-level to find out what is going on – and how best to intervene.

Consumers can already see AI solutions that do not always provide answers they expect.

Velamakanni points to video services that select content for users. However, even if the selection is not optimal, such AI systems have a lot of control over the content viewers access.

“Eventually, we might have a scenario of my AI versus your AI,” says Velamakanni. “In recruitment, we are already there. Fractal hires only around 0.4% of the people who apply, so we use an algorithm to sift CVs. People can get advice on how to beat the system.”

The basic issue for effective use of AI, Velamakanni says, is framing and solving the right problems.

Banks and AI

A recent blog post from the Bank of England argued that AI and machine learning are particularly powerful. And this is because they’re “better able to account for complexities that might be present in the data”.

But do banks have enough, and good enough data, to feed good AI systems?

“Banks have amazing data. In the early 2000s they were also ahead on AI risk models, especially in the US, the UK and Australia. Then, they lost ten to 12 years to the financial crisis,” says Velamakanni. “Since 2003, tech companies have been redefining the art of the possible in data analytics.”

Is there demand from UK banks for AI expertise?

“There is a tremendous amount of interest, but they are also quite well served – both internally and by a huge number of people wanting to serve them. Their needs are vast…and it could be game-changing for them.”

Related content

Centre for Digital Banking and Finance