Artificial intelligence needs a human touch
Banks' AI programmes are only as good as the data underlying them and the intelligence of the human beings doing the interpretation, writes Brian Caplen.
Every technology goes through a period of hype when huge claims are made about its disruptive potential. This is followed by a cooling off period when a clearer assessment is made of its realistic application and worth.
Blockchain has been through this cycle and while there are hundreds of blockchain projects being worked on, and many exciting developments afoot, few now believe that it will revolutionise finance from top to bottom.
The same is true of artificial intelligence (AI). Huge claims that 30% to 50% of banking jobs would disappear are turning out as something rather less ambitious with projects that impact specific parts of the business. A recent survey in the Financial Times reported a mixed picture among banks as to the scale of their investment and what they expect it to achieve.
The important thing for banks in devising an AI strategy is to remember that outcomes are a function of how good the data is and that it will still require human beings to interpret it. Some banks have employed highly paid behavioural psychologists to make sense of the patterns thrown up by AI reports. This is critical in areas of fraud detection, where information will be passed to the regulator who expects to receive an explanation not a computer print-out.
With all the excitement around new banking services and platforms, much attention has been given to front-end applications such as automated assistants and chatboxes. In fact, many of the best use cases are middle and back office, such as detecting fraud, improving risk management and simplifying compliance. In these examples, routine work looking through data lakes for patterns or scanning masses of documents can be done more effectively by machine.
HSBC just announced an initiative to use AI to assist in the prevention of money laundering, fraud and terrorist financing.
Regulation and compliance, where banks have seen a huge rise in costs and a subsequent drop in productivity, is an area ripe for AI investment. But only if the human talent is on board to make sense of the output. Otherwise the regulator may suggest scrapping the programme and re-employing all those staff that have been let go.
Brian Caplen is the editor of The Banker. Follow him on Twitter @BrianCaplen
Register to receive my blog and in-depth coverage from the banking industry through the weekly e-newsletter.
