For years, people have complained that when they call their bank's call centre they have to listen to a robotic, disembodied voice telling them to key 1 for current accounts, 2 for new business enquiries, 3 for payments or 5,423,534,667 for complaints.
I jest, obviously, but you know what I mean. Eventually, you do get to speak to a human being. They may, or may not help, but at least they are human. However, as we know, human beings are prone to all sorts of failings – transferring money to the wrong account, swearing at customers and generally providing poor customer service.
Moreover, human beings cost money, require comfort breaks, coffee and sometimes a squad of HR professionals to keep them happy. So it’s perhaps understandable that banks, looking to keep costs down, want to use robots to communicate with their customers. The problem is that they don’t (yet) work properly.
Take swearing at customers for example. One of the world’s most “intelligent” computers, IBM Watson learned to swear a few years ago. More recently (March 2016), there was the infamous Microsoft AI chatbot that went “rogue on Twitter, swearing and making racist remarks and inflammatory political statements.”
However, the real proof of the pudding is in customer service. For all our human faults, with some training we’re usually pretty good at this, whereas robots haven’t quite got it yet. As the Telegraph reports today, “Two key problems have presented themselves so far. The first is that the artificial intelligence robots are new and, frankly, not very good.”
That said, as the Telegraph also reports, the former CEO of Barclays believes “half of all jobs in banking could be chopped in the next decade as automation takes hold, underlining the scale of potential transformation.” Although – as my next paragraphs shows – there is still some way to go, this will eventually happen.
At present, the AI “robots” don’t always get it right, but they are improving. RBS’s early efforts in this field led to 90% of the robot’s answers being wrong when they were asked to answer 20 questions, but now they can answer 400 questions with 90% accuracy. The only problem is that if the 10% it gets wrong are fundamental ones (e.g. “how much money have I got in my account?”) then that’s no use. However, help is at hand. Way back in 2012, it was discovered that swearing at Apple’s automated customer service led to you being put through to a human being quicker than you can say “expletive deleted” so, while we are by no means encouraging profanity, there is what our techie friends call “a short-term workaround”, should you wish to use it! However, be warned, we suspect it doesn’t work in every instance…
Ami Islam, Be-IT Resourcing