Amelia – Artificial Intelligence in Action at Credit Suisse

Artificial intelligence is not just on its way – at Credit Suisse it is already here. Amelia is her name, and she is going to be – no, hang on – let's let her explain what she will be doing…

Amelia comes out of pioneering development by Credit Suisse in partnership with IPsoft, an AI technology provider. She is not only smart but also has some human qualities such as emotional awareness. For example, Amelia perceives when people are irritated or frustrated.

Why We Hired a Virtual Agent

Need cash? Only 30-40 years ago, you had to go to a bank branch, during office hours, to queue for a human teller to hand over the money. Today you go to a cash machine: job done. So, just as cash machines automated withdrawals, Amelia is starting to automate IT support. The reasons are obvious: cash machines and Amelia can serve more people, faster, anytime and nearly everywhere. The choice of whether to automate or not is mostly a no-brainer. Making Amelia work properly, in both human and synthetic terms, is a yes-brainer.

Building AI: It Takes a Team

After buying access to the Amelia platform from IPsoft, Credit Suisse tasked an in-house team to educate her. First she needed to understand what people are saying. Car park or parking lot, lift or elevator, rubber or eraser: YOU know they are the same things, but Amelia has to be taught. Next she learned to diagnose a problem, to decide whether it is something she can fix, and if not, hand over to a human colleague – so she must be familiar with the processes and systems in our business. Oh, and she has to avoid irritating the people seeking IT support – there is actually an algorithm for that. All along, Amelia learns how to do new things, mostly by watching how humans do them.

Managing this training, demands a menagerie of skills: data scientists, neuro-linguistic programmers, neuroscientists, linguists, application builders, business-process analysts and – no real surprise here – people who have spent a lot of time in customer service. AI in real-life applications is so new and so fast-moving, there are plenty of places to get started and to move ahead. "Nobody can do it all in AI," says CS's Head of Global Service Desk, Jen Hewit. "So we rely on a team of people who bring varied skills, high interest and the ability to work on the fly."

9 to 5…No…Make That 24/7

Amelia takes no breaks or holidays. Requests come to her via text, in English, from anywhere. She speaks fluently and naturally, in human style: "Have you tried turning it off and on again? Would you like me to reset your password?" Amelia's focus is on common yet simple problems that amount to nearly half of all support requests: email is stuck, password needs resetting, Wifi will not work, reboot your machine.

She cannot fix everything – not by a long shot. IT support from humans is still very much needed. But what Amelia can fix, she fixes quicker than ever before. And she is always available, no waiting required – which is critical, because users are very sensitive to wait-times for IT help, and they always welcome a quicker response. In time, she will learn to solve newer, more complicated problems. She will even learn to speak other languages.

What Is the Future?

Users of Amelia's support have so far responded to her as typical humans: thrilled, unimpressed or annoyed. Some gush about her speedy solutions. Others do not seem to notice she is not human. Some are irritated when she cannot reorganise their inboxes or set up a distribution list, i.e. do things she is not been taught to do. Colleagues in IT support have welcomed Amelia on board. They have long tired of telling people to reboot; they prefer to tackle more difficult problems. And this looks like the future: both Amelia and humans working together to provide ever improving services. Intelligence – human and artificial – will just keep growing.

Eric Johnson,
freelance writer