The growth of AI is set to be one of the most disruptive technologies in the coming years for businesses of all sizes. Even in these relatively early days of AI, it is already affecting customer experience.
The younger generation have grown up in a world where they are used to a quick and effortless service, and as a result, will lose interest with anything that takes too long or requires too much effort from them.
Here, Prof. Steven Van Belleghem, an expert in customer focus in the digital world, examines the evolution of customer service under Artificial Intelligence.
An insight into the future of customer service
While researching my new book, Customers the Day After Tomorrow, I wanted to find out more about how we can expect AI to affect customer experience and the role human agents play.
One of the world’s leading experts in the field of artificial intelligence is Pieter Abbeel, a professor in AI and robotics. My interview with Pieter revealed some fascinating insights into what we can expect as customers in the future.
Separating AI fact from fiction
First things first: I wanted to separate fact from fiction. Things often move so quickly in the field of AI, and there is so much speculation in the media, it is sometimes hard to know where we really are in terms of robotics, so I asked Pieter for some clarity on this subject.
He told me that at present, AI is focused on a very specific domain. The technologies are skilled at analysing specified set of data, but not so great at using wide sets.
They have learned to understand a complicated pattern of inputs and outputs to figure out what they should do, but still need to expand on their ability to analyse large amounts of data from unrelated fields to make more complex decisions.
How broadly can AI measure data?
I wanted to know how broad AI could go in analysing data. Could a business giant like Google, for example, measure the entire costs of all their revenue streams and improve the cost structure of their business accordingly?
“It would be difficult for an AI to do at this stage,” Pieter tells me. “AI would be better used for things like customer support, rather than huge tasks such as analysing a company structure.”
Customer support is a more specific domain to master, so artificial intelligence is much better suited to this. They can slowly learn what the correct answers are through listening to the responses that agents give.
The more questions that agents answer, the better the understanding of the machine: the AI learns to match the questions with the correct answers until they are able to help customers independently.
“In a class in Georgia Tech, the teacher had an online help form for the students to fill out. One of the teaching assistants would answer any questions that came through the form – and the students had no idea that it was actually an AI.”
The key was that the questions the students were asking were from a fairly narrow field. AI can similarly be very useful in customer service as we can learn the common questions and problems, but they might not yet be developed enough to run across entire companies.
I wondered whether we might soon expect human-like bots to serve customers. Pieter says that this is something is something that should be coming soon, and is already an area receiving huge amounts of research and development budget.
Amazon Echo are aiming to have a voice function that you can easily have an extended conversation with – but the AI required hasn’t quite reached this point yet.
At the moment, to build an AI to answer a simple, contained question is very achievable, but when a conversation refers back to things that have been said in the past, it becomes too complex.
Once AIs are able to do this they will be a lot more realistic and easier to interact with. The aim is to one day be able to have a natural conversations where the AI is able to successfully maintain a dialogue and provide meaningful answers.
Robots as part of the family
At some point in the future, Pieter believes that an AI will become almost like another part of the family – although he is unsure of the timeline for this.
He can see a time where AIs are jumping in and participating in conversation, but to achieve that level of natural speech, the AI would have to get to 99% speech recognition. At the moment it is at around 90% speech recognition.
If the level of speech recognition is too low, people easily get frustrated. “It needs to be better than typing to enjoy having a conversation,” Pieter says.
The challenge is that the final 10% accuracy is the hardest part to achieve, so while we can expect this level of speech recognition from AIs in the future, it is difficult to estimate when it will reach the mainstream.
And until that time, the key role of AI will be all about ‘intelligence augmented,’ improving relatively simple processes and enabling humans to deliver truly outstanding service in more complex situations.
About the author
This article has been written exclusively for ByteStart by Prof. Steven Van Belleghem, an expert in customer focus in the digital world. He is an award-winning author, and his new book Customers The Day After Tomorrow is due to be published in December 2017. Follow him on Twitter @StevenVBe or visit www.stevenvanbelleghem.com
More from ByteStart
For more on the subject, try these other ByteStart guides and articles;