How far should you automate your customer service?
Updated: Sep 29
Way way back in 1637 Rene Descartes was doing what he did best – philosophizing. Could a machine become so intelligent that it could replace a human? These were very early days to be drawing conclusions about the power of artificial intelligence but Rene was good at thinking ahead.
”We can easily understand a machine being constituted so that it can utter words, and even emit some responses to action on it…but it never happens that it arranges its speech in various ways in order to reply appropriately to everything that may be said in its presence.”
Fast forward nearly four hundred years and it looks like Descartes was almost right. He may never have envisaged chatbots but he did understand how the human mind was not something that could be replicated so easily.
Of course, plenty have tried and plenty more are still trying. Some are even getting close.
’A bot is only as good as the information provided to it and if you can provide that bot with enough intelligence, then it can do some of the job,” says Magnus Zimmerman, CTO of Subtonomy.
”In telecoms, where we work, this can work for very straight-forward questions like “what’s my current bill?” or “how much data do I have left this month?” but it can also be for more technical questions. For example, if there is a data roaming issue in Spain and the bot is made aware of this, then a customer contacting support from Spain can be immediately informed.”
So far so good but Zimmerman is also quick to point out that unpredictable customer queries can’t be solved by bots today.
”Bots are just not quite there yet to take on a customer’s query 100%. I think they may be in the future but right now we see a lot of possibilities today where bots can resolve simpler issues or even start a conversation with a customer before passing it on to a real person. When things get complex, there are just so many unforeseen variables in human communication for a bot to manage today.”
In a recent Wired article*, the author compared two ’meeting organizer’ services, one using a chatbot and one using a person to solve the problem.
In an effort to throw the bot off guard, the author, who requested a meeting, refused the first suggested time claiming his father had died. The human offerred her deepest condolences.
The bot’s response?
”I’m so sorry, but I am unable to respond to your last message. It’s possible that it isn’t related to scheduling a meeting or that I was unable to understand it.”
And therein lies the problem. Research published in 2012**showed that we’re quite happy to use a digital channel to find information but if we need a creative solution or if we get stuck on something, we will always go to a human. The task or request has to be completely clear and unambiguous if the digital channel is going to work.
Bots may seem attractive from a cost-savings perspective but then the entire cost needs to be taken into account. If your bot doesn’t understand your query, you’re going to get annoyed pretty quickly.
T-Mobile recently announced that it was getting rid of the bots and dropping all automated call menus from its customers. The reason? Widespread frustration with bot-based systems and lack of contact with care representatives.
”It’s worth pointing out that we can add a lot of intelligence into the network itself these days” says Magnus Zimmerman. Subtonomy’s own platform picks up failures in a network and can give recommendations to customer care reps. ”When the support team is equipped with the right data, you can dramatically cut down the amount of time they spend with each customer,” says Zimmerman. ’The platform can be of great assistance and do a lot of the job but it’s still important to put a human touch on your customer care.’
So, to answer the question that started all this article off. How far can you automate your customer service? As of today, the recommendation is to let people and bots work together to get the best possible results. Tomorrow? 100% automation is an absolute possibility but artificial intelligence needs to come further in its development.