Could AI actually make customer service better?
Opinion
If AI can read our emotions and respond with the right tone, while solving our problems at lightning speed, do we need humans for customer service?
In Closer to the Edge, our report from startup conference Slush, we remarked upon a set of emerging platforms employing profound human values as their operating system.
We named this trend “gentle tech”, describing tech solutions that were using humanity as an antidote to artificiality, with kindness and empathy baked in.
When we talk with our clients about gentle tech, they get it immediately. In an era of war, pandemic, climate change and inflation, consumers have enough on their plate without the additional frustrations of unnavigable websites and long hold times.
Brands have a responsibility to ensure they are not adding to their customers’ woes with subpar customer service, meaning they need to care about dissatisfied customers and play back those frustrations to demonstrate empathy and understanding — and, of course, propose a solution.
The best customer service is two humans truly connecting.
However, I recently came across an intriguing book that implies the opposite could be true: that, in the future, the best customer service could come from AI, not humans.
We are talking about the field of affective computing, where computers can detect our feelings, understand our woes and emotionally engage with our issues — and then solve them with maximum efficiency.
Is that possible and should we accept it?
The future of customer service
The book in question is The Future of Professions by father-and-son team Richard and Daniel Susskind, who start by arguing that the idea of the “profession” is under threat from technology.
“Professionals”, they state, can be increasingly viewed as a cadre of experts who protectively hoard their knowledge, all while charging other people a premium to access it.
While many of us may have no problem with this “grand bargain” — “you know something that I don’t and I’ll pay you to apply your knowledge to solve my problem” — the Susskinds propose that technology will systematise, productise and redistribute that expertise and make it accessible to more people, more often, for less expense.
For the authors, the roles of lawyers, accountants and university lecturers will start to look radically different, with automatic will-writing services and accounting software pointing to a future where professional services are disintermediated, democratised and de-premiumised.
Chief among the drivers of this change are new AI interfaces capable of distilling the wisdom of millennia-old professions and liberating it from behind the “paywall”, making it easier for people to access.
It’s not unreasonable to assume this transformation might extend to customer service, too, with AI increasingly capable of triaging queries and serving up best-practice responses to solve consumer complaints.
Initially, this might seem unworkable. Consensus states we prefer human interaction over artificial ones: we want sympathy, we want empathy and we want a human to say: “Oh my goodness. I am so sorry. We have let you down. Let us put it right for you straight away.”
Surely an AI could never fulfil this function: to fully engage with humans the way other humans can.
But to this, the Susskinds are saying: “Possibly.” And here’s why.
The rise of affective computing
First, the reality is we already train people in the art of customer service via processes, scripts and playbooks. Couldn’t we convert customer service playbooks into a decision tree of AI responses that de-escalate and pacify in the same way?
This may be no more difficult for an AI than responding to any other query. Affective computing means AI will be able to read our emotions by inferring mood from our tone of voice, word choice and facial expressions.
Understanding when it has angered, it can switch its language to placate, while in the background also working logistically to solve the actual complaint — all without a human customer representative.
New AI entrant Hume, for instance, prides itself on both the realistic tone and timbre of its AI voice, but also its ability to decode the tone and timbre of its human interlocutor. Its bright and sunny responses sound a world away from the synthesised robotic voices of the last decade.
There is an obvious objection, of course. A computer selecting from pre-programmed responses has no real sense of the meaning of its words. This is mere “customer service karaoke” with no emotional content — the very thing we want from our interactions.
But the Susskinds counter this. We can never truly know how deep human sympathy goes, they say.
How much does the solicitor working on the sale of our house really care for our wellbeing? Isn’t he just running through a complex legal checklist in his head?
Take counsellors too. Surely, if they became totally emotionally invested in their patients, they would no longer be able to offer a valuable subjective view. Isn’t that necessary emotional detachment the same as the detachment of an AI?
Also, is human-led customer service currently faring any better? Customer service scores are at an all-time low, costing UK businesses £11.4bn a month, according to the Institute of Customer Services.
Sitting in front of a database dealing with irate customers is a tough and stressful job. Like the solicitor, there are checklists to run through, except with a time pressure. Like the counsellor, total emotional involvement could be dangerous, but neutrality is difficult if you are being yelled at. The ability to follow due process and detachment are key requirements.
But could AI and affective computing be our saviour here: running through those checks at lightning speed while being utterly unfazed by an irate customer, at the same solving the actual problem?
A recent report from the UK Institute for Public Policy Research think-tank said that customer service was the second most at-risk role from being replaced by AI. Is this pointing to a future where interactions with non-humans are normalised — even preferred? Will efficient, emotionless exchanges be the standard moving forward?
Well, fortunately, probably not.
Human x machine, not human vs machine
Despite the ability of machines to replicate human endeavour and improve human experience, history shows us that there are limits to what technological augmentations we will accept as a society. Even when an automated alternative is “better” than its human equivalent on paper, we still drag our feet over the upgrade.
Recently, we have seen supermarkets replace self-checkouts with human cashiers. We’ve seen chatbots offer users the chance to cut straight to a real-life representative. We’ve seen the pandemic demonstrate the limitations of remote learning and the importance of physical socialisation. We’ve even seen a resurgence of “dumb phones”, as people seek the peace and quiet of an “off-grid” monochrome screen — a wilful downgrade to our personal technology.
Sometimes, we just want the less efficient option if it brings intangible human benefits.
In the same vein, although there may be a role for affective computing in customer services, it’s not certain that it can be a direct replacement for human-on-human interaction — and certainly not physical interaction. We want to look people in the eye, not look them in the AI.
We want certain interactions to be with humans who have lived experience and at least the capacity to care. Would we ever want an AI doctor delivering us bad news? I would venture no. Would we ever accept an AI customer services bot telling us “I’m sorry, but there’s nothing more we can do for you”? I would venture we would immediately seek out a human.
On this point, we return to the Susskinds for further clarification. Even though they have built their thesis on the automation of the professions, they are keen to rebut the idea of replacing jobs wholesale and prefer instead to talk about replacing tasks within jobs.
We should use machines to augment the endeavours of real people, they say. Get the union of artificiality and humanity right, and this might make for better customer services.
This is a better way to think about it. It’s better to use AI to prevent issues occurring in the first place: ensuring that our product is in stock, our discount voucher works, our package tracking is accurate, our billing is correct. We must not forget that what we want more than empathy from customer services is to not have the need to speak to customer services.
Then, and only then, if we can free the system of most glitches first, can we place more emphasis on the human-to-human element when things do go wrong. This is humans and AI working in tandem effectively.
We are not there yet and AI will no doubt continue to improve. But I’m hopeful that humans can too.
Phil Rowley is head of futures at Omnicom Media Group UK and the author of Hit the Switch: the Future of Sustainable Business. He writes a monthly column for The Media Leader about the future of media.