Can AI resolve customer service disputes?

Discussion
Jan 16, 2017
Matthew Stern

A phone call with a customer service (CS) representative can devolve into a nightmare for any number of reasons. But to the extent that problems are caused by a particular customer and CS representative just not getting along, one startup aims to ease the tension with artificial intelligence (AI).

The startup, called Afiniti International Holdings, has created an AI-based system that builds a profile of a customer service caller based on aggregated information. When routing a call, it uses the profile to determine which customer service associate will most likely achieve a successful interaction, according to Engadget.

The system pulls from a customer’s purchasing information and customer service contact history as well as external for-purchase informational databases and from whatever publicly available information there is about them on Facebook, LinkedIn and Twitter. It also examines census data. The company has implemented its AI in 150 call centers.

Retailers and tech companies have recently been experimenting a great deal with AI on both sides of the customer experience. On the customer-facing end, for instance, companies like 1-800-Flowers have enhanced their search capabilities with IBM’s Watson. On the back end, Google has introduced an AI-based product that can understand and rank how angry callers and emailers are so that companies can handle their calls appropriately.

The Afiniti AI has raised concerns from privacy advocates, as reported in Engadget. While Afiniti claims associates can’t see the data the AI processes, some worry it opens the door to subtle discrimination. And while a positive customer service experience for a customer may mean getting a problem solved, from the company’s perspective it may mean a successful upsell.

Whatever constitutes an ideal customer interaction, many of the complaints people have about their call center experiences might not be solved by assuring they get the right rep. A 2015 survey by Nexogy cited being stuck on hold, boring hold music, multiple transfers and having to repeat information when transferred as the top four call center customer complaints.

And no amount of data can lead to a positive customer service experience if a business isn’t empowering its reps to solve customer problems.

DISCUSSION QUESTIONS: Do you think personality-matching AI can be used effectively to reduce negative customer service experiences?  What do you see as the pros and cons to retailers implementing such a solution?

Braintrust
"Although I’m a big proponent of AI in customer support roles, this implementation has the potential for error and abuse."
"Customers don’t like the idea of talking to a machine, so the AI interface would have to be able to beat a Turing Test."
"While this type of technology may eventually evolve to work well, nothing can replace excellent training!"

Join the Discussion!

10 Comments on "Can AI resolve customer service disputes?"

Notify of

Sort by:   newest | oldest | most voted
Ken Lonyai
BrainTrust

Although I’m a big proponent of AI in customer support roles, this implementation has the potential for error and abuse. Pulling data “from whatever publicly available information there is about them” is a very slippery slope. I’ve been involved with AI personality insight tools, but none that would pull public data trying to re-allocate it in a different context. A consumer that uses, say, Twitter for tongue-in-cheek quips or posts about their interests in, say, slasher movies might be misinterpreted in the system because the context is entirely different and from what I’ve seen so far, AI is not that good with interpreting sarcasm or humor.

On top of that, the whole thing sounds creepy. In this instance, I would vote for better human CS rep training and a better mechanism to escalate or switch calls to alternate reps rather than the hazards and privacy issues this platform might instill.

Steve Montgomery
BrainTrust

The role of a customer service representative is to solve the caller’s problems to the extent that is possible. In my experience most are reasonably good at doing so. Could they use better training? Sure.

However, as the Nexogy survey stated most of the complaints do not have to with the problem-solving capability of the representative but with issues in reaching one or the right one in a timely manner and with whom you can effectively communicate.

Max Goldberg
BrainTrust

It’s an interesting concept, but will personality-matching AI really help solve customer issues and create a more positive experience? Most negative experiences come from poor training and not enabling first-tier customer service reps to solve customer issues. Management should be required to call their companies’ customer service centers to see how issues are handled. AI will not solve these issues. Training and empowerment will.

Manmit Shrimali
BrainTrust

Sourcing data from publicly available information, etc. … these are precisely the reasons why seven out of 10 so called Big Data-based AI projects fail to show an impact on the bottom-line. Nobody is even looking under the hood on the accuracy of these predictions — which is really sad. Recently, we conducted an analysis on AI technology and concluded that it actually did more harm than good as it was wrongly interpreting the social feed data — there were many data holes for which AI was making incorrect assumptions.

The future is more about how AI can help self-serve using historical interactions. We built an AI engine where we took the past call records and created self-serving servicing that replaced 70 percent of agents. Even after analyzing data from the last four years, our accuracy is still at 84 percent which is good but still not 100 percent. We highly encourage retailers not to blindly rely on publicly-available data and so-called data fusion without careful considerations. Bottom-line — the tech mentioned in the article will actually do more damage.

Ryan Mathews
BrainTrust

AI systems — today at least — are only as good as their programming. So this isn’t a yes or no question. If the question is if we could redesign an effective personality matching AI system, the answer is a resounding, “yes … eventually.” If the question is if such a system exists today, then I’d say the jury is still out. Customers don’t like the idea of talking to a machine, so the AI interface would have to be able to beat a Turing Test. My guess is the majority of customer service calls are from customers with problems and so the issue of “personality-matching” is really mission critical.

Camille P. Schuster, PhD.
BrainTrust

The AI system described does nothing to resolve the top four consumer concerns. Matching personalities does nothing to reduce wait time, get rid of boring music, reduce number of transfers or reduce the number of times having to repeat information. AI will take off when it addresses consumer problems. In terms of personality mismatch, the training of call center reps to use active listening skills and empathic language will provide much greater rewards.

Ed Rosenbaum
BrainTrust

I can see the further development of AI in customer support roles in the future. But I can also see where there is an avenue for abuse. Yes, they say our personal information is secure. But we have heard this from our banks and credit card providers in the past. That has proven false. I read an article last week that St. Jude was found to have security flaws in heart devices. Why anyone would want to hack into defibrillators and pacemakers is beyond my comprehension. (BTW: I have a defibrillator implanted and my doctor did not know anything about it until I called his office.) So if the hackers are smart enough to do this why would anyone believe our information is secure in a customer support department that is connected to the company’s main server?

Doug Garnett
BrainTrust

This is a clever idea. But I’m skeptical. My agency has managed telemarketing for nearly 20 years with all flavors of clever call routing theories being presented to us. And what we’ve learned is that no matter how perfect the theory looks, they tend to fail in the reality of a call center. What I’d expect is:

  • The routing software will only rarely be able to choose best available and usually only be able to route to “somewhat better.”
  • The challenges of load leveling will negate the opportunity for improvement from software like this. Customers far prefer to talk to a human soon over waiting on hold for what can seem an endless time.
  • It’s unlikely that routing to a different rep makes more than a slight difference in call quality or impact.

My bottom line is that there will be a few situations where something like this can help. But the majority of situations will end up with a retailer paying big bucks for a solution that offers only small or negligible advantage.

Craig Sundstrom
Guest

“And no amount of data can lead to a positive customer service experience if a business isn’t empowering its reps to solve customer problems.”

Need we say more? Probably not, but let’s anyway. This sounds like it might either work great, or be a complete disaster. But I’m having trouble picturing exactly what’s involved: how does AI even know who you are … what if you’re calling on behalf of your company (or some other person)? Might all this surreptitious gleaning of info be even more upsetting that unhelpfulness?

Brian Numainville
BrainTrust

There are many concerning elements here. First, who knows what data will be pulled from the “available public information” and its accuracy. Then, what if the idea rep to help a customer is tied up for a long period of time? I suppose it routes to the “next best choice” but how good will that rep be for a given person? While this type of technology may eventually evolve to work well, nothing can replace excellent training!

wpDiscuz
Braintrust
"Although I’m a big proponent of AI in customer support roles, this implementation has the potential for error and abuse."
"Customers don’t like the idea of talking to a machine, so the AI interface would have to be able to beat a Turing Test."
"While this type of technology may eventually evolve to work well, nothing can replace excellent training!"

Take Our Instant Poll

How likely is it that solutions such as Afiniti will be implemented to route retail call center calls in the next 10 years?

View Results

Loading ... Loading ...