054: CX horror stories and AI armor for contact center agents
24 minAudio
Damien and Nikola talk rising aggression towards contact center agents and the protective role of AI automation. Highlighting alarming stats like 65% of people abandoning brands after one bad experience and a 90% increase in customer service abuse during peak seasons, they address the issue and how AI can deflect initial customer rage and effectively manage high volumes. As holiday-induced pressures mount, hear how AI offers a valuable solution to preserve both employee well-being and customer satisfaction. Follow PolyAI on LinkedIn Watch this and other episodes of the Deep Learning pod on YouTube
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
Damienshould
Nikolaabsolutely punish companies that provide you a bad service or a bad product by not returning your business to them in the future.
DamienAnd what is it something like 65 percent of people will abandon a company completely if they have one bad experience? That's it's wild to see the level. Yeah.
NikolaAnd that's why those companies and those contact center agents do their very best to. Tolerate you even when you're being absolute scum, but remember the agent is not at fault. They're just, they're trying to help you.
DamienHello, and welcome to another episode of deep learning with Polly. AI, the only podcast that is spooky. I'm just kidding. You're probably hearing a lot of spooky stuff coming in because it is the start of spooky season. I'm your host, Damien. And I'm here with as usual, Nikola Mrkšić, CEO and co founder of Polly. AI. Hey Nikola.
NikolaHey Damien, how's it going?
DamienNot too bad. Not too bad at all. As mentioned we're entering sort of the Halloween era, which also serves as the entree. I've actually seen that a few retail locations, some Christmas decor that's going up as well. So
Nikolalook, I basically look like Harry Potter with this like half scar.
DamienSo is that your costume this year? That's awesome.
NikolaYeah, no it's approved. I'm still an adolescent, but
Damienwe are, we all are at heart, especially here on deep learning with Polly. Yeah. But I did think that this was a good opportunity to talk about two things. One, some scary stories that are coming out of contact centers with regards to what is being reported as a pretty significant uptick in aggressive consumer behavior going toward human agents. People are having shorter. Tempers having quicker expectations for resolution and that's being reflected in, for lack of better words, a bunch of Karen's calling into contact centers. So I wanted to talk a little bit about that, how it, and we anticipate that as we approach the holiday season, that's going to get worse and worse.
DamienWhy it happens and more importantly, what you're hearing from some of our clients and just leaders of contact centers are around the world with regards to how automation is possibly offsetting that by providing what you termed in a recent total retail article, AI armor. Which is automation as a protective layer against that sort of abuse hitting human agents in the contact center. So first and foremost, happy almost Halloween. Do you have any scary stories that you've heard or heard from from some of our clients or customers, or even something that you've experienced recently that would call out to that theme?
NikolaYeah, different things are scary to different groups here, right? But one thing that really got to me when we started visiting contact centers, because, we used to work on voice assistance more in the consumer context, right? Where, a double digit percentage of calls interactions with Google assistants, Siri, Alexa involve profanities, right? Whether it's because people are frustrated or just because people are. Fooling around. That's a high number, right? But what really surprised me when I went to just shout out human agents talking to contact center agent to call.
NikolaJust contact center interactions, right. Where people were like taking calls and dialing in because of a problem with a device they bought or a delay of a shipment or something like that. I was shocked just by how rude. Maybe not the average person, but maybe a quarter of them were
Damienreally
Nikolalike flat out. Just they're pissed off. They're calling because there's been an issue with a service or with a product or with delivery. But how like the contact center is just like that extension of the company that's there to take that hit immediately. Like I was shocked with the percentage. I could not do the job. Right. Like the reality of being contact center agent is that people are yelling at you, that they're complaining that the moment they can't do something with you, they're like going on aggressive rants.
NikolaAnd, some people do it very deliberately, right? Cause they know that you might give them a refund or you might. Right.
DamienThe squeaky wheel kind of approach. You
Nikolamight give them a credit to make sure that they don't churn. It might be encoded in your rules, right? If you're a utility or a telco, right? Okay. And then, from that point onwards, I was like, whoa, this is. When people talk about automating a job, I think that many a contact center agent would love to not have to have that interaction. Right. Right. But the real issue is you don't know which ones they are. Right. So it's not like you can selectively say, Hey, Damien is giving off a pretty negative energy today. Maybe we leave him in an AI conversation for longer. Right.
NikolaBut I think many would do that if they could, right now, we can't do that just but what we can do is just. Create a layer around the contact center to, absorb the first hit, right, hopefully deal with as many calls automatically as possible. And, season matters, right? So like the high retail season from black Friday onwards, right. It gets hairy because the number of calls grows sometimes by a factor of 10. And then companies staff up, there are a lot of part time people, temporary workers there.
DamienAnd those folks aren't coming in hell bent on protecting brand. They're not coming in on, on being entrenched in corporate culture or whatever they're coming into sort of absorb the blows and to serve as that front line. And that's really not fair in an already grueling job.
NikolaYeah. They haven't had the time to learn, right? Right. Like a nutritionist is already hyped. Like when you triple the size of your contact center with temporary staff, your quality of service inevitably degrades and then your callers get more aggressive. Cause maybe they're calling for the second time, third,
Damienright. Or they've been on hold for quite some time. It's interesting though, when you think about the history of just the utility of a contact center, I probably. In its advent, it was okay. Yes, we want to provide customer service, but then as it became the predominant channel for customer service and volume began to outstretch capabilities and bandwidth, right, then it becomes a deflection agent. It becomes, how can we get people off of the phone regardless of whether or not we provide any resolution to their issue, how do we make things as complicated for them so that they end up hanging up the phone and that becomes like almost a pursuit and it seems counterintuitive, but it's We're past that era now.
DamienAnd the good news is that we have the capabilities to offset that increasing volume while still dealing with the fixed staff size. Is that fair?
NikolaLook, we have more capabilities. I think that when it comes to user behavior, right, like you said it, right. And it's been, I think, happening for two decades now. We're just like, companies are competing by providing a better service, right? So they're trying to get to, and people expect a high level of service. And they are given. Leave to behave in any way they want, right? Because companies are fighting for their business. Which is a shame. We need to be better people as consumers. Absolutely. But, if you were applying for a mortgage and then you start being unpleasant, I'm not going to hang up on you.
NikolaRight. If I did, you would probably go to a different bank. Right. Especially in businesses where the service is somewhat commoditized. Like you don't really have an app think of like utilities telcos, especially right? Like they will compete on service and they will compete to have you state right? So I think that the one thing we can do is create like an instant pickup deal with a lot of issues automatically in a uniform way and then hand off to a human who is ready to take the call and equipped, right. And with enough preparation, context, et cetera, if you increase first call resolution, people are happier.
NikolaYou're basically like, you're just lowering the blood pressure of the whole contact center by giving time back, by giving the right people enough time to deal with those complex queries and by not leaving people to wait for too long. What's interesting I moved about two months ago. I picked the A broadband provider that promised to get me the earliest installation, right? And then they, just magically changed the date to two weeks later. And obviously I wasn't particularly happy and knowing what I know about the contact centers, I called and called and called and very quickly it became apparent that the answer I was getting, which was that the.
NikolaEngineer needs to check the outside of the house and then something we'll have. Basically, like I called enough times to figure out the journey and it turned out that there was basically like giving me a generic answer that would get me to stop calling and that their process were so
Damienthey're still stuck in the deflection era.
NikolaThey're absolutely deflecting. And then I was like, I usually, especially having shot so many contact center agents. I don't like to be that color. But I, I was like, Hey, okay. If you can't access that change in the CRM to write that thing down for the team visiting. Can I please speak to your manager? Right. Can I speak? And basically I'm asking for a tier two agent, the greater level of access, a
Damienbasic escalation request.
NikolaYeah. I literally, it's the only time in my life that I've done it. And I'm cool like that.
DamienOh, really?
NikolaYeah. You, you do it often.
DamienI don't really do it often because I think I explore every other venue before I tend to pick up and make a phone call and it's just my way. Yeah. There's some things that I have to, especially those. Either high margin, things like the worst and necessary customer experience that I had in the past was calling in about a home insurance. And so it was something that I couldn't handle online. I had to talk with an agent and it was a very similar experience to that. So I got pretty frustrated, but I didn't mean to interrupt continue.
NikolaBut what happened there was, I was placed on hold for 30 minutes, went back to the same person. I was like, Hey they're still on a phone call. So that might be one mighty phone call. I hope they're right. And then I realized, and I've since spoken to a few of our clients, just, kind of like, protect the conversation. What really happened there? What do you think? Right. And from people I've, what I've heard is It's deliberately designed, like you're not going to go through, right. Cause they have a playbook of like when and what and how, right. And then if you ask for that they're just going to leave you on hold.
NikolaThey'll make sure that you're just like fizzle out. Right. And yeah, you're spending someone's time, if they escalated me to an engineer, that would be a lot more expensive, right. And they probably have a very complicated equation of my lifetime value as a customer and what the probability that I churn is relative to not meeting that request then. There I was. They, so it's interesting. I think that, companies already have policies in place to somewhat protect both the people and just to manage the whole thing and manage the cost of the whole thing.
NikolaBut yeah I think in the, in the call we played initially, you hear different people, right. And I think that there's like aggression towards automation, but there's also just like aggression with humans. And I think what is really interesting about AI conversations. If people know that it's AI and they tend to know even when it's really human is that they don't try to manipulate AI as much, right? Whereas I might be giving you a sob story of why I really need internet right now, right? How I come home and, my kids at home because she has a cold or something, whatever, right?
NikolaYeah, it's okay, I do not care. Yeah, I can actually listen and can change the stone, right? We have systems where our agents change stones when they're dealing with a bereavement situation, a loss of a loved one or something like that, we can do a really good job there, but. That's just the surface thing. And as a colleague of mine likes to say like customer service is often about saying no several times in very pleasant ways and explaining why the answer is no, right. You can't always meet that request.
NikolaAnd when you can decipher which of your calls are those kinds of like complaints and default in that quadrant of calls that don't do much for the caller. Or much for your business. They're neither sales calls, nor are they like essential support calls where you can do something about it. Right. That's a good place to put it. Right. Because. To be really honest, those calls should not exist, right? They're not true, but that's where that, and
Damienthat's where that AI armor comes into play, whereas it's a protective layer, because honestly, if 90 percent of the calls, or let's say 80 percent of the calls that come in, are these rote routine, like intermediary, you can't apply a certain touch that will preserve that person's brand loyalty or anything like that.
NikolaAnd, just to give you an example, right? If you call and call and call. And an agent tells you something, right. And you keep pressing, pressing, pressing, and they're really trying to, explain something to you. Maybe they don't have the right answer. Maybe they express themselves clumsily, but then you press on, like a lawyer and you're like, but you said that, and you said that, and you said that maybe they like offered something that they have to go back on and the calls are recorded. So if people want to press it, they can't, right. This is where AI is again. Really principle, right?
NikolaIt's like a good lawyer. It will not say the wrong thing, right. If programmed the right way. It's similar to I think we, we had an article a while back about security and what voice assistants can like, it's harder to manipulate them because they can have very hard rules around getting tricked, right. And when to make an exception in a special situation. Similarly here, right. Even if someone's adamant to really just press on and on and on A, with AI, they're a bit less likely to, and B, they're not really gonna catch you off guard in the same way, because The thing is programmed to be very explicit in how it phrases certain things.
NikolaSure. Yeah. And again, like you have data analytics from all that. What people are calling, you can see where they're pressing and, it's it's just a good idea, which would be
Damienharder to get from a human agent who got rattled. By, the umpteenth upset customer that calls up because they had to wait 45 minutes. And the sort of post call analytics are not probably as clean if you're dealing with assuming there's a manual process in place and there's not things like, calling. And humans
Nikolaget rattled and then they quit. And then your service gets even worse cause you don't have enough people. Right. And it's just slippery slope to oblivion all the way until the end of the Christmas season. Right.
DamienSo we've, and speaking of the retail, I want to touch on that in just a sec, but it harkens back to the sense of appreciation that I've heard from a few of our particular customers, but I think who Are always keen on systems or technologies or processes that offer that layer protection and not just because dealing with churn in a high churn environment is something that's taxing on resources in general, but because people managers are still people managers and they want to see their people. Happy and succeed. And I think back to an earlier episode of the podcast in our conversation series, where we talked to Brian Jebson over at Fertitta entertainment.
DamienAnd he mentioned about how, some of the employees were so thrilled because they were able to dedicate more of their call time to. Bonus generating activity. So affecting their income as opposed to handling those routines that they nominated the AI agent that was there, the, the poly AI voice assistant that was in house as an employee of the month, because it was making their job a lot easier by handling all of those, we've had
Nikolaeverything we've had cakes made with like anthropomorphic versions of the agent put on top when, yeah, yeah, absolutely. Like we've We've had thank you notes. We've had like employee of the month awards for the agent. People love it. Right. Cause it's just it lowers the blood pressure, but it allows you to do more. Right. And it gives you a lot, it gives the agents a lot back and makes their lives better. And yeah, and especially it can also detect and quantify where customers are getting really angry. Right. And that's something that feeds back into the ops, right. And it says Hey, look, there seems to be a spike in.
NikolaRedelivery notices not being handed, right? Stuff like that, where if you just say, Hey, we're aware of the problem, right? Then you can really get to the point that you are making the business better as a result, right? So over time again, you're further lower the blood pressure and there's no overnight fix here, lower waiting times more like concrete answers that are always. Praising things the right way. It just helps you get to a better outcome. Right. And especially in these high season kind of settings, it's a really good way to not have your stuff completely burned out and quit.
DamienVery true. So yet another stressor that we can help to remediate, right? I think that people in general can help to remediate by just a going into these peak seasons, like the holidays and whatnot, and just be a little more jolly, be a little more holly jolly when you're talking to any person that you interact with commercial or otherwise, but we can't ignore the fact that. Like the holiday season is stressful overall. So I'm sure that these things are amplified as a matter of fact, one of those studies that was quoted in the article that you submitted to total retail came from bring over and they reported that 90 percent of customer service staff reported an increase in abusive behavior, especially around this particular time of year.
DamienRegarding the positive or previous year. So a 90 percent of a customer support staff report an increase in abusive behavior year over year. Why are we such animals when it comes to this stuff? And I really want to express a little bit more or dive in a little bit more about how you would think that if I had an automated agent on the phone that I would probably unleash a little bit more, but people are less aggressive. To automation. Why is that?
NikolaThe caveat is that the first term like off people screaming their heads off is much worse with automation, right? People will try to get out of it any way they can. That's a big reason for why. At PolyEv, we spend tens of millions on better voices, on lower latency, on higher quality of understanding so we can precisely relate to what you said rather than reading you a sentence, right? Just earning that right. The trust to stay in the conversation. So yeah, like the first thing, if it was a tennis match, you'd probably have, I don't know, it was the best on the tour I don't know, pick a very told and I don't know, it was Roddick really good.
NikolaI don't know. My tennis knowledge fails me, like the world's best servers on the other side, right? So like the thing starts. And it's quite difficult. Right. The caveat is that, those very tall tennis players that have an impossible serve, Ivo Karlovic comes to mind and a few others, right? If you return, they're not as fast, right. And they're a bit clumsier. They're used to just winning the points when they serve. Right. So if you stay in the points, you probably have a higher percentage chance of winning it. And the same is true for for AI interactions in the contact center, where if you are able to like.
NikolaGo back and forth a few times, then the aggression really drops people. Oh, this works. Okay. Then they get surprised. Like they either like even compliment directly. Right. But they figure out that they can use more than a yes, no, like refund. Right. And they start talking, they see that the system really picks up. Then it becomes a magical moment because everyone's hearing about AI all the time. And frankly, I don't think we're seeing enough of it just yet. Not in daily lives, right? Right. Sure. Like I wonder what percentage of the global population is aware of just You know, the recent jump in performance of GPT I really wonder how many people are aware of it and can actually figure out that Oh, one a is like what it is.
NikolaSo I think when people like are on the phone and they realize that it's AI, then they start asking questions. They go on tangents, they have a bit of fun with it, but also like they're bought back in. And then the anger is just. They're in a completely different mental state. Right. Cause now I'm talking with like better AI than I've ever spoken to before. Right. And that's,
Damienand that can be a cute, like, why do they come? Yeah, it must be, and it probably is. And there's that level of amazement too, that probably tempers that anger a little bit quicker. Absolutely. You get
Nikoladistracted away from being angry.
DamienYeah. That's a really interesting point. If you're hyper
Nikolaballistic, right. You're probably going to. Ask for that handoff anyway. Now to all prospects and existing clients, right? If you actually want to not take those calls, like we can detect a much higher level of frustration coupled with semantic profanities, you could have those calls. Go into a keyword, they wait for an hour, right? You could very well do that, right?
DamienThat's why all the calls that we played at the intro, which is just a lovely collection, by the way. You hear that most of them, when they hear the level of profanity that's there, say, Okay, let me get you to somebody else. Because like you said no, no,
Nikolathose calls are actually like, all but the last one. Which for the ballistic, we do exactly. What we're supposed to, right? The system works regardless of profanities, right? And that's like part of that armor, right? Like you just didn't have to have people doing it. Now, if you're going to be serious about profanities and a different elevated emotional state, getting you either different agents, maybe more skilled ones, right? Or if you're going to maybe just stack rank people by their overall pleasantness, there's no reason you couldn't just. Push these people to the end of the queue, right?
NikolaNo, for most businesses, someone very angry is probably a higher turn risk. So they probably want to put them to the front. I don't know. Not many companies actually action this in ways that we have signal for. That's something that we're excited to explore in the future.
DamienOh, that seems pretty cool. All right. So in general, like the fact that more turns equals more tempered anger, which is nice to know people are probably coming in hot, especially if they've been on hold and they have an issue. So in this, that initial blast is probably worse than the longer the call lasts here to your tennis analogy of that, those aces that come across. If you can return one, then you got a higher chance of winning the point. So try and keep people on the phone. One higher chance of resolution. It calms them down just a bit. And then three obviously people are largely unaware of their true capabilities of something in this level of application for modern conversational AI and generative AI, like it's only going to get better from here.
DamienIt's only going to get more capable from here. So we should probably. Get used to finding another outlet for preventing frustration as opposed to the poor human service agents that we've unleashed on in the past, because their jobs are about to get a lot better. You should
Nikolaabsolutely punish companies that provide you a bad service or a bad product by not returning your business to them in the future.
DamienAnd what is it something like 65 percent of people will abandon a company completely if they have one bad experience? That's it's wild to see the level. Yeah.
NikolaAnd that's why those companies and those contact center agents do their very best to. Tolerate you even when you're being absolute scum, but remember the agent is not at fault. They're just, they're trying to help you. Right. You're probably going to get further by being nice. Right. And I think a lot of people are just in either. They can't calm themselves down and re engage basic humanity, right. Or they think that they will get a better outcome, a better refund, a higher incentive to stay with the contract or something like that.
NikolaAnd to those people, well. Do you have nothing better to do?
DamienDo you have nothing better to do? And for companies out there, don't enable that sort of behavior. It doesn't help anybody in the long run. The squeaky wheel should not always get the grease.
NikolaYeah,
Damienno, good stuff. Nikola. Always a pleasure to talk to you. Happy Halloween as we head into that season. And I'm sure we'll do another retail focused topic as we approach that peak holiday season. We have
Nikolaa dashboard and our retail clients Tend to really, really take over and call volume this time of year. Yeah. Oh,
DamienWe should pull out some fun little interactions from that and see what kind of trends we can detect in in what our bodice, what our bots are picking up within the industry. I always like to find out how many people are thanking robots more and more. It's just, to me, that's the big tell. Cool. Appreciate your time, Nikola. We'll talk with we'll talk with you soon and hope to see you all next week. Have a good one.
Three years ago, Simplyhealth's PolyAI deployment was only resolving simple FAQs over the phone. Today, they run voice, in-app, email, chat, and WhatsApp conversations on the platform. Nikola Mrkšić talks with Paul...
An agent that gets better every day compounds into a very different agent within months. That's the idea behind Wren, PolyAI's new release: an agent whose job is to make your other agents better. Some of what it finds...
Companies are pouring money into AI, but not everyone is getting the returns. Ian Jacobs, VP and Lead Analyst at Opus Research, thinks he knows why. In this episode, your host Nikola Mrkšić sits down with Ian to work...
When PolyAI built its Builder MCP, one client helped shape it: Vixxo Facility Solutions, one of the largest facilities providers in the US. In this episode, guest host Damian Sasso talks with Vixxo's CTO Derek Neighbors...
PolyAI just launched Dialog-RSN-1, its first audio-native model, and your host Nikola Mrkšić sat down with its builder, Matt Henderson, to unpack why it’s a game-changer for building voice agents. Most voice AI either...