In this episode of
Customer Service Secrets, Vikas Bhambri, Senior Vice President of Sales and CX at Kustomer, joins Gabe Larsen in discussing how both human customer service agents and artificial intelligence (AI) are mutually beneficial in the development of real and positive customer experiences.

Artificial Intelligence Customer Service: Bots Cannot Solve Problems Alone

The artificial intelligence bots of today’s world are not only growing in popularity, but they are also growing in capability. They are the focus of various customer service conferences around the globe; but are they being utilized in the correct way? Vikas Bhambri, with his 20 years of experience in customer service, claims that even though they are receiving growing amounts of attention, people do not seem to understand where AI bots show their strengths.

Bhambri discusses how everyone is “hyper-fixated… it’s getting kind of buzzwordy… [but] the key to me is, let’s think about the customer. Let’s start with the customer and the experience that they want, whatever you want to offer them, and then let’s figure out where you appropriately position the bot versus the human being.” Only in the future, when innovations permit even further data for both bots and humans, can they coexist in beneficial harmony.

Knowing Your Customer

Human reps and bots will more successfully coexist when the bots are able to recall previous data from individual customers. Doing so will enable customer service branches to personally help clients, rather than run everyone that calls for assistance through the same AI loop. Vikas goes on about the necessity of “AI machine learning… [bots should be] looking at the results of anybody who’s ever asked a similar question and what has been offered to them and what actually resolved their issue… that’s where it gets… more in-depth.”

He also gives an example of treating customers differently by comparing clients that have different demographics. Your company will have “all [of] these different issues that have been resolved across [your] entire customer base and now a multimillion-dollar customer comes to [your] website and asks a question.” Your bots of today are “probably going to offer them the same solution as [it] did to the last 20 people that asked that question… they’re only looking at the question, and they’re not looking at who [they] are.”

Where the Bot and Agent Best Work Together

Once the customer has been personally identified, it is vital that both the bots and reps focus on the customer’s needs. Not only is the client’s problem important, but the means, or channel, that your company uses to resolve it should be an additional focus. It is essential to understand the customer’s situation and realize whether a personal interaction with a rep or an automated conversation would be more efficient.

Vikas and Gabe talk about the idea of whether the customer wants “to speak to somebody [or] if [they] don’t, and want to do it [themselves].” Vikas gives the potential example of using new tech advances like geolocation to aid in the customer experience as well. He says, for example, “I’m an airline, and when you’re reaching out to me from an airport the moment I kind of initiate that, [I] should be like, ‘Oh Mr. Barry, I see you’re booked on the flight from Orlando to New York because — and I know you’re in the airport right now. You know what, we’ve already rebooked you. Just head over to gate 43.’” The future holds great potential for the merger of AI and human reps, but the customer experience can only really be elevated when the customer’s needs and situations are understood by both.

Do you want to enjoy more in-depth ideas about how to better the customer experience? Listen to the Customer Service Secrets episode “Bots Vs Human: How to be Successful in AI Customer Experience” to hear it directly from the experts.

Listen Now:

Full Episode Transcript:

How to Combine the Best of Both Human and Artificial Intelligence to Kindle a Successful Customer Experience

Intro Voice: (00:04)
You’re listening to the Customer Service Secrets podcast by Kustomer.

Gabe Larsen: (00:11)
Hi, welcome everybody to today’s show. Today we’re going to be talking about bots versus humans, all things customer experiences. We’ve brought on Vikas Bhambri where he currently is the SVP of sales and customer experience over here at Kustomer. Vikas thanks for joining man, how are you?

Vikas Bhambri: (00:26)
Glad to be here man. My partner in crime. Guest number what, 56 on the podcast?

Gabe Larsen: (00:32)
No, when this comes out man, this is, you’re going to be earlier than that.

Vikas Bhambri: (00:37)
I asked to be number one. I would like to quickly dismiss it… Now it’s like 56, 57, somewhere along those lines.

Gabe Larsen: (00:45)
If you could see me right now, my face is red. I did tell him that but I’m not going to fulfill that promise. Well you’ve been on vacation for like a whole four days. So what do you expect me to do, wait?

Vikas Bhambri: (00:55)
I’m glad the place is still intact, you know?

Gabe Larsen: (00:59)
So I probably didn’t do justice introducing you. Tell us just a little more about your background, some things you do over here at Kustomer, etc.

Vikas Bhambri: (01:04)
Sure, I’ll give you the short version. You can tell me if it’s not short enough or if you want me to go into more detail. Twenty years, CRM, contact center veteran. A lot of people don’t actually know this about me, but I started my journey, or my career in the contact center. I was a guy who carried a pager around, got the call — the page at two in the morning that something was wrong with my application. So back in the day, you have to actually be the coder and the QA, and the help desk for your product. So I did that, but then found myself actually implementing contact center technology. My first client was CSX transportation. If you don’t know them, they’re a big commercial railway on the East coast. And I actually implemented the contact center solution where if you were at a railroad crossing and the crossing was down or broken or the gate was smashed, you call the 1-800 number, it would route into the platform that I implemented with the agent.

Gabe Larsen: (02:10)
What was this 1970, 1960?

Vikas Bhambri: (02:15)
I’m not that old. It was probably just around the .com, so 2000, 2001. I went from there to implementing contact centers, like Bank of America, UPS. So I’ve been on this like CRM contact center journey since its inception.

Gabe Larsen: (02:31)
Was that on purpose or was that just by accident? I mean, are you that passionate about the space?

Vikas Bhambri: (02:36)
You know what, I’ve become passionate about it. I mean, you know, initially it was a job. Oh this is interesting. And you know, for me it was more around I love technology. So it was the perfect role to be a business analyst or project manager working with technology. But then as I got into it more and more and spent more time in the contact center, in the trenches, and then in the CRM world, which now encompasses sales, marketing, etc. And just seeing that evolution. So it’s been fun. I’ve worked across the globe, I spent five years in Europe. I’ve done CRM sales service marketing, you name the industry: retail, TELCO, financial services, insurance, healthcare… so it’s really been a great run over 20 years.

Gabe Larsen: (03:23)
I love it, man, that’s right. It’s funny, you and I have known each other for a few months now, but I forget that history, that’s a pretty rich history.

Vikas Bhambri: (03:31)
Yeah, a lot of people, they get caught up in the title, the most recent title, right. Like, oh you’re sales and CX leader and reality is, I actually started my career as a developer. It’s been a wild ride.

Gabe Larsen: (03:42)
Yeah, that’s a little bit of a change, right? Board room to dev room. So let’s dive in: Talk bots and humans for a minute. So obviously it’s a little controversial, isn’t it?

Vikas Bhambri: (03:55)
It is, because I think, of late, everybody is hyper fixated. You go to any conference, you go to any meeting and everybody wants to talk about bots. It’s getting kind of buzzwordy right? And everybody now says they do it. Everybody says they’ve got one. The key to me is, let’s think about the customer. Let’s start with the customer and the experience that they want, whatever you want to offer them, and then let’s figure out where you appropriately position the bot versus the human being. And I think ideally, and I think that the future will actually be where, they coexist. And so we can stop having this…

Gabe Larsen: (04:37)
One eliminates the other, one pushes the other out.

Vikas Bhambri: (04:42)
Right? And even the way some people talk about bots is they’re like, “look, we’re going to — we’re going to implement the bot and they’re going to solve the problem.” And then what happens when they don’t? Now the customer’s frustrated, right? Now, they pick up the phone or they called the agent and the agent has no idea that they just went through an eight step process with a bot and it failed. So even understanding like how do I take a journey that may start out with a bot, and actually escalate it to a human experience.

Vikas Bhambri: (05:11)
And what nobody talks about is, when did it start out in a human experience and then maybe kind of escalate to a bot, right? So you actually empower the human agent with more information, more data, more automation for them to give intelligent solutions back.

Gabe Larsen: (05:27)
Let’s go back to it. So maybe take one step back real quick because I want to dive into those two use cases. But when you say bot, how is that different than chat versus AI versus… give us a little click on that.

Vikas Bhambri: (05:43)
Sure. I think for me, when you think about bot, I kind of liken it to just robots, right? It’s technology that does a task. Now when you get in a chat box that’s just serving technology through a medium, which happens to be chat. But I would argue chatbots are already outdated because chat is only one digital interaction. Why wouldn’t you do the same on Facebook messenger? Why wouldn’t you be the same on WhatsApp or SMS? So even the whole nomenclature now it’s already outdated.

Gabe Larsen: (06:13)
Well and it did feel like chat — chat’s been around for so long. It’s like wow, is this really something that new, adding a little more of a bot or a push notification in a bot? But it seems like maybe as we take it to different channels, that would be one thing that would certainly be different. It’s this automated interaction in a channel, chat particularly, that allows you to potentially deflect or get rid of some of the human interactions.

Vikas Bhambri: (06:39)
Chat was rightfully kind of the first kind of place to offer it. Because at the end of the day, people are already used to doing pre-chat surveys and asking certain questions. So it kind of made sense to offer it there, right? And you know, you’ve got companies like Drift and others that are doing it in different styles. So it makes sense. But why wouldn’t you offer some sort of automation when somebody goes to your knowledge base? So now we call that — now we’ve kind of pocketed that into self-service deflection. At the end of the day, it’s still a bot. It’s still technology that is looking to the customer to answer certain questions or make some self identifiers and then offer them a solution.

Gabe Larsen: (07:21)
Got it. Got it. Okay, perfect. That’s great to just get the fundamentals. And then one step above that, where do you feel like it’s, maybe it’s where we are currently or where we should be going, but there’s stuff that’s like pre-programmed, like branching stuff you could put in. So they like press a button or they answer yes and then it delivers them a message versus true intelligence, like they write something, the bot reads it and actually responds back in an intelligent way. Are both of those happening? Is just one of those happening? Where are we in this evolution of the bot, so to say?

Vikas Bhambri: (07:57)
Sure. So, to me it’s kind of the if, then, else, right? Like the choose your own adventure. For those of us that are old enough to remember those.

Gabe Larsen: (08:04)
Those books were good. I should get one of those for my eight year old, actually.

Vikas Bhambri: (08:13)
But here’s the thing: So the if, then, else, the branchable logic that’s there. It’s been done. I think you see that quite often now.

Gabe Larsen: (08:22)
That’s pretty table stakes now.

Vikas Bhambri: (08:24)
Right? That’s table stakes. I think true AI, where you’re looking at the question the person’s asking, analyzing it, then comparing it to other questions… When we talk about true AI, machine learning, it’s now looking at the result set of anybody who’s ever asked a similar question and what has been offered to them and what actually resolved their issue. So that’s where it gets a whole much more in-depth. Now, I still think the problem with even that concept and why I’m excited about some of the things we’re working on, is that it’s still very limited to all the problems that people have asked and answered. It still doesn’t really take into account who that customer is. I think that’s still one of the things when we talk about bots and you’re only as good as your data. So what I describe to people is… look, imagine you bought a robot to clean your house and you only put it in one room of your house and said learn and then you unleashed it on the whole house. You’d probably end up with a wreck because the dimensions of your one room are not all of the rooms. And I think that’s when people create these algorithms, they’re only thinking about one problem area and then all of a sudden they unleash it. For example, I’ve got all these different issues that have been resolved across my entire customer base and now a multi-million dollar customer comes to my website and asks a question. I’m probably going to offer them the same solution as I did to the last 20 people that asked that question. Now taking into account that they’re a $5 million customers, now I’m going to wreck my house.

Gabe Larsen: (10:00)
Oh, interesting. Almost like a tiered… you know, we talk about like tiered support where if I’m a gold member, I call in and I’m treated different. But you’re not really treated different with a bot because they don’t know a lot about you, and they’re only looking at the questions.

Vikas Bhambri: (10:14)
They’re only looking at the question, and they’re not looking at who are you.

Gabe Larsen: (10:16)
So that might be one of the future trends, as you think about bots and how they… I can’t think of anybody doing that, that’s pretty… wow, that’s different.

Vikas Bhambri: (10:27)
That’s it. The more data you can feed this, the more intelligent it’s going to be. I think the problem is when people are thinking about it, they’re not thinking about what data am I going to keep using with the robot? Because that’s easy for people to say, “what information you’ve giving the robot?” And if you’re not giving them all the details, they’re going to make foolish decisions.

Gabe Larsen: (10:47)
What else do you have? If you had to kind of say a couple of years from now, I mean I just thought that was interesting. Kind of the personalization of the bot around the individual, the company, whoever it may be, and then treat them slightly different. Any other things you see in a couple of years from now, where the bot is going to that might be a little outside of the norm?

Vikas Bhambri: (11:06)
I think the big thing — well, before we even get there is I think there’s going to have to be this harmony between the bot and the human experience, which I don’t think exists today.

Gabe Larsen: (11:17)
So lets click into that, and then we’ll come back to the trends. So right now people are kind of thinking about it: as I interact with the bot and then there is a chance I would maybe escalate to human.

Vikas Bhambri: (11:31)
It’s still clunky. The hand off is clunky because a lot of times, well we’ve all experienced that as consumers. I get asked a bunch of questions by the bot, I get served up to human agent because the bot can’t actually answer my issue. And the agent actually asks me all the questions again. That’s like the most fundamental failure of the hand off because they have no visibility. They may know that you did communicate. A lot of brands won’t give their bots names. So like you, you asked Jeeves or you asked Elsa, right?

Gabe Larsen: (12:06)
Elsa is a Frozen reference in case anybody’s wondering.

Vikas Bhambri: (12:11)
Right, anybody whose kids are listening, they got it. All of a sudden, they talked to Elsa before me, but you don’t know what they asked and answered. So that’s a very fundamental flaw, but people are getting better at that. They’ll at least give you the tree, showing everything that the person went through with the bot, right?

Gabe Larsen: (12:27)
Do they? I sometimes question if they’re even getting that, but fair. Yeah, they could get that far.

Vikas Bhambri: (12:34)
There are some people who are a bit further along, if you look on a maturity index, so now I know what questions you asked the bot or answers you gave, and why they can’t resolve it. But to a degree, I almost have to still go and do my own due diligence and figure things out. So I think that’s step one. Now the other thing is, how do I take that data that I did get, plus what the agent captures and now offer up intelligent suggestions to the agent to resolve? That’s where I think you start getting true harmony is automation on the front end, smooth pass off, but then also helping the agent be smart by giving them smarter answers.

Gabe Larsen: (13:16)
But help me visualize that a little bit. So what would that look like? The first part I get, so you get the automation. I like the second part, the smooth transition, because that just feels clunky in my own experience with bots. But that third part. It’s like, ooh, how can we enable the effectiveness of the agents so they are responding back smarter? Any examples, like tactical examples that may come to mind?

Vikas Bhambri: (13:38)
Think about this and let’s just use your cable box provider. You just went through an eight step troubleshooting process with the bot. It failed. You’re on the phone with the human agent and the technician is saying, “Ah, okay I see that you went through this process.” Maybe the agent gathers one or two more details from you. You know what I mean? You check the remote, you know the batteries in your remote or whatever. Now, the intelligence to the agent says I’m going to take all the steps that the customer did with the bot, plus what you gathered and now I’m going to offer up a solution. Take both sides of that discussion and then offer up a solution.

Gabe Larsen: (14:20)
Cool! Interesting. So now flip it, because that’s kind of the standard idea, that can we deflect –? Well, do one more quick double clickback on that. So I liked your three step process. You have automation. If you need to escalate, you pass it off smoothly and then you kind of provide real intelligence or a recommendation. A lot of people are wondering how far you can go with a bot before you have to escalate. That’s, I guess, the elimination conversation. Where do you kind of recommend companies who aren’t thinking about that? Try to get rid of the small stuff, focus on the return? How far can you automate that bottom part of customer service?

Vikas Bhambri: (14:58)
I think the two factors you have to look at are one, what can the customer do themselves or can you kind of use the bot to guide them through to conclusion? So that to me is number one, because at the end of the day, as much as brands don’t want to talk to customers, customers don’t want to talk to brands either. Right?

Gabe Larsen: (15:20)
Do you think that’s true? I mean is that kind of where we are? I mean people don’t want to really do it.

Vikas Bhambri: (15:25)
They don’t want to. It’s not a bad thing. And I think we need to get away from that. If I’m booking a round trip flight from New York to LA, I don’t want to talk to anybody, I want to go to a website, I want to go to a mobile app, I want to book the ticket, get a reasonable fare, select my seat and I’m done. It’s paid for it and everything, right? I don’t want to ever speak to a human being and the airline doesn’t want to speak to you either.

Gabe Larsen: (15:53)
It just sounds bad.

Vikas Bhambri: (15:54)
But quote unquote, we’re talking, right? Because we’re obviously transacting business, but we don’t have to have an elongated discussion. Right? So that’s number one. Number two is when do you want to get that human being involved? Because now I’m booking New York to San Francisco, to LA, to Portland.

Vikas Bhambri: (16:19)
It gets more complicated. I want to be able to speak to somebody if I don’t want to do it myself. Number two is there’s an adverse event, right? My flight to San Francisco gets canceled. Now everything is going to be botched. I want somebody to jump in. And third, you have customers whether it’s ato demographic, whether it’s a high end customer, that you want to offer, that additional level of service, if they choose to use it. And that’s why I said bots can be a one size fits all because if you’ve got a premiere business traveler, you want to be able to say, look, if you want to go and book that round trip ticket yourself, go for it. But by the way, we’re here for you.

Gabe Larsen: (17:00)
And maybe it’s just where I am. Don’t get me wrong, I like to book stuff on my app and things like that, but I’m in like a Delta premiere or whatever that is, gold or medallion, and maybe I’m driving, you know, I’m just gonna call them up and have them walked me through it and book my roundtrip ticket or something. I like that sometimes, so I do like that option. I like the complication. The emergency totally resonates, right? It’s like when your flights booked and you’re trying to go home for Christmas, the last thing you want to deal with is a bot. Is my flight canceled? Just help me, I need to talk to someone.

Vikas Bhambri: (17:35)
So that’s why I think brands need to look at where is that inflection? Where does that point where the customer is going to yell? Now there might be some customers, they don’t care if they’re yelling all day long, right? But certain segments of customers, we care, the brand absolutely does care because they want your repeat business. Right?

Gabe Larsen: (17:54)
So do you feel like if you implement some sort of automated bot program, are you affecting negatively or impacting negatively the customer experience?

Vikas Bhambri: (18:02)
No. If it’s done thoughtfully where you’re thinking about that journey and go back to the customer journey, or customer map. It may actually benefit the customer. If I want to change my address, do I want to speak to somebody by changing my address? No I want to go in, I want to punch it in and I want to hit submit and let it go.

Gabe Larsen: (18:25)
I think the problem people are running into is because it’s such a trendy word now, I think I’ve run into this in the past a little bit is you’re like, well, let’s throw a bot on our website or let’s throw a bot somewhere and you don’t watch the rest of that customer journey and that’s where you drop off on kind of points two and three. We have a bot, but the experience actually got worse because we didn’t help them.

Vikas Bhambri: (18:42)
I think like anything, A, you need to AB test, and B, you need to do the what if scenario. What if a customer wants to do this? What if they do that, and you need to really think it through. But the easy thing to do is just… you almost need like a program management around iit.

Gabe Larsen: (18:59)
You really do. What I learned in my last gig is, we got a bot and it was cool. We threw it up there and pretty soon I was like, I need someone to own this and own the journey. It’s not just a side gig that someone else can do by themselves..

Vikas Bhambri: (19:16)
Like in marketing, right? You have somebody who does your search engine optimization. You need a bot optimizer.

Gabe Larsen: (19:28)
So flip the other way then. Is there a reason or a method to go to a human, then to a bot? Is that in our future, that certainly would be kind of a side scenario or side use case. But is anybody doing that? No.

Vikas Bhambri: (19:47)
No. I don’t think anybody’s doing that. I think right now it’s about bot to human. But where I think is a missed opportunity in the near term is to empower that agent with more choice for automation, where they can do things. And you’re seeing in some industries I think TELCO is actually ahead of this where your agent will take action on your behalf, like so you don’t have to get up and reboot your cable box. There’ll be like, we can do it from our side.

Vikas Bhambri: (20:19)
It’s things that financial service institutions are able to do, where the agent can initiate fraud detection and things like that. So I do think there is things happening on that side, but I’d like to see more of that across the board.

Gabe Larsen: (20:31)
See if we can’t bring that together. Okay, last two questions and I’ll let you get back to your day job. So one is for people who are starting to go down this journey, human versus bot, I think you’ve given them a lot of material. Where would you kind of say, if you’re starting this journey, here’s a couple things I think about or if you start, here’s the baby step you could do now, what’s kind of that easy step that you could take starting the journey of maybe getting a bot into your program and your customer service journey?

Vikas Bhambri: (20:59)
I would start with my knowledge base, your FAQ’s. The reason I think people should be putting FAQ’s or their knowledge base together is they’re like, oh this stuff is so darn easy that I expect my customers can do it themselves. So start there and start putting that into your initial bot journey. Where you’re basically pointing them to existing artifacts, things that exist. And then as you start triaging through those, then it’s like what are the next level of… let’s actually sit down with the agents or the reporting, and look at what are people reaching out to us about. And ideally you want to look at the end of the day, you want to fix the end solution. But if you can’t do that in the near term, what can we do that can automate the solution.

Gabe Larsen: (21:54)
I love that, I love that. That’s a great place to start. Okay. Last question is, we touched on it a little bit before, but there’s a lot of movement in the space. A lot of new technology is coming out, all different languages. Obviously some buzzwords. Any kind of predictions as you move into the future, thinking about humans, bots, anything kind of on your mind that says, I think it’d be fascinating if we saw X or Y in the future as bots evolve and iterate?

Vikas Bhambri: (22:20)
I think the biggest thing is, how much data can we feed? What I mean by that is, look, if I’m on my mobile device and you’ve got so much information, whether we believe it or not, the brand potentially has access to my geolocation. They have access to certain data about me on my phone. They have a profile on me. They understand the question I may be asking. Where to me almost get to the point where you’re doing predictive analysis. Before I even ask you my question, you know the question I’m going to ask because we have so much data about you. So we’re like, wait a minute, most of the time when somebody is reaching out to us, I’m an airline, and when you’re reaching out to me from an airport. The moment I kind of initiate that, you should be like, “Oh Mr. Barry, I see you’re booked on the flight from Orlando to New York because, and I know you’re in the airport right now. You know what we’ve already rebooked you, just head over to gate 43.” That’s the Nirvana.

Gabe Larsen: (23:28)
You know the funny thing is I used to be nervous a little bit about the data thing and giving too much data, but now I’m like, I want to give, and I think there’s people like me in this world who are willing to give up less privacy. They’ll have less privacy to get better service, to get more personalization. I’m like, dude take my social security number and take whatever you want, but give me that type of service.

Vikas Bhambri: (23:51)
I think that’s ultimately it. Like look, GDPR, you’ve seen the California Consumer Privacy Act, all of this stuff, people are still hitting every website. You know, I was in Europe, and every website comes up with a pop up and everybody hits accept. Why? Because I’m giving you data because at the end of the day I’m hoping you’re going to market to me better, sell to me more intelligently or are you going to give me better service. There will always be people that will opt out. Most people think, “if you’re going to offer me more value, I’m willing to give that.” And there is so much you can do with it to benefit the consumer.

Gabe Larsen: (24:31)
Could you be proactive? You’ve heard some of those stories where you know people are the target. Did you hear the target story where they were buying this family was buying different things and then they sent them like a gift card or a coupon for… I won’t get into the details, but basically send them a coupon and they were like, “Hey, we’re not, we’re not actually experiencing that. We’re not doing it.” Well they went and asked their daughter, and it sounded like that person is sick or is not working. But based on the behavior, their AI triggered, and sent a coupon for something, this father got it so it can get a little bit out of hand. But my goodness, the stuff you can do with data, wow. You can take this pretty far.

Gabe Larsen: (25:17)
Cool man. Well, I appreciate it. So if someone wants to get in touch with you, learn a little bit more about what you do, you know, continue the dialogue, what’s the best way to do that?

Vikas Bhambri: (25:26)
LinkedIn is always a great place to hit me up. You can hit me up at kustomer.com as well, either or.

Gabe Larsen: (25:34)
Love it, man. Well, I appreciate you joining. Great times. Audience, have a fantastic day.

Exit Voice: (25:47)
Thank you for listening. Make sure you subscribe to hear more customer service secrets.