How to Solve 40% of Customer Inquiries With Kustomer IQ

Summary

The Kustomer webinar featured Vikas Bhambri, the SVP of Sales and CX at Kustomer, as the moderator. He introduced the co-presenters, Peter Johnson (a.k.a. P.J.), the VP of Product, and Julian Zatta, the AI Customer Success Manager. The main focus was on how to solve 40% of customer inquiries using Kustomer IQ. The discussion delved into how Kustomer IQ can scale businesses with intelligent automation, chatbots, and knowledge-based deflection. Julian demonstrated how their self-service model works, with AI-driven deflection capabilities, proactive suggestions, and advanced sentiment analysis.

Key Takeaways

1. Kustomer IQ Revolutionizing Customer Service: Kustomer IQ, powered by the acquisition of Reply.ai, aims to transform customer service by achieving an impressive 40% deflection rate, resolving nearly half of all initial customer inquiries without the need for human agent involvement.
2. Four Steps in the Customer Service Journey: The customer service journey comprises four key steps: self-service, intelligent automation, routing based on intelligence, and empowering agents. Each step leverages data and AI to improve customer support efficiency and effectiveness.
3. Scalable and Easy Implementation: Kustomer IQ emphasizes easy implementation with drag-and-drop features, making it accessible for businesses to integrate intelligent automation and deflection capabilities into their support processes.
4. Personalized Support through Intelligent Routing: Kustomer IQ’s advanced features, like sentiment detection, automatic language detection, and intent identification, enable businesses to match customers with the right agents, offering personalized support and improving customer satisfaction.
5. Proactive Reach-Out and Future Expansion: Kustomer’s platform roadmap includes plans for proactive customer reach-outs and the expansion of support to multiple channels, such as SMS, in addition to the current offerings like chatbot-based deflection and omnichannel support.

Transcript

Thank you everyone for joining this customer webinar on how to solve forty percent of customer inquiries with Kustomer IQ.
I’m your moderator, Vikas Bhambri, the SVP of Sales and CX here at Kustomer, and I’d like to introduce my co-presenters. Peter Johnson, a k a p j.
Customer’s VP of product. P j say hello.
Hey, everyone. And with us hey, PJ. And with us is Julian Zatta, Kustomer’s AI customer success manager. Joanne, say hello. Hey, everybody.
Let’s get started.
So according to Gartner, seventy-two percent of Kustomer in actions will involve technology such as machine learning and chatbots by twenty twenty two.
Two years from now.
Just take that in for a moment.
At Kustomer, we recently acquired a company called reply.ai to advance our intelligent automation capabilities add them to our KIQ offering.
Together, we’re gonna revolutionize customer service with Kustomer IQ.
Here’s our agenda for this discussion.
How to scale your business with intelligent automation, chat, bots and knowledge based deflection, omnichannel deflection, what KIQ can do for you, and, of course, we wanna make sure we keep time for questions and answers.
So as we’re thinking about this journey, really we think about it in four steps.
Number one is self-service, that opportunity to eliminate conversations that either you, the brand don’t want to put a live human agent behind, or actually what consumers will will most and tell you there are certain inquiries that they have that they’d rather self-serve themselves. So how eliminate those inquiries where you don’t want to have a live human agent engaged.
The second piece of that is there does come the time where automation, a bot, etcetera, doesn’t necessarily resolve the customer’s inquiry. So how do we then take what intelligence we have about that individual whether it be the language they’re speaking to us in, their sentiment, what information we’ve they’ve given us or we know about them. And use that information to route that conversation to the right agent so that then the agent can have a compelling interaction with that individual. And then let’s not leave the agents behind.
We wanna give the agents the tools and the intelligence to also have efficient productive conversations. So those are the three key steps of the journey that are all based on a core foundation around having a platform that scales that allows you to have the data that you need about not just the conversations you’re having with your customers but all your relevant customer data as well. So those are the four steps to the customer service journey.
It’s been amazing to see how our customers have adopted the KIQ suite and are already seeing tremendous results.
With an astounding forty percent average deflection rate KIQ can successfully resolve nearly half of all initial customer communications without the need for live interaction with a service agent.
Let that sink in for a second.
So that’s what we’re gonna be talking about today. And with that, I’d like to pass over the baton to Jillian to walk us through some of this self-service.
Yeah. Sure.
Thank you, Vicus. That was a wonderful tee-up. So, as we mentioned in the beginning, reply AI was just recently acquired by a customer to help power some of this self-service deflection.
And The problem that we’re seeing on the front end is interesting and complicated. So teams are experiencing increased ticket volume.
The self-service model, while there’s great articles and resources and community pages for customers, it creates more work on their end to find and so many of them are bypassing these resources and going on to create more tickets, chats, and messages in your system. So is there a way to shift, that responsibility away from the customer, but still give them the answers they need without accepting more tickets and efforts on your end. And then on the other side, customers in general are demanding more personalized immediate attention.
The world is very customer-centric right now. Social media and mobile have given customers the power to feel as though they can have immediate access to brands and that they can talk with other customers and consumers and friends, and put a lot of pressure on modern companies. And finally, you know, the way that everyone’s been traditionally dealing with this is to scale their customer service teams And, unfortunately, that creates second and third-order problems when you’re dealing with larger teams, more resources and costs. And so The solution, that customer IQ provides on that front end is really designed to do zero-touch resolutions, deflect some of those really to easy to answer questions and get customers the answers they need so they can go on, with the rest of their day.
So how can customer IQ help?
We’re working in two channels right now. One is with chatbots and messaging deflection, to answer customers in a very conversational format to answer those frequently asked questions and also automate those low complexity interactions like Where is my order? Doesn’t require, it now requires the, human interaction, but it’s a low-complexity conversation.
We can automate that now.
And then on the other side, we can do deflection based on your current knowledge base. So we can lift out answers and not just articles to address quest address customer’s questions using their natural language inquiry.
So all of this is really intended to substantially reduce inquiry volume. So I mentioned answers not articles. On the chatbot side, this is a totally conversational experience. It’s in your brand voice. It’s automating the conversations that you’re already having with customers. Other tools out there in the market, are offering a chat-like experience but it’s not the same.
And then if you’re in the knowledge-based deflection product, Instead of just giving your customers the articles that they could have found on their own, the system will actually lift out not just the right article, but the right answer within it. So the right sentence or paragraph that directly addresses the inquiry. So you’re also saving the customer’s time by having to read through a long or lengthy article.
And then as I mentioned, with zero-touch workflows, you can automate those low complex interactions on in both systems. So in the chat, it’s a quick back and forth to figure out where the order is. And then in the knowledge base, what’s been traditionally a static experience where if someone has a question for where is their order, the article would usually tell them wait a day or two, and then go talk to an agent. But that doesn’t actually answer the question. So with the deflection, tool for the knowledge base, we can accept the order number and update customers on where their order is in real time. So both channels, can leverage this more intelligent way to provide self-service.
Now, we have a top international delivery customer, who is using these deflection capabilities right now. And they focused on just one automation to start something really simple. And based on this one automation, they were able to deflect over seventy thousand conversations, and which made them reach an eighty-four percent deflection rate, which is huge. This was a very, very common question. We asked them to focus on what were the most common questions. And so by applying this automation, we were able to seriously reduce their inbound volume.
And I actually love this stat by Gardner, because forty percent of interactions will be handled in the first contact by 2025. So we’re all shifting towards this, deflect and customers are becoming more accustomed to it because they just want their answers.
Now, I’m gonna pass it over to PJ who’s the VP of product who can take you into the rest of the intelligence automation cycle.
Thanks, Julian.
So super excited about the reply AI ad functionality and chatbots and that inline knowledge based functionality that you showed It’s a huge expansion to our customer IQ suite.
In addition to that, there are other ways which customers can contact you that you need to have the functionality. So Julian spoke about the chatbot, advanced functionality that allows you to do things like Wizzmo lookups, whereas my order, integrates to the back and in all the data we have in customer, There are a few other ways that you can do self-service. So out of the box for all customers free today, we do allow basic web and mobile knowledge base deflection. You can put that today for free up to fifty deflections per user per month that you get for everybody. And if you look at the competitors out there, they don’t let you get any free deflection. So this is a this is a great feature.
For the simple basic functionality.
Additionally, people contact in other channels email or they create forms we have a few ways to enable that deflection. So for email, if somebody contacts you, we allow that email to come in, but we can auto-reply while they’re waiting with some potential articles and if the customer has their problem solved by one of those, that’s great for them because they’re able to get their problem immediately without having to wait for a human being. They get it in a in a faster time than what your average first response time is. And for your business, that’s also great because that means it lowers the number of total issues that are coming in and let’s the issues that still need to be solved by a human being be focused on those agents, and ultimately improves the velocity in in the in the it will lower your handle times or first response times for all the other conversations.
Web, email, auto reply, and then, lastly, suggesting articles inside of a help center form, are all things that are live today. And they’re free up to fifty deflections per month that you get. So meeting your customers on the channels that they have, not just chat is is super important and There was a question, if this involves SMS channels today, it does not, but we are rolling out more channels, as we’re continuing to build up functionality on this.
PJ, before we proceed, there was one question that came in. And, I think you addressed the SMS question that came in from Britney. Oh, Yeah. P large quick.
Yeah. So how would you know that this ticket was solved by the first automated response? I don’t think we have a screenshot of it here, but there’s a few things. We’ll get to it from a reporting standpoint.
Every conversation that’s solved by an automated response, we actually show the articles that were suggested to that person, and we show which if they clicked on a specific article, which ones they read, and we automatically can mark that conversation as done. We also have properties on those conversations. So you can search and see all conversations that were deflected automatically through these channels.
So there’s a great use case of being able to, look at the individual records that are being solved if say that customer contacts, again, later on, the agent can see previously what articles they looked at. And again, it gives them that entire context and able to maybe suggest a solution knowing what questions they’ve had in the past and what articles they’ve already looked at. So inside of our interface, we show that. And then lastly, in a minute, we’re gonna talk about reporting. We use that data to enable your business to know how many conversations are you deflecting. Which articles or what situations are resulting in deflections and which ones are not?
So we’ll get there in just about a second. And then just one of the questions just came in as you were speaking, which was, WhatsApp as a channel in terms of being able to use some of this capability?
Yeah. So we are working with Facebook on WhatsApp right now. They have restrictions on allowing bot messages to be sent in WhatsApp.
So they’re changing those rules, especially in terms of being able to lower these reflection. So we are planning it, but we’re still restricted, but kind of, but what you can do with with with with WhatsApp. So, we will plan on hopefully having that later this year. One second pulls out the water.
So Ultimately, not every customer and every problem can be deflected by human beings. As we or buy a a bot or knowledge base or whatever that is. Even the largest companies in the world that have the biggest investment in their customer organizations and automations still ultimately have customers that have complex problems and need to be routed to a person.
The best way you can continue to provide support by getting them is by matching the customer who’s got a problem with the right person staffed on your team that can solve their issue and is skilled to solve their issue. So how we do that There’s a lot of features that we already have today to be able to say, look at what orders they’ve placed, maybe look at what type of customer they are, and enable you to use those properties to route them to the right person. However, we also have natural language processing functionality that we’ve added recently. So first natural language processing that everybody gets today, is sentiment detection.
It’s about taking the question that they have, the problem that they have, and being able to categorize that and using those categorizations to match them to the right person. So sentiment based off of how frustrated they are, the level of urgency, you could route on that that information and send it to a high priority or or or highly upset customer team, maybe a a a different tier team, tier zero, up to say tier three that we see customers do. So that sentiment detection is stored on every single message that comes in, and that logic can be used to route. Next is language detection.
Automatic language detection. If a customer contacts in via email or say certain channels where you don’t actually get the property of, like, what language the browser is. In chat, we get that browser property. And even then, sometimes that can be wrong. They could be on an English browser speaking a message in Spanish. So what we do is we we enable each message to automatically be run through a language detection engine.
And based off of the language that’s stamped on that conversation, you can route that to a a a staff that speaks specific language. And is able to handle their problem. And then last feature that we’ve recently rolled out that we were really stoked about is called our intent identification feature.
Intent identification is thinking about how can we categorize conversations in a way that’s unique to your business? So whereas language or sentiment, those aren’t necessarily unique on a business-to-business basis. Your company may have features like maybe you’ve got customers that are a buyer versus a seller. And you need to know is this a buyer-related question or is this a seller-related question? Maybe it’s the delivery or It’s about a specific product in your product portfolio.
We can enable your company to use your historical data that you’ve already categorized. So if you’ve said, well, this question is about buyers or say this question is about this type of product. That can build a machine learning model, and this is all in our interface doesn’t require any coding and development. We’ll talk about that in a second, that enables you to set it up so that as those next questions come in in real time, that intent identification can automatically categorize that conversation as, say, if your company has, let’s say you’re an airline and you you have questions about baggage or questions about reservations, they may not even use the word baggage or reservation in that question, but it can accurately identify it, tag that property, which can then be used to route to the appropriate person.
So a lot of really cool ways at which you can make sure that customer’s question makes sure it gets to the right person instead of having them to go to some person that’s gonna need to transfer them and rebuild context. You wanna eliminate the time spent getting to that right person, and that results in better customer satisfaction. And less handle time for your team overall, which lowers your business’s cost.
So eventually, then, that conversation is hand off to an agent, either from a bot or from that previous context that was being asked, and that agent gets to see all of the previous messages that were asked or communicated inside of that conversation and in any previous conversations that they have. So I, as an agent, may have never spoken to this person in my entire life. I get the sentiment they have for this conversation, but also for all of their previous conversations, any questions or articles that they read any orders and things that they have, all of that data is ex exists in the format so that you’re able to provide really personalized commerce customer support with somebody that you’ve never spoken with before.
Next slide.
So we’ve intentionally focused on making this easy to build.
So all of the features that I’ve spoken about so far are very simple drag and drop non code. Just you can go into our setup configure them, and it’s it’s easy to make these work.
The other thing that’s important is that you need to be able to understand your deflection rates, how many questions are being deflected which ones aren’t which articles are being offered. So in order to improve this over time, you need to have that sort of insights into it. So we have standard reporting excuse me, that is automatically out of the box checking and can build reports based off of the percentage of deflections that you have what’s successful. Which ones are not? How much time this is saving your business? How much time this is saving your customers?
And then even further down, it’s not in the screenshot, but we have article specific reporting. So you know Well, here’s a bunch of articles that are being shown a lot. Here’s articles that are not being shown.
And then in addition to what’s being shown, we have separate report for what’s being asked. What are the questions that are coming in frequently? Of those questions, which ones are being solved and which ones are not? So this data is super important for you to understand.
Well, here’s what I need to do to tweak and improve my deflection rate over time.
So, PJ, while you were speaking, two questions came in from the audience, and I’ll open this first one up to both you and Jillian. How does this differ in shape from other services, for example, ADA?
Yeah. So the the first thing that comes to mind for ADA is that it’s just chat on ADA. And as I’m sure many customers here are can can validate this you get customer inquiries on all these different channels and and whatnot. So, being able to provide this in say email or in help center deflection things. That’s that’s super important.
So that’s that’s kind of the first one that comes to mind. There are definitely similar functionalities.
In terms of being able to suggest things.
The other thing that’s super differentiated in terms of what we spoke about is the ability to integrate with, say, order management or the fact that we are a platform that has your orders or your products and services, whatever that is, know, the example that Julian spoke about, the customer that implemented that chatbot, they don’t have, like, in a traditional e commerce use case. Their product is actually a service of a delivery.
And so there’s not like a there’s not like a Shopify store that you can integrate to to to make this work. The fact that they’re able to have the order or the the the service data inside of our system and allow customers in a chatbot interface to be able to access that information and see the status on on that that that delivery means a a larger amount of deflections can be can be done because it it maps closer to what their business is instead of some cookie cutter out of the box use case. Jill, and I don’t know if if you have anything else you wanted to add to that. That’s exactly right.
So this is gonna be built on top of everything that would be in your instance of customers. So all of that data data and other providers are third party platforms to whatever other service platform you might be using. So this is gonna be very this is gonna be native, and you can leverage all that data. And then to, piggyback on that original point, as we’re talking about deflection and zero touch resolution in multiple channels.
I’m sure not all of your customers reach out to you over chat. You get emails, you get folks filling out the contact us, and you might even be talking about different messaging channels like SMS, and social media. So their are other ways that this can be implemented not just in traditional chat.
And, actually, I think One of the other questions was for SMS channels. The Jillian chatbot functionality that reply has in in our customer IQ plus now tier that supports SMS and and is in am I correct? Totally correct. Awesome.
Okay. So I was wrong initially in my response. Feel free to jump in and correct me. I I should No problem.
Excellent. Excellent. We’re all for learning together.
Another question came in, which is how is CSAC captured in the bot? I e survey questions at the end, etcetera?
Yeah. So our our CSAT feature which is available to all customers already.
After a conversation is resolved, we capture CCI Csat score, and it’s not it’s not shown here right now. But, it it’s a feature that’s simple to configure. You can choose how you want to collect CSAT. So whether it’s one through five or thumbs up, thumbs down, you can shoot it’s also automatically localized So every language that the the customer’s in, we support that CSAT to automatically show in in the language that they’re speaking in. And then lastly, you can define a question to be asked, after that CSAT score is inputted.
To be able to collect raw a CSAT data about, well, maybe they’re saying, if they had a negative CSAT, they can explain why specifically, they have a negative CSAT, and you can use that data ultimately to improve your your your systems over time. So we do have out of the box native CSAT functionality.
That’s on every channel including social today.
So, yeah, Let’s see here.
Yeah. There’s one question. Is there potential with reply dot ai to offer proactive reach out to customers browsing to improve conversation rates.
Jillian, do you wanna answer that one?
So there can be in the future. We’ve traditionally focused on customer service, but I know in speaking, with some customer customers that it’s in important to after you have all this service data to figure out how you can be more proactive.
So with the channel open and automation possible? Who knows where this road map will go?
Yeah. And Brian, or to the questions here, to add a little more insights into what we’re thinking about with what the future of replying customers looking like. We see there being three types of communication specifically for chat right now that we’re thinking about.
There’s right now which is inbound questions initiated by the customer where you let the customer ask the question first. And based off of that question, you can do things like run them through a specific bot flow or try to deflect them with WISmo or things like that. Additionally, there’s inbound customer questions where you wanna guide them through that flow. So you can ask the first question If they click on the chat icon, you ask them hi.
What which team what do you have a problem without a category or what’s your name and you can click force them through a specific flow based off of user attributes. What we’re thinking about next is the last step, which is how does our bot define sending a message first to that customer, which would then also take them through specific flow and letting you define criteria that says, if they have a item in their cart or they’ve been on this page for that long or previous earth items that they’ve purchased, you can proactively define, criteria, which would suggest a message for them to be actually seen without without initiating a conversation and then be taken through that flow.
So that’s what we’re thinking about right now.
So another question that came in. I’ll take this one is what systems already need to be in place for you to implement or do you help build from ground zero? So everything that we’re talking about here is based on the customer platform.
Much of what we’ve talked about, really depends on the maturity of what you’re trying to do. And, really having your your knowledge base set up will allow you to do a lot of the omnichannel deflection that we talked about upfront, and then you can get more advanced with the bot capabilities, etcetera.
As you kind of continue that journey through your, you know, your maturity on self-service. So that’s certainly something that we would walk you through and and talk to you about in terms of what are you trying to achieve? And then what is, what is the best way to do it? But the very basic necessity is obviously the customer platform.
And for most of the use cases we’ve talked about here today is having the the knowledge base set up.
Next question in here, or I can get to that next. And just a recap of what the objective is here. Is that that when I think about the goals for a business and specifically for support organization in a business, you’re tasked with improving the customer satisfaction, delivering the highest quality support that you can because ultimately that will result in retention and more purchases and beating out your competitors in terms of what products you offer. Right? The if your customers are happy with your communications, the business, it’s very clear they’re more likely to to stay and buy more things. So support organizations very directly impact that communication So first and foremost, your business’s goal is to improve customer satisfaction.
How do you do that? While customers like to have answers as quickly as they can. And so a lot of these systems enable you to do things like deliver answers immediately. And be able to get that time from question to solution down to almost instantaneous.
But again, at what costs our businesses trying to deliver these quality experiences. I always say, well, any company could provide the greatest customer support in the world if they had one support person for every single customer they had. Right? That agent just focused entirely on one customer and if they ever contacted, they would probably be able to deliver really quality support.
Obviously, that’s super unscalable. You can’t have a single agent for every customer and most organizations operate in like a zone defense model. Right? One agent can handle more and more questions.
If you can improve that ratio, then you can maintain high quality of customer support and improve your satisfaction while lowering the cost and the number of people you need to have as you have spikes either seasonal or daily or weekly spikes to to to maintain that while lowering your costs. So so this is a tool that is gonna really enable your scaling to be on the business side of things and your products and software and not always requiring you to scale up and down support organizations.
Great. Thank you. And in terms of question that came in, what about basing the routing and categorization based on the inquire category issue tree, for example, payment, problem with the app, etcetera, plus intent, plus sentiment.
Is your ticket’s categorization based on tax? Yeah. Great question. This is very specific. I imagine to our custom intent classification feature that I was talking about, the intent identification.
So right now, that intent ID feature uses the first inbound message that comes in from a customer, and that’s the body of the message.
At the moment, that’s the only property that’s actually training these machine learning models.
It is a very from what from the research and testing that we’ve done. It’s a very high indicator of what people are contacting about.
But in the future, we’ve got other data such as, like, the customer type or the question type or category and things that we wanna allow you when you’re building those models to choose which attributes should be put into the training so that and potentially also what level of importance do each of these attributes have in building that machine learning model, which can then do suggestions. So really love that you guys are thinking about that already because we are too. For now, it’s the first version of this that we’ve released and that our customers are live on today. We’ve started off with the first inbound message.
Yeah. And that intent, identification is a very powerful engine, but also you know, in terms of routing, the routing engine and routing logic within customer platform. You can take a lot of the different, variables or datasets that you have to route to very specific teams. So you can do something to say, look, show me all my unhappy, customers that, you know, our VIP that have this particular issue and I wanna route them to to team x.
So you can do some really cool things in terms of taking data that you have about the customer, about the status of an order, and combining it with other pieces of data to route to the right team. Yeah. I didn’t that’s a great call out, Vocus. I should have made it more explicit.
Routing today can use any property. That includes a categorization or issue tree. It can use sentiment. It can use anything that we’ve got. Specific the the intent machine learning model, is is what I was talking about. So, yes, you can build your entire routing logic today based off of what questions or categorizations that are being done already. This is specifically using ML to auto-categorize to add a property which will then be used by the routing logic.
So look, all all of this, you know, we talked about, you know, twenty twenty-two and, you know, where, you know, seventy two percent of interactions, would involve technologies such as machine learning and chat bots. And the key thing for brands everywhere is why are you going to do that? Is to absolutely, solicit savings while delivering amazing customer experience.
According to Juniper research, that number, you know, globally could be as high as eight billion. So tremendous value that people are getting from implementing these these programs.
I know that we have taken quite a few questions within the session, but we do have time for a couple of more if anybody does have any that they want to, chat in we’ve obviously got PJ and Jillian here, so I’m sure they’d be happy to take a question or two before we wrap up.
Christine, is anything from the audience?
It’s not looking like we have too many more questions coming through. I think you guys have an awesome job of answering Oh, here we actually had one come through from Wellington.
The keywords report and deflection. Yeah. So the question was Do you have some updates about when the keywords report will be available at the deflection session of the reports?
We are turning the on in, it’s in beta right now with some existing customers. We always wanna make sure before we roll this out, we test it with certain customers to make sure that, it takes the quality is high there. So it isn’t beta. The feature should roll out next week.
Next question that came in, is what’s your pricing model based on? So we have three tiers of the KIQ suite.
The basic tier is, available at no cost to all customers.
And then we have two additional tiers on plus on top of that and they’re based on a per agent model with all the functionality inclusive. So it’s publicly available on our pricing page on our website.
And it’s fully detailed for you to have a look at there.
In terms of Neil’s question, if we wanted to test this in a bit, where would we get started please contact your customer success manager or account executive and they will happily walk you through the setup of this so that you can get started and experiencing the power and capabilities of the KIQ suite.
Question for you, Jillian, how long does it take to train a bot to deliver a good customer experience?
Great question. So on the deflect widget side for the contact us, that actually requires no training based on the way that our model is built. So it is good to go as soon as it’s implemented on your site.
And that is only really honestly a few days. For a chat bot, we train as we are building your conversation flows. And so if there is data that you have, conversations that you’ve had in the past, we’re training as we’re building. So once it goes live, however long it takes to actually physically build your bot, it is intelligent from day one.
Great. So I think that’s all the questions that have come through and wanna give everybody time back in their day. So thank you so much. It’s been a very productive conversation.
I want to thank PJ and Jillian again. And Christina, with that, we’re ready to wrap it up.