- Sentiment analysis
- Next best action and analyzing customer journey
- Chatbot vs. enabling call center
Tracy Waller: (
And I'd like to introduce to you our moderator and panelists. So moderating today, we have Peter Grant, co-founder of Foresight and our two panelists today are Tim Carlson, assistant vice president AI Accelerator at Travelers and Sanjeev Chaudhry, CEO of Gigaforce. Thanks all.
Peter Grant: (
Thank you so much. Okay, welcome everybody to this session, really excited to dig into this topic, so let's get going, with increased frequency insurers are using advanced machine learning to drive smart automated applications in the field, such as healthcare diagnosis, predictive maintenance, customer service, automated data centers, self-driving cars and smart homes. Machine learning can be effectively applied across structured semi-structured or unstructured data sets as well as the, right across the value chain to understand risk claims and customer behavior. With high predictive accuracy, working with machine learning does of course have its challenges in the customer service space. We, look forward to digging into those, challenges and the opportunities within this panel, to introduce myself I'm Peter Grant co-founder of foresight. Foresight is a worker's comp in short tech providing innovative coverage and value for core industries like construction, agribusiness, manufacturing, and light industrial in the middle market through our proprietary safety coaching service and technology safe site. We give our policy holders the power to drive down their incident rates and insurance premiums through safety engagement, adopting technologies for our customer service has allowed us to not only, be proactive in servicing our customer needs, but also continuously drive success efficiently as we scale joining me on the panel today, as we heard is Tim Carlson AVP at travelers insurance and Sanjiv chaudhary, CEO of giga force. So gentlemen, please introduce yourselves and,, love to hear what brought you to dig in and what you're excited about.
Tim Carlson: (
Thanks Peter. So, Tim Carlson yes I'm as in the previous panel as well, for those of you who recognized me. So I've been at travelers for, 16 years. I'm part of our enterprise AI accelerator team. I worked in, pretty much all our different lines of business, over
Tim Carlson: (
The past 16 years, coming to dig in, I think, can say the word I'm here, so I, think just, learning, hearing more what's going on, in the space, obviously, we're deep and my team's really deep into a lot of the application of AI at scale, understanding, what others are doing, is definitely something that we're interested in hearing.
Tim Carlson: (
Excellent. Thank you.
Sanjeev Chaudhry: (
I'm Sanjeev Chaudhry from California and, CEO of Gigaforce, we do have a platform which is built on top of blockchain and AI. very specifically we focus on subrogations or first use case subrogation salvage, and couple of other use cases. So, we basically support the insurance ecosystem, basically the companies who support the insurance company. So that's our product. Yeah.
Peter Grant: (
Very interesting. I'm sure that, both of you, gentlemen have, some very interesting insight into how both AI and or AI and the latest technologies are influencing, customer service. So I'd love to hear if you can give specifics about, how you've done that within your businesses and your roles.
Tim Carlson: (
Absolutely. So, I think the example I give was primarily on our, virtual assistant, and providing some really automated service for individuals who are, coming in and looking to interact with travelers in a number of, of different ways, looking to, we can dive into some examples, some, specific examples as well, but I think it's really all about, providing real time information, to customers for what they're looking for, and very often when you're coming in talking to your, insurance carrier, you're not doing this on a daily basis, not like you're dealing with your bank. So it's really figuring out for when people need us and need information. How do we quickly resolve what they're looking for? so they can go on the day.
Peter Grant: (
Sounds like there's a lot of return on investment opportunity within, that deployment. Absolutely and, sanjiv.
Sanjeev Chaudhry: (
So, very specifically what we are, using the EA part, right. If, you're familiar with what subrogation and salvage use cases are, like this is the one when the, insurance company, sometime they don't want to do it, or they will pass it onto somebody or the subrogation companies or TPS, right. That basically take over the claim there. So one of the thing that we do using AI, part of it, just doing the assessment part of it, basically whether this has a subrogation potential or not. Okay, of course you have to go through the whole data set just to train the system there and what is the recovery potential there? So this is just one example here, the couple of other places where we have used it very significantly, is like, so a lot of information goes as part of the, when the claim goes outta the system, like a lot of personal information, P H I and P I information using the part of it, we do reduction of those information.
Sanjeev Chaudhry: (
If the customer wants it to be this way, or sometimes, when the, you even in this, all the digitized work, we know we have all kind of customers, all kind of companies, some will give the claims as part of the, very well organized, different put in different folders. And somebody will just send very send you, as digitized as it is, like one long PDF. And, we use some of the AA part to split the documents, like, in various categories, whether it's email or invoices or whatever it is. Right. So that kind of places where we use the AI today
Peter Grant: (
And presumably deliver those insights back to the service provider or TPA. that's correct.
Peter Grant: (
Right, and that sounds almost essentially the tip of the spear or core part of your business, that AI and that delivery. That is one piece of it there's one piece of it. Absolutely. I'm sure there's more around that. Yeah. so, I think the interesting element for me, and I mentioned this, the return on investment, should be clear, or if it's not, you should be able to demonstrate it over time. what sort of resistance have you faced in deploying? Certainly if it's not core business, like evolving a business, and how have you been able to demonstrate the return on investment?
Tim Carlson: (
I think with anything new, with, any change initiative there, there's gonna be, certain aspect or resistance, how we've gone about, again, past that. I mean, I think a lot of it is just, small wins and starting to show some, some small wins and, then looking to build upon that. So, in our virtual assistance space, we, you went in, you, roll out initial model, you roll out something, that's able to, you understand what a, customer's saying, you look at, how the customers are interacting with that. You look at the data, understand what customers are looking for. Maybe things you hadn't previously, thought of, or, had addressed with, the initial deployment and you improve upon that. And I think, you're providing transparency back to the business and showing, here's the types of things that are being asked. Here's how we're handling it. Here's where maybe we have some gaps with our own technology of, w, we're looking to actually do certain things aren't, you, possible today, you with our current tech stack, but I think it's all around, getting some of those small wins and then just being, really transparent with the business.
Peter Grant: (
And are you able to share maybe some of the KPIs or SLA's that you sort of focus on as a group at all?
Tim Carlson: (
I think you definitely wanna understand, how often, people are asking something that, that you had anticipated and you understand how often are they going on kind that happy path versus unhappy path. And you'd be amazed in looking at the data, the number of edge cases that people have are various things that are happening across your business, and really understanding, there trends here? Is this something that's kind of one off, is there is a situation here that, we actually need to, pass some of these conversations off to other individuals to deal with certain situations? but you, I think it's definitely understanding, how often we're, we're helping someone. I think, we've definitely had the ability to get kind of survey feedback from the customers themselves.
Tim Carlson: (
Although I'd argue that, you, we should be able to tell by looking at the data, if we were helpful or not, but we definitely also have that ability to say, do we, we resolve it? Are you satisfied with what you received? and then, even beyond that, looking at the, business outcome, so if someone says they wanna, pay their bill or, or add a vehicle, I mean, are they actually doing that via the digital experience first, call center or, some other, some of our methods. So I think there's, there's a number of different ' to look at from there.
Peter Grant: (
Absolutely. Thank you. And Sanjiv, I guess some of the challenges that you've, seen in, deploying this AI on a customer service basis.
Sanjeev Chaudhry: (
So, one of the biggest challenge, right, is on the data part, right, is like the old theory that we have it right. Garbage and garbage out, that still proves in this world, like, more in the data science world, it's, with all the like information which comes here, people are getting a little bit prejudice or they're getting biased towards the data, and that can take your model off. Right. So somewhere, I think somewhere it, consistency has to be there or somewhere, some regulation has to be there, which can at least control the data quality part of it, And also one of the challenges that we see it, right, mostly from the data perspective, it is the enterprise or the companies, they gain the data, they gain the information, they gain the knowledge, or they make the money, or they get their ROI.
Sanjeev Chaudhry: (
Right. But it is very limited for the consumer to get the benefit. And that's the reason they don't come so forward. in terms of sharing the, some of the personal information, which they feel, it could be misused or whatever. Okay. On other part of it. Right. I think, if somehow, companies make the policies, picking it very clear, like, their data, whatever the data they get, it is only for the aggregate purpose or consumer perspective from that doesn't hurt the consumer at the end of the day. So they get a little bit more open. I think that is the biggest challenge that I see in the data industry right now. Yeah.
Peter Grant: (
The, so with subrogation, presumably there's potentially conflict with the, outcome of the assessment of the AI. Do you have, I guess, downstream, challenges against that, and maybe you're recalibrating the AI,
Sanjeev Chaudhry: (
Actually, you have to do that. Like, any other model part of it, but what happens here is most of the subrogation that we do it right now, it's mostly B2B. Right? Sure. So, the consumer doesn't get reflected or as, at least as of today, the moment they come to know, if I put some extra keywords, I'll get better compensation or they might play a little bit differently, but at least at this point of time, I think, one of the thing that in the last panel also, we just heard about it, right. There's just so much of, work needs to be done in the insurance space, but there's not enough technical staff to do it. So we have a lot more to do a lot more to cover there. There's a lot of potential there, but I think fundamentally it comes to, how many models you want to take it as part of, your solution there.
Peter Grant: (
Absolutely. Yeah. Thank you, I can certainly see the advantage in investing in that, over the, as I mentioned, adjacent to the core business, over time, so we, we have chatted a bit about chat boxes and I believe that's an oversimplification of what we're talking about here, I believe, but, certainly I guess, what are, some of the other challenges, I guess, how do you believe that, chat boxes themselves where they've come from and where, do you think that they're going to in the future?
Tim Carlson: (
I touched about it a little bit in the previous session too. I think it's really, how do you integrate back in with your existing tech stack? So, it's great to understand what someone's looking for, what they're trying to do, if you can't actually take action on that and, do it for them, based on your existing tech, it's really hard to integrate that back, into the business process. So, you can build a great technology, great thing that model that understands what people are trying to ask for, but then if you can't actually do the action and have the ability to, go out on that path it's not gonna be, as helpful for them. So I think there's, there's that aspect. I think there's the challenge of, culture.
Tim Carlson: (
And I think I've heard that a few times through some of the sessions here of, if you're not building a culture, that's able to think, through kind of iterative improvement, continuous delivery, having the ability to see, something, maybe a new trend of what people are asking, different types of questions and update your models and roll those out and just kinda have that mindset where it's not necessarily a, 3, 6, 12 month roadmap. It's really a product that you need to be continuously you improving upon and rolling out. Those are, things that, are top of mind for me
Tim Carlson: (
Go set and forget that's for sure. Yeah, no, we can't do that. and Sanjiv,
Peter Grant: (
Are, are you implementing chatbots at all.
Sanjeev Chaudhry: (
With yourself? Well, not in this current in my enterprise right now, but I have done like some work on the chat box historically, and that's one area actually, which is pretty close to me, what we have seen it, if like I just was reading the numbers there in from 2019 to 2020 or so, right. We saw 40% growth. Actually the number went up to 188 billion, all the sale were done online. Okay. And there is no companies like, which will be able to support that kind of a growth there. Okay. Just in the month of November, December itself, that was the peak time when the COVID hit the online sales were a hundred billion dollar, which was never had seen it. So there has to be somehow the chat box has to come into the place to resolve that problem there.
Sanjeev Chaudhry: (
Okay. And, good part of, I'd say the insurance insurance industry also lagger in terms of the technology adoption for obvious reasons. Okay, but this is one place where we can learn a lot okay. From how the chat box should be implemented, and it's not, it's not just the chat box is one part of the technology, it has to be integrated with a lot of other things too. Like if a customer is calling again and again, like you, somehow the chat box has to know that this is the frequent color. So we have to do it a little bit differently. Right. The impatient guy. Right, he will probably won't even wait for another five line, five minutes to talk to somebody live. If we can get the answer done. I think putting all the things together, there's a huge scope that chatbox can do or are able to do it.
Sanjeev Chaudhry: (
And one of the thing which probably in last, I would say three years or four years, which has made a big impact from the technology perspective alone is the NLP part of it, the natural language processing part of it. Now that can put lot of things in context and chat box could be a lot more smarter. Okay. So I think, there's basically two parts of that, right. Is like how much chat box is able to solve the problem by itself and how much they're able to assist it with existing, like call centers all. So I think that's where I personally feel chat box has a lot of growth opportunity and, they can do a lot more than they were able to do it before. Yeah.
Peter Grant: (
And certainly high volume, call centers or customer support, centers, travelers, as an example, I'm sure during COVID as, consumer behavior or customer behaviors changed or evolved, if you weren't employing a chat bot or technology at the risk of talking about chat, chat bots too much, if you weren't employing a chatbot, service or solution, with labor shortage, and just the increase in services that you had to provide, it was a must have as far as, I understand it
Tim Carlson: (
Well, and I think it validated a lot of our investments. so I remember that at first, mid-March, everyone was home for, two weeks, so everyone's home, I'm at my kitchen table. Yeah. Which, which was the desk. And,, you, I was looking at, chats in real time. So we'd enabled a capability to look at that. And one caught my eye like, oh, I just lost my job. And I can't, can't pay my insurance this month. And I went back into the logs and like, oh, we had a few more of those that day, which is something I hadn't seen before, and then, so something, that night I actually shot our business partner and know like, Hey, this is something we gotta keep an eye on.
Tim Carlson: (
And the next day we got like 12 of those messages and so we actually, had to redo, re, update our model to, have a response like this, this COVID response for what was going on. And, ultimately, as, we, realized kind of where this, where COVID was going, and, travelers had a 15%, discount for, basically all, all the policyholders. So we actually were, you had additional questions coming in now related to that. So we had to update a lot of our, our models and messaging, that that was code specific, but yeah, we definitely saw an increase, with message volume as well. But I think it just validated a lot of, kind of the things that were more hypothesis of where we're gonna go, just made it happen a lot faster, so yeah, I think, there was definitely that validation that, that was good. Yeah.
Sanjeev Chaudhry: (
If we could add, this is interesting, you, in my past life, we used to work with a product company, which is part of SAP now, and this is the company was trying to set up their call center, in India at that point of time. And, the group, which was supposed to take this, initiative was out of French, France actually. And they were not only supporting Europe, they were supporting pretty much the whole world actually, including Australia or whatever. Right. Okay. So, and there was a little bit of resistance, right. Or Indians people have accent, so do I, but that's it. Okay. So, should we even have the call center set up in India there at that point, this is, I'm talking about 2004 or five or something like that. And that's how the SVP there in this company, he said, you got a choice either we don't respond to the customer or we respond with the accented one.
Sanjeev Chaudhry: (
Right. So which, what do you think your customer will make it happy? Right. So I think with chat box, that problem is also kind of taken care. Right. So, that's another add on to can do it. And the funny part was this, SVP who was out of Canada and he said, you know what accent we are talking about, every part of Europe has a different accent. Right. So, so anyway, the, as I said, the right side of this is right now is like chat box has taken care of that part of the problem too. Yeah.
Peter Grant: (
The, accents certainly. Yeah. That makes, that makes total sense. so I guess,
Peter Grant: (
I'd love to hear, your impressions talking in the customer service space and some of the technologies that you've seen out there maybe coming down the pipeline, or, some of the more forward thinking, technologies, or even, a technology that you may have seen, that's been adopted, ideally outside of insurance that has been super successful, or setting the pace, what out there is interesting to you as, from a customer service perspective, I guess in the interest of efficiency customer satisfaction.
Tim Carlson: (
So I could go on first. Sure. No, but, so I mean, there's, there's really, a couple cold path. You can take it as I kind of describe it when you think about kind of the tech stack, there's, I think some of the, larger cloud you carriers that really provide kind of all in one services, like top down that they, have services that cover a lot of the capabilities that you'd be looking for, and then there's more of kind of a startup world where, I think there's, a number of really interesting, , companies out there that as you start to kind of piece together, some of the capabilities, you can start to create some pretty compelling products. so a lot of it's really gonna depend on what's the appetite for, doing that.
Tim Carlson: (
What's your, team makeup? do you have the, skillset, cause not every, team is going to, like maybe some teams are gonna have, the specialists that, that have the ability to do that, but others maybe just, you're maybe working with a business partner with a really small it budget that, was really looking for something that's just a bit lighter touch on the it size. And there's like kind of those tradeoffs that you have to, consider, and that's just not for virtual system. I mean, it's really for any type of, kind of AI, initiative, I mean, theres a couple just considerations you have to think through. Yep.
Peter Grant: (
Agree.
Sanjeev Chaudhry: (
Yeah. So, I think the questions on the customer experience part, right?
Peter Grant: (
What are you saying out there? That's, that's interesting. And perhaps, inside or outside of insurance where you you've seen a technology, deployed that has been, effective at customer satisfaction or customer service.
Sanjeev Chaudhry: (
I think the fundamental thing right now, since most of the things are online, right. Customer experience becomes very important. Right, onboarding experience, how long the customer retains on the site and all that. Right. I think, those are some of the, I would say probably a good reflection, of the model, if you could say that and then, how much people are able to refer you, okay. To the other, other parties and all that. Okay. Yeah. I think probably these are the, some of the three or four things I would say, which should probably reflect that. Yeah.
Peter Grant: (
I think I tend to agree certainly, it relates to obviously your business process and the user journey and the, whether your CRM or your E P or however you are deploying your services or technology to your market, that, I agree Tim, that sort of is where all the data or the process is centralized. And if you need to have intelligence on top of that, to be able to deploy, whether it's AI or, an update to your process, or automation, so certainly, yeah, some of the big marketplaces, within that space, I think are really effective. One company that, we've used in the past, Intercom, I'm not sure if anyone here is familiar with it, pretty big naming customer support customer, service. They've been excellent, at, at providing all of the tools that we've spoken about, but there is obviously a resistance for them to completely integrate with other CRM services.
Peter Grant: (
And I don't, I know why I'm not gonna talk about why or why I think about why, but, it is, it is a challenge for us to integrate that data and information from our customer service technology, into our CRM technology. Because as you said, you need to have those learnings, you need to be able to respond, you need to be able to iterate, so, I think as, long as you're able to, whoever is whichever system is touching your customers, needs to be integrated. The data needs to be integrated with your business processes so that you can, you can integrate it, so with that, before we, before we, throw the mic out to the audience, any closing thoughts on this particular topic that you wanted to, announce to the world?
Tim Carlson: (
I mean, yeah, I don't think there's any silver bullets for a lot of this. I mean, I think, it really goes back to, looking at data, really understanding what customers, want and, then kind of it early improving and building out products.
Sanjeev Chaudhry: (
So I think one of the thing that I'm seeing in the space here, right, in terms of the AI, the companies are then like which is on the pretty much on the leading edge part of it. Right. I think with all with no biases one way or the other on this one. So I would probably say, probably Microsoft and, Google, they're pretty, pretty strong on the, AI part of it, whatever they have to offer it in terms of technology. And some of them gives it a very reasonable price too, open AI, which was supported by Elon Musk go and where the other, it's not so open, but again, ID probably say they're also probably one of the good leaders in this space and, couple of, schools, or I would say like CMU or probably Waterloo, Canada and all that.
Sanjeev Chaudhry: (
They're pretty, strong in putting all the algorithms together there. So I think something, if you're a data scientist here, something you have to follow, I think these are the places which you should look into from the following perspective there. Okay, now the other places where I personally feel where AI will make a huge difference and which is across, right, whether like using with the blockchain part of it, actually that could be a huge significance. It could be used for the, in the distribution. And we use very extensively as part of the claim system. Anyway. So integrating the not only AI from, just as a single piece, I would say probably integrating with other technologies, including RPA, you can do a lot more there. So I think one of the biggest thing that we have as insurance companies, right.
Sanjeev Chaudhry: (
We have data, we have a lot more to play and, all these things we can do to play, but I think somehow we have to see our consumer gets benefited and I think we should be a little bit transparent here. I think we should do a good job of, saying it or know why we have to use what we have to do it and how you are getting benefited. So in the beginning I itself, I said right, if the consumer doesn't give you good data, that you cannot do process much there. So I think that could be somewhere, it could be a good relationship, not only from the consumer perspective, even the small business players, like even the brokers or whatever it is. Right, just big them part of the system. Okay. So that way, at least they contribute more and, you can help them and you can help everybody else in the picture there.
Peter Grant: (
Excellent. Thank you very much, I would love to double click on the blockchain application, bit of the risk of going down a rabbit hole. we'll turn it out to the, to the audience for any questions that you might have.
Audience Member 1: (
I thought, Sanjeev raised a really good point, concerning accents with the, chat bot. And I wondered, because, access and, language in AI is such a hot topic right now. I wondered if there are any other, trends or things you think we should be aware of as concerns, access and language in customer service or things that you guys are working on in all three of your companies, concerning language, and accessibility in customer service right now, as concerns, AI, or, chat or any other resources enhancements as concern language.
Peter Grant: (
One part of our service is that we, aggregate historical safety information from a client and, bring industry related safety information and historical loss trends for the client. We, we put it together in what we call a safety success plan, which is designed at targeting those loss drivers with our service and technology, a lot of our customers within the mid-market, particularly in the construction space, English is not the first language. and so we use AI, to make sure that all of our documentation, all of the follow on processes are in the relevant language. Obviously Spanish super important, obviously, the Spanish dialect is, as an Australian I'm learning this more and more is very different, regionally, and it's been talking about customer satisfaction as we've done that, the, amount of engagement amount of, customer success that we've seen off the back of that we call it safety success, has been tremendous because it is much more accessible, for those particular companies and individuals,
Sanjeev Chaudhry: (
One thing, which I would probably say, I think I touched on the NLP part here, and there's something new concept is coming. The transformer part of in the AI space. I think that's something we should watch out for. This will probably bring lot more revolution in, the probably near feature there N LP has been there for like last six, seven years or whatever it is, but I think the way people are able to use it, the people are able to use it for different purposes. I think we are going to get a lot of benefit out of that one. Yeah.
Peter Grant: (
Any other questions looks like a no. Excellent. Well, thank you, Tim. Thank you, Sanjiv. Thank you everyone for joining us for this session. I hope everyone's a little bit more informed, about AI and customer support, customer service and success, looking forward to, engaging with the rest of the conference.