How AI Chatbots Are Transforming Customer Engagement – Making MarTech
Unknown Speaker 0:02
Welcome to Making martech, where we spotlight innovative martech companies and leaders who are driving growth, highlighting valuable insights and actionable strategies to stay ahead in this industry today, we’re here with Anushka, co founder and Executive Chairman of botcope.ai,
Unknown Speaker 0:24
hi, Anu, how are you? I’m good. Thank you.
Unknown Speaker 0:28
Awesome to have you on today. Thank you for having me. Yeah, no problem. Can you introduce yourself a little bit about what you do? Sure, I’m a senior entrepreneur, and I’ve had about a billion dollars in exits of mostly SaaS companies in martech and ad tech. And in fact, interestingly enough, I had one of the first martech companies. I was a company called rubrik, more than 20 years ago that really introduced the for the first time, some concepts like drip marketing, etc. So, and it was entirely built for the internet. Internet was just getting started, even at that point. So yes, it was a company called rubric that got acquired very quickly for a large sum of
Unknown Speaker 1:08
money. And so currently, I’ve got my fingers in a lot of places, but mostly I am working on a co founder of a company called botco.ai
Unknown Speaker 1:18
where we are building DAAS platform that’s used by enterprises to build chat bots for better customer engagement, for customer support, and we focus very much on the vertical markets of healthcare, behavioral health, Senior Living, Pharma is a big segment that we go after, and city and local and state government applications wherever people want to engage with customers, improve the customer journey and close more more sales faster, or they want to reduce costs and offload some routine support work onto an Agent and automated agent. So that’s what our agentic platform allows them to do. And can you walk us through that? What does that usually look like? No code platform that you can use to quickly put together the flow of a chat, and then you can ingest
Unknown Speaker 2:19
documents and PDFs, you know, emails, whatever you have, you can ingest that to train the bots. It’s all AI based, so it’ll train exactly on your materials. And then we have, like, built in guardrails for, you know, if you want. You know, enterprises are really not
Unknown Speaker 2:37
very tolerant about hallucination, so we have different ways to prevent hallucinations, so that that chat bot is accurately representing your company.
Unknown Speaker 2:49
And at the same time, since you know, we have created a sort of an agent that’s interacting with customers on your behalf, either answering the support questions or telling them about your services and your products to take them to the next level in their customer journey, hoping eventually they will get much faster and more streamlined way into actually buying your products, or revenue oriented sort of activities. We are also cognizant of the fact that information changes, so we have the ability for a chat bot to go in and get refreshed by themselves, by taking a look at the latest information on whatever data source you point them to, and refreshing their answers accordingly. It’s really for an enterprise. It’s, you know, for fairly robust bot that’s going to reduce your costs quite a bit, ROI, or increase your sales quite a bit. The ROI is very clear.
Unknown Speaker 3:43
It’s not a very long process, a three to four week process where it can be built, a bot can be built and launched with about 96% accuracy, or 94% accuracy, in some cases, and then you can fine tune it. There’s a training interface where you can see what answers are not correct and what sources are not being pulled up correctly, because it actually references the source from where it’s getting the answer,
Unknown Speaker 4:09
and you can fine tune it more to get it up to 98%
Unknown Speaker 4:13
now, we’ve got some long standing chat bots that are running for a while, And we are we’re reporting or 100%
Unknown Speaker 4:21
on those that have been running for over a year.
Unknown Speaker 4:26
Can you dive in a bit more about that training aspect of it? Yes. So as I mentioned that you can train your chat bot by ingesting all the documents and information that you want the chat bot to learn and then answer questions. And we have an interface that’s showing you how the entire conversation. It’s pointing out to you which ones were answered correctly, which ones they answered could be better, which ones there was a fallback, fallback, meaning I can transfer you to someone or call this phone number to get your question answered. Or I’m not trained in.
Unknown Speaker 5:00
That subject, so you can look at the fallback, so you can produce better answers for them and ingest them right away. And the minute you ingest them, the bot is trained on that. So this, this is the way you get it from a 95% accuracy up to, like a, you know, 98% accuracy by fine tuning it. And also, you know, because it’s an enterprise and data is not static. It’s not a static model. It’s trained. It retrains itself on new information, so you can add more information to it, and it’ll now get trained on that information as well, and you can do this in an automated fashion. So this is how enterprises can trust that a chat bot that bears their brand name is sitting on their website, or people are accessing it through SMS. For example, is always going to give in be available. It’s always going to be on, it’s always going to be giving, giving you the right answers to your queries without hallucinations. You mentioned that they can use this through SMS as well. So yes, they can. What channels can this chat bot actually
Unknown Speaker 6:03
use? The chat bot can be on your website, it can be on Facebook using the messenger chat channel, or can be on WhatsApp. It can be on Instagram, or it can be just access through text. But unlike dumb SMS, where you can just press buttons, it’s actually AI trained, so in as on SMS, over any mobile phone, you can carry on a conversation where the chat bot is actually understanding what you’re saying, and you’re not limited to a button based or a scripted flow.
Unknown Speaker 6:35
That’s interesting. Where do you see the biggest value in that?
Unknown Speaker 6:41
In chat bots? Well, the ROI is incredible. You know, we have something called an ROI calculator that is used by our clients to determine if they want to make an investment into a chat bot. So for example, I’ll give you example of a medium sized city in Texas, right? They were getting three to 400 calls a week. Most of these calls were related to
Unknown Speaker 7:02
trash.
Unknown Speaker 7:04
You know, when is my trash day for this neighborhood? What can I put in trash? Can I put glass? Is that glass recyclable? Or is it trash I have a big sofa I want to get rid of. Can I schedule something for trash pickup now, three to 400 calls, frankly, even the mid size city didn’t have the bandwidth or the resources to have a human being answer every one of those questions, plus those questions were, I would say, 80% repetitive, the same right now, this was a very good option to put in a chat bot. So they had about 1200 pages of information on their website about the city and the city services and the chat bot, we crawled the entire website and trained it on all the information that the city has on their website, and they gave us some additional documents that we also ingested, and now we were able to launch a smart bot, an agent that was able to offload 80% of those 300 to 400 calls per week and also be available seven by 24 in the middle of the night, early in the morning, and answer these questions about trash and about other things and on a weekly basis, when it saw information had changed. It picked up. The new information crawled again, refreshed, has new information. So if you may the mayor change, or the trash day change in certain neighborhoods, it would have that information. So saved an equivalent of two or three full time people that cost a lot of money, but also increase the customer satisfaction because they were getting the answers that they needed right away.
Unknown Speaker 8:46
And what do you think some of the biggest misconceptions are that people have about this? Because clearly, there’s a lot of value here. But do you think that people fail to grasp this? Sometimes
Unknown Speaker 8:58
there are generations of chat bot that technology is advancing. So it’s a good question. What misconceptions people have is that it’s very expensive, and I won’t be able to afford it, and I have to have aI engineers in my company to actually use this. And so no, you don’t, because these no code platforms like barcode.ai take care of all of that stuff for you. So you pretty much have to look at a dashboard and verify that the answers are what you wanted them to be or provide the information. So no, you don’t need to build an AI infrastructure and a team to utilize AI agents or AI agent platform, right there is it’s a no code platform. There’s literally no learning curve. If you can learn Word and Microsoft PowerPoint, you can learn using a no code platform, right? You do definitely need somebody to walk you through it, but it’s not a big push training. So this is not a heavy lift using products like ours. Second misconception is that people are not comfortable with a chat bot. They want.
Unknown Speaker 10:00
Talk to a human being. Actually, the vast majority of people just want their answers. They just want their tasks to be done quickly and efficiently. They don’t really want to spend time chasing down an IVR system for an actual human so no people. People want to get the job done. They don’t really care if they talk to they don’t think that you’re a bad company because a human being didn’t talk to them right away. That’s that’s number two. Number three. Enterprises worry about security. This is data, especially healthcare enterprises. They have to worry about phi, therefore worry about HIPAA compliance. They have to worry about SOC two. And companies like ours. Have all of those audited and verified tests done. So we are SOC two level two compliant. We are HIPAA validated. We are, you know, we adhere to GDPR, CCPA, all the privacy laws. We know how to treat and handle data. So that’s another misconception, is that all my data is going to get out on the internet and everybody will know it. No, it’s very secure, and you and your customer and your healthcare information is safe when you use a vendor that has all these certifications. So that’s the third misconception is that it’s not safe. And the fourth is really by by virtue of the generations that we’ve rapidly evolved from earlier charts, including ours, were pre scripted, right? They were pre scripted, and they relied heavily on a small amount of training. And with the advent of larger, more expensive llms, like,
Unknown Speaker 11:40
like, chat, GPT, Gemini, you know now they have much more advanced capabilities, and so you don’t have to worry about having a scripted card rail there, which is limited usefulness, because it runs our esteem. So these are all the misconceptions that are lingering out there that we do our best to try to shed some light on.
Unknown Speaker 12:01
And how do you think that getting rid of those guardrails that you mentioned? How do you think that affects the personalization of chat bots? Do you think that enables companies to have a more personalized interaction with their customers because you’re talking to someone one on one? So I’m not talking about getting rid of guardrails. I’m talking about putting some guardrails in place on the AI so that you can control the experience for safety and security, which I just mentioned, and privacy, which I just mentioned, are very important to enterprises, right? But you have to put you can put guardrails in place without sacrificing functionality. And yes, a chat bot is one of the most personalized interactions you can have. Now, compare a chat bot to an email. An email is a one way thing, right? You send an email to somebody, and most likely, you send the same email to 1000 people, right? But here you’re having a personalized conversation because somebody is asking you specifically what concerns them, right? I’m so and so and I, you know, I need memory care for my parent. Or I, do you think I have this eating disorder, right? Or do you think I have alcoholism? Like, how do I tell if I’m an alcoholic or not, right? Okay, tell me your symptoms. You drink alone every day. How many drinks do you have if you ever blacked out? So these are all PERS. You can’t be more personalized than a chatbot having a one to one conversation that’s tailored to your questions and interests, and the guardrails are in the in place to protect their information and to protect the type of information you give them. So for example, in certain cases, our Chatbot is trained to get some signals that somebody is maybe having suicidal thoughts, and giving them immediately a phone number on the National Suicide Hotline. Say, please call this number and for help, it might help you to talk to a human being now, right? So things like that are, what are called superior guardrails and going off that you mentioned them being able to talk to the chat bot about their symptoms, right? And when I hear chat bot, right, I think of customer support chat bot that tells me how to fix this technical issue, yeah, right. But a lot of the times I don’t really end up thinking about, yeah, this is the chat bot that I can talk to about my symptoms, that can help me know what’s going on. And I think that’s really unique. How do you think that that’s affecting the healthcare industry having that kind of functionality in there? Well, it depends. The healthcare industry is very big, right? And so they are products that are potentially helping diagnose we try to stay away from diagnose diagnosing. We just want to get them to the right person. So for example, if you’re approaching a behavioral health center or addiction center, they’re asking a few questions that they would have asked anyway if you had called them right if you talk to a human being, they would ask you.
Unknown Speaker 15:00
Those questions just to get to the right counselor, right? You know, very simple thing in healthcare is like, Okay, you may, you may have an addiction. Is it covered by your insurance, right?
Unknown Speaker 15:11
Or you should know how much it’s going to cost. Otherwise, there may be other resources for you, right? Go to. So this whole intake process has a lot of things. I have heard of, chat bots that are specialized just for diagnosing certain ailments. And there was specialization in that right and I don’t think we at Barco AI are claiming to have that specialization. What we are claiming to do is to get them to the right human being for a proper intake, saving them time and saving you time as well, because you’re not talking to qualified people, and the people that are not suitable for you have already dropped out, right because you found out that they have an insurance that you don’t accept. How do you guys end up minimizing hallucinations in this it’s by these fine tuned models that we do. It’s by restricting the documents that the chat bot is trained, trained on so it can’t hallucinate if it doesn’t know that there’s something beyond this universe.
Unknown Speaker 16:16
So those are the different techniques that we use. And you talked about this web crawling process, what does that usually end up looking like when you’re training the chat bot, it’s very it’s very simple. It’s just to save time. We basically allow our chat bot to crawl site, because we’re also crawling for changes. So we train the chat bot this week, next week, some information change. We want to know what the latest information is so we set up, you know, regular crawls to get the update information. This is just simply a way to save time and allow a chat bot to be maintained without human, you know, human input every day going off this, what are your biggest recent successes that you’ve had with these chat bots? Well, I just looked at, you know, an application that we’re doing for a client, and the chat bots been functioning for about a year, and this was having 100% success rate, a success rate, and also that they’re getting using it for referrals from social service agencies and other organizations, and they’re basically getting over half of their referrals. You know, through the chat bot, that means it’s been a well accepted people like to use it. They’re less prone to error. They’re saving them time and money and
Unknown Speaker 17:36
and it’s, it’s having, it’s able to answer any free form questions, they have all of them 100% of the time. That’s really good. That’s a great success. Secondly, we also, just recently, have been working on a over the counter medication that’s being launched in a particular region or territory, and we were able to take a high volume of calls and give accurate answers and engage with people and answer their questions about this new medicine without giving them wrong information, and while protecting their privacy, because when you asking about a drug, you don’t really want other people to know what ailment you may have, right? So you have to really protect the information has to be really anonymized, and data has to be treated very carefully. But there’s a great deal of interaction, high volume, a lot of excitement, and lot of questions and answers were being asked to the bot about about this over the counter medication that was coming newly into the market, and people were able to get the questions answered. They took a lot of surveys. They took a lot of surveys to see whether this medication would help them. They also asked for, you know, where can I buy it? And they also asked for, like, a introductory coupon. We were able to answer this question about this new medicine, and also increase the sales behind these two successes that you talked about, you find that there was one specific factor that stood out to you as a driver to that success. Well, in the case of over the counter medication, your audience is millions of people. So you can do a broadcast communication with them. You can go into social media, and you can take television ads, you know, put it, put print ads, but if somebody has some questions, you really can’t have human beings talk to them. So really, an AI agent is probably the best suited for that, because, you know, you’re talking about 1000s of concurrent inquiries. So this is a perfect example of training a bot using AI and creating an agent that can answer these questions and thereby move people on to realizing Yes, this would help me, because they can’t ask the doctor. It’s over the counter, right? It’s over the counters. So that, I think that’s one of the reasons putting.
Unknown Speaker 20:00
A chat bot into the mix really helped people to find out enough to go in and try it right. So there was a very good format for it. And in the case of referrals, this, the next case we talked about, is just so much more convenient not to to, you know, try to type in a form or to send faxes, which was the method, or try to get somebody on the phone, so you can discuss that with them. There’s a lot of phone tag back and forth. Faxes have to be examined and signed. So this was just like convenient, a way of getting, you know, sort of mistakes out of the system. So I think that’s why it worked so well. Wow. So I know you mentioned that you found it really helpful when you had this customer base of millions of people, right?
Unknown Speaker 20:52
So do you find that for that reason, you guys typically go after these bigger enterprise level companies? Well, we do have big enterprise level companies, and, yeah, we’re very good for the mid market as well, right? So it’s just a question of, you could have a mid market company which doesn’t have millions of customers, they may have hundreds of customers, but each customer is very valuable. Let me give you the example of senior living, right? So right now, I think 65,000 people are turning 65 almost every day in the USA and so many parents, families seniors are looking for senior living. Sometimes they need Senior Living, independent living. Sometimes they need assisted living, sometimes they need memory care. Is Senior Living?
Unknown Speaker 21:39
Person A memory care person who comes into a facility is probably going to spend, or be worth the ASP for that is about $200,000
Unknown Speaker 21:53
right? Especially with memory care, it’s maybe 250
Unknown Speaker 21:56
so you may have only hundreds of people inquiring, but every one of those that you can effectively capture is worth $200,000
Unknown Speaker 22:05
to you. So that’s why it makes sense to have human beings, chat bots and whatever you can to make this journey streamlined, whereas the people who are buying the over the counter drug are probably spending 20 bucks, right? That’s why you need a million of them. It’s clear this brings a lot of value either way, though. Yes, I think that’s really great. Thank you.
Unknown Speaker 22:30
And one of the things I like to say a lot right, is
Unknown Speaker 22:34
there’s two sides to the coin, right?
Unknown Speaker 22:37
If you have one side that has the successes on the other side, you’re also going to find challenges. So what are some of the biggest challenges that you guys have been facing recently? I think for us, at the end of the day, when we’re telling people, we’re going to get you more customers, or, you know, you can convert more of your customers, it’s also dependent on
Unknown Speaker 22:56
the kind of traffic. Let’s say we put us on our website, we have had customers who just don’t have enough traffic to their website. So they’re like, Well, I only got like 20 leads this week, and we’re like, well, but you didn’t drive any traffic. So the chat bot was not seen by anyone. So one of the challenges for us is to pick the right customer where this is actually going to make sense for them. I know we work with with a firm that was going to start a new behavioral health practice, and it was telehealth. It was during covid. They’re like, we’re going to do this through telehealth. It’s a new investment for us. We’re going to create a new website, and we need a chat bot. And we built a really great chat bot for them. They never actually got their telehealth practice off the ground or got enough traffic to their website to really make it worth their while, right? So this is one of the challenges that we have to face, is that we have to get a customer that has enough things to contribute to the process.
Unknown Speaker 23:55
And talking about that customer acquisition process, what do you find has been working recently when it comes to customer acquisition. So we actually do not spend a whole bunch of money on customer acquisition.
Unknown Speaker 24:09
We have a B to B sales cycle that we follow,
Unknown Speaker 24:13
and it starts from our own chat bot, bot queen who sits on our website. In fact, we recently had a webinar saying we supercharged her to be a lead generator, right? And here are all the tips and techniques that we use. So we use our own technology to do our own leads to get our own customer. We have a lot of referrals. We have partnerships out there, and we also obviously have a website, and do some marketing around direct mail, and we do webinars, which is a funnel for us, and we get on a bunch of podcasts so people can get the word out. We have PR so we do different articles and contributor articles. We do some industry research studies and publish those. So, yeah, we have an active and integrated martech program within our company to help us scale. Yeah.
Unknown Speaker 25:00
And you think that’s been working well.
Unknown Speaker 25:03
I mean, it could always work better. But we’re limited by the amount of budget we have also, so we do a lot with the limited budget, all right? And when it comes to your actual products, right now, you have this platform for building these AI chat bots. What are some innovations you’re making right now, anything you’re planning on coming out with. You know, we recently launched SMS. We’ve got new personalized dashboards that have been launched. We are working on different kinds of agentic models for different types of verticals. And, yeah, those are some of the ones that I can actually talk about on the roadmap right now. Of course, it’s doing a lot more with video and training on video and voice that’s also advancing because people like to speak to chat bot. So we’re working on all the voice technologies as well. Can you talk to us more about that audio and video technology? Yeah. So the ability to like we I told you how we train the chat bot by crawling a website where you could actually feed in a video, and the chat bot would learn from the video and then be able to answer questions about the video content. Okay, gotcha, yeah. So that’s one of the focuses we have. The other one is to have the chat bot create videos. You know, it’s just like a home improvement kind of bot that we’re working on for this particular case, which can actually tell you by video, here’s how you fix something. So that means we have to go to a vast library of videos, find the right video, find the right clip of the right video, and then feed it to someone when they say, How do I repair my Bosch washer, which is making too much noise, right? So we like, Okay, well, let’s diagnose it. So show it to me. You put your phone there, and you show me the video. Then I’m going to tell you what the answer is, and it’s all be done by another video or a human being chatting with you. So being able to ingest videos to train and being able to produce videos to show you what’s going on. There’s some huge cases there. And then voices, of course, speaking, being able to speak your questions and getting a voice response back so that you can maybe access it while you’re out on the field, repairing something, or you are actually driving, and you still want to know, and you want to just talk versus type anything. So trying to enable those are, those are fairly advanced now use cases,
Unknown Speaker 27:31
and those really supercharge the amount of personalization you can get right, like it’s, it’s almost as if you’re talking to somebody, yeah, once you get there, yeah, that’s right, certainly advancing rapidly, I would say yes.
Unknown Speaker 27:46
And you also mentioned these personalized dashboards.
Unknown Speaker 27:50
What are those? So basically, as a senior living, for example, your metrics that you care about are different versus somebody in addiction, then it would be different as well. So you can actually personalize or customize these dashboards to for each client based on the metrics that you want to track and see. Like for example, for some people, it’s very important to see how many people that I engage with today or this week. And for other people, they want to see what time of the day did people really engage with me? Right? So I don’t want to know what they did in the day. I want to know, you know, what did they do at two o’clock, three o’clock, four o’clock, right? So you can customize, so obviously, you can’t have everything on a dashboard. It’ll get too crowded. So you can customize what’s most important to you as a company, or as the as your role, your individual role in the company, what’s most important, and have that show up on your dashboard. That’s really cool.
Unknown Speaker 28:45
And I know we’ve been talking about baco.ai for a while now, right? But if we zoom out from that, what are some trends that you’re currently seeing in the industry? What recent developments have you seen in the industry that caught your eye?
Unknown Speaker 29:02
Well, I mean, everybody in AI was kind of surprised by deep sea, the Chinese LLM model that was introduced. There’s another one too from Alibaba, which, you know, basically, were very interesting because they were developed with a lot less money, and they’re pretty good, right? So I think that that’s kind of a shake up, like, Okay, what is this? What is we here? We’re spending billions of dollars and a lot of time. So you know that’s kind of interesting to see how much cheaper it can get, right? And also that you can actually do it without this huge expense, the compute expense that goes on in AI, right? The amount of energy it uses to actually produce to do AI is a lot, right? So the fact that you can do it cheaply is very interesting, and how this company did it, and what’s the future of that. So, you know, compute being cheaper.
Unknown Speaker 30:00
These models being developed so quickly and with such advanced capabilities that opens up the remnant of more applications and more usage, right? So at this point for startups, I think that kind of investment that it takes to get a chat GPT may not be that big anymore, right? Because we have deep seek that apparently did it in $6 million so that makes it very, very interesting, right? The application usage of it is growing so more and more interesting uses everywhere, because they always say, like, it isn’t the people who invented refrigeration that made money, it was the people that made coke and figured out how to use refrigeration. So these applications are going to get more interesting and more ubiquitous, right? So those are the trends that I’m seeing for AI agents.
Unknown Speaker 30:55
Yeah, no, well said, and once again, back to the coin. What’s the flip side of that? What are some of the challenges that you’re seeing in the industry? Well, I think that the challenges are that it’s still expensive, and there’s a lot of unknowns out there, and there’s this fear of sentience and like terminators and all that. So ethical AI to examine what’s that and privacy is still important and privacy and sensitive data, and who owns the models and who owns the data that made the models and what those relationships look like. Those are all the challenges that we all have to work through. And what do you think are some ways that we can address those? I mean,
Unknown Speaker 31:36
I think it’s being addressed by companies that are that are building these models or these bots by them actually paying people to use their data and creating those kind of relationships so that there’s a lot more work that will happen in that area. But yes, and shifting focus on that, this one’s to you personally, most overrated piece of advice that you’ve encountered recently. Most overrated piece of advice. Yeah, my God. What could it be? I just remember I talked to, it’s not recent, but I talked to a venture capitalist a long time ago who asked me, like, why are you an entrepreneur? Like, why do you do entrepreneurship? And and I said something, and he said, No, no, what I tell people is, because I want to make money and have fun. I mean that now seems a little bit trite.
Unknown Speaker 32:29
Yeah, it’s okay to make money and have fun, but I think there’s a little bit, there’s more impact beyond that. So I think that that advice is a little overrated. I feel all right? Well, thanks for sharing that. Thanks for coming on. It was great having you on pleasure. I know you shared a lot of really good information. You brought a really unique perspective onto the podcast, so I love that.
Unknown Speaker 32:54
All right. Well, I’m glad you liked it.
Unknown Speaker 32:57
Thank you. Thank you for giving me a chance. As we end today’s episode of making martech, we’d like to thank all our listeners. If you enjoyed this episode, make sure to share this and subscribe to our show so you don’t miss the next one. I’m your host, himnish Jindal, and I look forward to seeing you on the next episode.