Why Your Competitors Are Already Ahead on AI with Dvir Ginzburg

You're now watching B2B Vault, the Biz to Biz podcast hosted by Allen Kopelman, powered by Nationwide Payment Systems, the number one solution for payment processing. Hey everybody, welcome to B2B Vault, the Biz to Biz podcast sponsored by Nationwide Payment Systems. And today we're going to dive into the world of generative AI with Tavir Ginsberg with his company Insight.io. We'll put the link in the show notes so you can find his website and everything. And we're going to dive in, talk about AI, what's going on with AI, what can businesses do with AI, what his company does. And welcome to the show. So tell us a little bit about yourself.

Hey, Allen, how's it going? Good. Thank you very much for inviting me. And this, we're CEO and founder of the company, PhD in computer science by education. I've been building conversational AI agents at Microsoft and Facebook beforehand, before it was Meta. And we are building the generative AI layer for businesses-to-customers interactions, whether it's emails, chat, and of course, voice. Basically being able to either aid or augment conversation using GenAI.

Okay, so explain like how does that work for like a business owner?

Of course. So any business owner today, or at least I would say 95% of all small and medium businesses, have some sort of client relations, right? And these client relations can be over the phone, emails, chat, sometimes over the website. Clients sometimes fill in forms or just need aid with some sort of service or product that they have been sold. Now, these type of communications, that's a huge time drain for the businesses. And also, as businesses scale, making sure that the quality of these interactions actually stay in par with what the business owner expect them to be. And later on in today's discussion, I will give some stories that every time that I recall these, I'm amazed by how we are sometimes concerned with the quality of GenAI, but don't remember that sometimes the quality of the service that we provide today when we offshore it or when we pass it to someone else and we are not doing it ourselves as business owners, is not that good as well. So what we do, but GenAI in general is headed toward, is how can we take all these customer-facing interactions that businesses have and either augment or replace those using GenAI, whether it's in after working hours, in the mundane day-to-day interactions, or even when it comes to doing proactive servicing and proactive sales. So for example, outbound calls, whether it's cross-sell, upselling, seeing that something changed in the account and then calling. And of course, that would be a GenAI agent. And again, I can only assume that at least 30% of our listeners today have already spontaneously had a call with a GenAI agent and either knew they were speaking with a GenAI agent or sometimes didn't even know. They are speaking with a GenAI agent. So this is not the future. This is already the present. We are having these conversations. GenAI agents are there. And now the question is, how can we make them in a way that is not simply a one-off? They are doing the call center system. They are doing a very specific use case. But actually, in every conversation that we have with end customers, we have GenAI either helping us or doing it for us, for us to have more time to do other stuff.

Yeah, one, one thing I see quite a few companies doing, and not enough of them, is with customer service. So we have a few softwares that we sell, and a couple of the companies got really smart, and they— so when somebody has a customer service question, If they call on the phone, sometimes it'll say, oh, go to the website and click, you know, the frequently asked questions, and there's an AI there that tells them. Sometimes they get the answer, and once they start doing that, we've noticed that call volume by our customers calling into the help desk drops significantly. Like, one company calls dropped almost 50%, and that's a huge savings for them answering simple questions like, how do I add an employee, or how do I set up a recurring transaction, things like that, you know. And we've even added into our software video, so each thing, each screen somebody's on, it'll say, oh, are you not sure how to use this? Click here. And that's, you know, they're looking at a video versus calling us on the phone and waiting and or getting a callback. Everything's moving faster.

Yeah, yeah, 100%. And again, just to make it very clear, today it's not only chat, it's also voice, right? So the voice systems today are so good that you are able to have a natural interaction, voice conversation with GenAI agent that will feel as natural as speaking with a human. And they can— I hope I'm not frightening anyone, but they can giggle and they can pause and they can sound frustrated together with the client and give the experience that they are actually with us as we go through that either challenge or purchasing acquisition process.

Yeah, I've definitely seen like service companies, like air conditioner company, you call up and instead of leaving them a message now, you're getting an AI agent and the AI agent's going, oh, is this an emergency? Blah, blah, blah. Or even if you fill the form in on the website, you're getting a call back like in 2 seconds from the computer. And they're talking to you and you say, yeah, this is an emergency, my air conditioner broke. Don't break air conditioner. We're having super hot weather today in Florida, and so we don't want to wish— say something and it comes true.

But you know, you get these calls back, but if it will happen, they will answer in 2 seconds, you see.

Yeah, they answer right away. And companies that are instituting this are seeing, you know, better conversion of sales and things like that. And I know a lot of like large ticket companies use these kind of things too, like somebody selling a generator or solar. Someone fills out a form on a website and instead of it going to a call center in the Philippines, it goes to the computer. Computer's calling back. I think it's, I think it's good. I don't think it's bad.

Yeah, 100%. I, and I think, as you said, I think my responsibility here in today's call is to say that not only that it improves support and it generates operational reduction and cost savings, it also increases sales. Now, I will say that especially when we deal with more complex type interactions, you need to do more than just having the FAQs, as you mentioned beforehand, to actually generate good and quality experiences. Because when you are dealing with sales and, you know, we are all at the end of the day, we are all salespeople, right?

Right.

We're dealing with sales. You need to be able to understand what works and what doesn't and what works and what doesn't. That's a type of expertise that you need to be able to distill because, and you know, we work with many either auto lenders, service providers, banks as well, and we are sometimes speaking with debt collectors that are doing debt collection for 19 years. Simply reading a PDF or a form, zero chance you will be as good as these debt collectors. Right? And this is why to be able to do sales properly, and this is the foundations the company was founded on, you need to be able to not only digest PDFs, forms, you know, many chatbots, they scrape the website. That's not enough. You need to find a way to take conversations, real interactions between the business owner, the customer support representatives, the tellers, the salespeople. And distill the way they do it into your agent. So again, hope I'm not freaking anyone, but our agents, they sometimes they'll tell jokes that the business owner is telling over the phone when they are pitching. So we are able to understand what works and what doesn't in these conversations. And these will be the tools and techniques we will use as our agents speak. Because if you want to sell, if you want to convert, if you want to upsell or cross-sell, you need to have that type of natural— not only natural sound, because natural sound has become commodity already. You need to be able to be natural in the way you interact. And for that, you need to understand what good really means.

Excuse me. Yeah, it's very, it's very important. You know, when you're dealing You know, when you're making sales, every business— and I always say this— every business involves sales. There's no business that doesn't involve sales because you're ultimately selling. And I learned that when I was very young and I was a chef. And I'll never forget, I was a chef of a hotel and my boss said, oh, you got to come to the catering office to help us with the sale with the client. I was like, what? I never— I was like, the chef doesn't have to make sales. But it doesn't matter what, what job you're in. What business you're in. Sales is the activity that generates money. And then on the other end of that, you have to have good systems so that you can accept payments. Because, you know, the big talk is now about like also about agentic commerce, and sales is the beginning of the— of that journey.

100%. So, you know, when it comes to agentic commerce, you have 2 sides of the same coin. One, you have— and this I will say, I think that we are still a bit far, but we will get there very soon. Mastercard, Amex, all are already playing in that court— complete agenticommerce. And agenticommerce is basically the concept of I'm telling my chat I'm looking to buy a new surfboard, uh, this is how tall I am, this is my budget, just do it. And 2 weeks after I have it by my porch. And that sounds science fiction, but we're headed there. We are headed toward the place where the chat already has your credit card information, just like Google has, for example. And once you give it the right constraints, it go ahead and does even the acquisition for you. Parking that aside for a second, what is much more popular and already happens today is clients, humans speaking with AI agents over the phone, providing them credit card information. And that's something that we already do. So that, that brings a new set of regulations. So you need to be PCI compliant. You need to be able to mute the conversation when you give the details, all that. But that's all— that's not future, that's present. So everyone that is listening needs to know that from a compliance perspective— and I think that's one of the only moments in time where regulation was able to overcome or be fast enough to put the right guardrails in— business owners today can have AI agents collect credit card information, and that would be compliant, of course, if you do it correctly. And then go ahead and trigger the right systems in order to trigger the payment. For example, if they use, you know, Stripe or any other method.

Yeah, they have to ask questions like, is this compli— is your system PCI compliant? Because if not, you're going to have problems. And also ask about the encryption, making sure that this is an encrypted transaction and it's going to the, you know, going right into the gateway. 'Cause I've seen some systems where I've been approached like, oh, we can take an order over the phone for a restaurant, right? And then when I started asking questions about PCI and encryption, there was like a little pause in the conversation. I was like, listen, you know, if a customer of ours gets caught with something like that and anything happens, the fines would be like tremendous.

Exactly, 100%. I think that's exactly the point. You know, people are, people forget that being PCI compliant is hard even before GenAI. So many systems weren't PCI compliant before GenAI came and many people, many companies decide to not be PCI compliant after and then that's their problem. But with GenAI, you must be PCI compliant because the credit card information passes through so many points, the transcription mechanism, the LLM, the GenAI itself, the text-to-speech, the databases. So being PCI is a hassle, but some companies have solved that. And that's important to know. So, you know, when we were taking credit card information for the first time, I was so happy because that is in the front line of innovation. And the fact that we have calls— I can send you a few afterwards, you can play them if you want to— that clients don't believe they speak with AI, or they thank AI in a way that they say, you helped me so much. That's exciting, right? Because these clients, they wanted to partner with that business and they couldn't beforehand. And I can give you the 180-degree perspective where, again, we start every partnership with a new business with something that we call interaction mining. So we have more than 250 connectors to Google and Outlook and, and chat and all the other systems and of course voice systems and telephony and all others. Where we take these interactions, existing interactions between humans and tellers, and we analyze the current state. And again, you would be surprised how we sometimes find conversations that the business is embarrassed by. Tellers not speaking in the way that you want them to speak. And we even have one client— I hope I'm not disclosing by details, but they were offshoring their BPO. They were offshoring their call center, and the tellers in the call center, they didn't even know the name of the business. So they got the business wrong. So let's say that the business name was P2B Vault. They would answer, say, hey, this is B2B Safe, how can I help you today? So, and you know, that shows the poor quality of sometimes having human tellers, whether it's offshore or just lack of experience. So this is why we really like starting with interaction mining to let the business know where they are right now.

Yeah, there was actually some legal information that came out recently about US companies giving access to their databases to offshore companies, and what type of liability are they opening themselves up to by allowing people in another country to access their systems in the U.S. And I was like, wow, I couldn't believe it. And I was like, that's pretty crazy, because a lot of businesses have virtual assistants and things like that, and they're into the person's CRM, and you don't know how secure that person's computer system is halfway around the world, and it's kind of dangerous.

Yeah, yeah, exactly. And I think same thing with Gen AI compared to customer service that is offshore. Yes, AI is, is scary, but Tellers that you don't know can be scary as well. And at the end of the day, you want to be measurable. You want to have leading indicators to whether you are performing well or not. And luckily enough, I think another advantage that you have with GenAI, and again, I'm speaking maybe on the bigger end of B2B businesses, but let's say that you have 10 tellers or even 5 tellers, and you want to start to understand why they are failing, right? And many of these salespeople, these tellers, they don't write in the CRM why they think the sales dropped. They just, you know, they mark it as a drop lost, right?

Cold lead.

Yes, exactly. Exactly. The thing with GenAI agents that you can actually aggregate what is going on in a much better way. When we have 100 calls a day that, that failed, we know why, because the GenAI is not only doing the call, the GenAI is also writing the analysis and then aggregating, understanding why, whether it's the product, the offering, the price, the terms, the way we manage. So we have that continuous improvement piece, and if something changes The same day, the business owner can train or alter the requirements or the specs to have a different script the next day.

Right. So if someone's, like, just getting started with AI, because not everybody's using AI today. We use it for a few things in our office. And, and so We haven't really jumped into the part where we're providing support, a little bit of support with it, but not majority, because a lot of things need to be done over the phone when it comes to financial discussions. And, uh, but how many calls are you getting a day, Allen Kopelman?

How many calls are you getting a day if you can disclose?

We get probably about 10 or 15 calls a day. But a lot of— because of when, when we jumped to the— when we jumped from— I'm in payments 25 years, so I go back to analog telephone lines, you know, and now we're at, you know, the internet and EMV chip. So we jumped in 2015 to the EMV chip and encryption. A lot of the telephone calls that used to come directly into our office now have to go be routed through customer service because These machines and a lot of software that we use is all tracked by serial numbers, right? Or their account number. And because of that, we also don't have access to these databases that used to show us like almost like we never saw the full card number. But you could get access to the full card number. Now that's completely shut off. So unless you're Level 1 PCI compliant, You don't get access to that. So a lot of, so, you know, so we get more sales calls than anything or people that fill out forms and then we call them back. And obviously, you know, AI would be great to like call all these people back. Hey, you know, blah, blah, blah, you know, call them back on the phone. But if you're like a small business owner, you know, like we use it for SEO, we use it to help with writing blogs, writing emails, developing, you know, drip campaigns and different things like that, you know, A/B testing. That's this little simple stuff that we started doing even like 2 years ago. And, but if a business owner was getting started today, what kind of things would you recommend for a business owner to start thinking about?

Yeah, right. So I'm putting the hat of customer-facing interactions and needless to say, there are so many more opportunities from automating internal processes to HR, to marketing, to productivity. Of course, GenAI can and should affect every segment of your business, but from a customer service, customer interaction, sales perspective, everything that is customer-related and customer engagement-related, I would start with analyzing, and I think the key takeaway or the key lesson for me, and, and that's after doing it from with many businesses, start by being able to export your emails, chats, and phone calls. Because whether you use Claude or OpenAI or or Gemini, speaking with them. So first of all, uploading, and that's fully secured, especially if you have the enterprise license, that lives— they don't use that to train any models. When you upload these conversations, calls into Claude, it is able to give you such high-value insights into whether, where, and whether you should even start. And, you know, I'm coming from a position, right? I'm not objective, but there is a chance for Claude. And of course, our interaction mining might be a bit more complex, but the basics, it touches on the same pains. The LLM will be able to tell you whether and where you should start. Because there are use cases and flows that are more powerful than others. There are use cases and flows that are easier to implement. For example, the non-actionable, non-personalized. And the LLM will be able to easily tell you and inform, for example, 85% of all of your service use cases you can do with AI agents. Amazing. You now know the potential. Or it might say as well, well, 75% of all of your use cases are actually personalized and actionable. You need to expose an API, an endpoint for the AI to fetch the data. Now, for some small and medium businesses, that's just something that they don't have yet, right? Asking them to expose their CRM for an API, maybe they don't have it. Maybe from a security perspective, they can't do it. So start by being able to export your conversations. It will give you a very good visibility into what's going on.

Yeah, there's so much. AI is like, oh, and it's getting bigger and bigger every day, you know, and people are going to, you know, need to put it to use in their business, you know. And, you know, I always say crawl, walk, run, you know, 100%.

And then this is why I'm saying, and crawl, walk, run, and crawl, walk, run toward the right direction, right?

Yeah.

So don't do Gen AI just to do Gen AI. I'm probably the best salesperson ever right now, right? But please don't do GenAI just to say that you're doing GenAI. Do GenAI if you know and understand what is the value it's going to provide you, right? So I see so many people buying software to help them write better emails or help them craft the best LinkedIn post. That won't do the right— it won't do justice with you, at least today. Maybe in the future. But people are still looking for that personal touch there. So, and this is why I'm suggesting starting with exporting your existing interactions with clients, because LLMs will be able to help you and help you understand whether you have potential, because maybe you don't have any calls and maybe your calls are so complex that AI cannot fix. But you won't know. You need to be data-driven.

Yes, definitely. Everything's in the data. But, you know, from a simple standpoint, like hooking up a voice AI up to your forms, somebody fills out a form, and even in the middle of the night, they still get a phone call.

Yeah.

Yeah. After you're closed because you're on the East Coast and they're in California. And it's 5 o'clock here and it's 1 o'clock in the afternoon there. And then, you know, you have it notify you, hey, somebody called, somebody filled out a form, and then you can call, you know, the AI gets their information, texts it to you, and or something, and then you're able to act on that. Because when you own a business, owning a business is not a 9-to-5 job, it's 24/7.

Yeah, that's for sure, that's for sure. And luckily now you have people that don't sleep, right?

So if you had to give business owners, you know, 3 or 4 pieces of advice about AI, what would you tell them? Like how to get, how to get educated, how to get more information, how to, how would they get started, different things like that. What would you say?

Yeah, cool. So Only two. You asked 3 to 4. I, I'm condensing it. I'm, I'm, I hope two that are as good as four. One, if you are using Google Calendar, connect your Google Calendar to either OpenAI or Claude. I'm, I'm not promoting any of these. They both have what is called a connector, and your mind would be blown away by how good it, it understands where you waste time. So it is as easy as connecting Claude to your Google Calendar and then asking it, and then telling it, for example, I'm a business owner doing ABC, please tell me where I waste most of my time and what should I do to improve it. And the insights that you're going to get are, are mind-blowing, you know. It's like someone putting, uh, facts and reality in front of you, telling you as a business owner you should spend more time here and not there. And the second, which I think relates to the first one, and I touched on it beforehand, start with the data. It's the easiest thing. Gen AI is like a huge hammer right now, And everyone are looking for screws and everyone are hammering. You know, I have GenAI, so I will do everything with GenAI. People are trying, you know, build podcasts with GenAI. People are trying to build— it's, it's an amazing hammer, but you need to understand what it actually fixes or does. So understand your business well, connect it to the LLM. Write the right questions to understand where value resides. We are, and you know, it needed bravery to do that. But we today, we are not starting with a client, even if they want to partner and they want to run fast. We don't start with a client before we go through the interaction mining process. And that's time, right? And some of the clients are telling us, no, but exporting the calls is very challenging. We're saying no. If we are not going through the interaction mining process, there is zero chance we will actually understand your business. So there is zero chance we will actually be able to provide you meaningful value.

Yeah, you got to start somewhere. So you have to start with the simple things and then go, go all the way through the process to figure out where the best use case is in your business. To, to use AI. Another thing, one of my big pointers I give people is learn how to prompt and ask the AI the right questions, because AI is going to give you the answer based on what you're asking it. Not good at guessing, right?

Exactly.

All right, well, thank you for joining us on the podcast today. Let people know where they can find you.

Of course, the easiest route would be my LinkedIn, Davir Ginsburg. Of course, it will appear in, in the podcast as well as just shooting me an email. I will make sure to attach it here as well. That, that would be the easiest thing, trying to be as responsive as possible.

All right, so you can find him and also his website is insait.io. And thank you for joining us on the podcast today and talking about generative AI. This is a great conversation. And, you know, we touch on AI a lot on the podcast, and it's great to have experts on to talk about exactly what's going on in the world of AI. And everybody have a great day. Peace out. And as I always say, carpe diem, seize the day. Thanks for listening to B2B Vault.