Gong’s Amit Bendov: From Meeting Recordings to Revenue AI

0:00I get on a call with a customer, Gong already prepares me for the meeting. The way it used to work, I'd get five people from the account, we'd schedule a pre -meeting, briefing, they'd prepare a document with all the questions, and it's probably like five hours of some highly paid people. Now it's 30 seconds, right? It just means it's a really five, a question I ask. So it's a huge time saver. And we're not even yet at level five driving. So the potential is like way bigger. If you remember like 75 % of seller's time is not customer facing, that's the data. Everything else is the opportunity size.

0:57Hi and welcome to Training Data. Today we're excited to welcome Amit Bendov, CEO and Founder of GONG, which is one of the most beloved companies in the modern -day sales world, and also one of the very first AI native application companies. We interviewed Amit about GONG's evolution from a wedge into a platform, specifically from a transcription tool into a brother revenue intelligence platform, and heard more from Amit on his prediction for what the sales team of the future might look like and Gung's role in building that future. Enjoy the show. I mean, thank you so much for joining us today.

1:31We are delighted to have you on the show. You are one of the OG AI applications and absolutely beloved company in the sales world. And so looking forward to having you on training data. I'm excited to be here. Let's jump right in. I want to start with one of the hot topics in the AI sales world right now, which is do you believe that AI is going to take over the role of the SDR? completely no not in the near future but not with the current technology I do think that a lot of the BDR work will be done by AI and it's a good thing so it's it's not binary like but short answer not in next like few years definitely not if it's a complex outbound SDRs AI is not going to be able to totally replace that got it okay follow up question do you think that CRM will exist in five years time.

2:22In the future where we have a lot of agents running around doing a lot of sales -related activity, what is the role of the CRM in that future? Yeah, I'm sure CRM will exist in 20 years. It's not going away. There's so much built in there, just like, you know, mainframes are still there even though there's like much better technology. I don't believe that managing in managing customers, CRM is going to be at the epicenter. We're moving from a CRM -centric world to a more like AI -centric word. CRM manages customers, their data and a lot of other things, but it's not managing the operations, the activities, and how to engage customers.

2:59So that's for sure. It's still there. People tend to underestimate how much the reason serum there's quite a bit and it definitely will live for a while. I mean in that AI -centric role, what is the role of the software? What is the role of the humans? The role of the software we were trying was starting to understand the role of the human is actually less clear and I think we will be surprised it's hard to predict what it will be. I'm sure it will change roles. There are roles today, you know, prompt engineer as the obvious one, right? That didn't we couldn't even like imagine like three years ago.

3:41SDR did not exist. People ask you would have been right. It did not exist 30 years ago. Like this is like a modern invention and it might go away. So to be like different roles, you know, I'd say in general, the current technology, anything's based on transformer is fit for certain types of work. It cannot replace humans whenever there's like accountability or decision -making or creativity. We're not even close to that. So humans will still be in control, I believe, for a while, but AI can do a lot of the work. So, you know, definitely I could do an amazing job reviewing a contract right now in redlining all the hot topics even suggesting correction, right?

4:23We would not let AI sign a contract or even send a contract or send a proposal, right? There's always like a lawyer that will need to do that. So, I think the roles will shift to humans that are augmented and using AI as a tool for, you know, quite a few tasks. You know, in our industry specific, there is this number that goes around that 75 % of what sellers are doing is non -selling activity. So three out of four dollars that we're paying them is to do admin and non -essential work. All of that could be replaced by AI. That'll be pretty exciting. I mean, take us back to 2015 when you founded Gong meeting transcription.

5:08Why did you start with that? And what was the vision always to build a meaning transcription tool or to build something very different from that? No, actually the vision was never about transcription like so when I started a company we were like my previous company before starting gone We were recording calls and I was using the transcription service that the real reason That I didn't want to read transcript like there's like there's so many of them going on and who has the time to do it? I have a little bit of an ADD as well, so I'm not the person who they're like listening at great lengths or read.

5:48So I wanted something that would take old information and transcend his structured data. And transcription is just a way in a process in the pipeline that you need to capture the audio and video, transcribed, then identified topics and themes and other things and then use it in structured data. So that was the idea. We started with meetings because that was kind of the big black hole. We could kind of see other things that are happening and activity. But this was like a very rich data set that we knew nothing about and nobody was doing anything. So that's why we started with that. I mean, I remember in those early days, there was this period of time where GONG was one of many.

6:36And then it felt, from the outside looking in, it felt almost overnight. All of a sudden GONG was one of one. What did you guys get so right in those early days? I actually think we have like more competition right now. And I always tell the team, the more success we are, we're going to have more. So if you look at, when we started, we thought that we were alone because there are other companies in sales, we didn't know about it. Thank God, we didn't know. But we found that there are like two others. Now there are dozens of companies, and I hope to have like hundreds like in a few years. Like if you look at how many companies are in CRM, there are hundreds.

7:12So if the market grows, there's an opportunity. But that's a very important question. There's never like a moment. There are always moments where people predicted our death or doom or whatever. So, I think, 2017 or something. So we didn't know if we have competition. Like, I started going because I said, oh, like, I have this problem. I'm scratching my own itch. Nothing exists. Maybe we should start a company. And then I found there's another company that was like better funded, like a little bit ahead of us, you know, maybe like cooler technology. And when I realized that, I still have the recording from our old hands.

7:56We're like a very small company at the time. But by the end of this year, there's only one goal that matters. We're going to be number one. Or nothing else matters. Like, I'm not interested in this company. If we're not number one clear leader. So like in SAS, there is no second prize. At least nothing that I'm interested in. And we have to be in. We decide the defined metrics first, how much we're going to win. And second, it's not us. We're going to be the undisputed leader. So we even put metrics like how much media share or LinkedIn or social media share and we wanted to have with two X all the others combined.

8:32That was the goal. Right. And we hit it. And first the gaps first it's like very neck to neck. then you get a little win. It's very hard. And I think we started a little behind. But the strong gets stronger. Once you start opening a gap of like you know, let's say 5 -10 percent it becomes easier. The sales people become more confident. Customer start to see and then the gap widens. And then you get new competitors. So we started. There was like a chorus and executive isn't none of them is an independent company. Then people said, okay, like outreach and sales lofts are going to eat you. When we, before we raised money, it said Google and Amazon are going to eat you.

9:13So there's always like competition, but you have, you absolutely have to win. It's very hard at first, but once you start opening the gap, it gets easier and then you get new competitors. So you're never, you're never alone. If you're alone, you're in trouble. You're an un -arong market. I mean, my view as the, you know, the person on the side lines who was, you know, doing the investment memo work at the time was The thing that you all did that was really special was you won the hearts and minds of your customer and you were really really customer focused And you know, like this was at the time meaning transcription was kind of new -ish like the a lot of the, you know Models weren't weren't that say they aren't yet And so a lot of the other companies were thinking technology out and you were really going customer back and you knew how to win over those hearts and minds.

10:02And I think to this day, that is the thing that is so special about Gong. You go talk to any sales or app sales team. It's like it's Gong, Gong, Gong. And they get Google AI's when they think about you. Absolutely, that's a great point, Sonya. Thank you. It's not that the fact that you want to win doesn't win the battle, right? You have to do it the right way. And we're always looking at customer problems that we can solve better than anyone else, right? And always looking ahead. And our goal is always to be almost like every two year to be a new company, right? I'm a big Beatles fan, sorry, that's my age.

10:41And I told the team, I always tell them that I want to be like the Beatles like every two years. It's almost like a new band, totally different kind of music. And I said, if we don't do it, we're dead. So if all we do, like, meet in recording today, we'd be dead because like, oh, that's like everybody can do that. So we always have to reinvent, take great care of customers. And people think the competition is like a zero sum game, almost like a football, like bowl possession and all that. It's more like an Easter egg hunt, right? If you see, don't know where all the other kids are running. Just find the ones that are your own that nobody's seeing and get as many of them as you can.

11:21I just watched my toddler try to do his first Easter egg hunt. and he didn't know it was paying very many of those. So when you started the company, this was a pre -chatchy PT era at least, and there weren't really great foundation models to build upon. And so back when we invested, you had very state of the arts, kind of model research lab in Israel that was doing a lot of speech to text and model technology. To what extent was that a source of competitive advantage for you then? and is it still a source of competitive advantage for you today? Great question. So we went back and forth on these, I think like two or three times, that I'm still not sure we've seen the last chapter in these.

12:02So when we started, we did not have our transcription engine. We're using just like a web API. I was at opinion now, we should have one. It's so core to what we do, but Elon, my co -founder, the smarter half said no, like it's, well, like six million, I don't want to like invest in something like it, let's, you know, 3 % more accuracy. Let's see what people want to buy first, develop the app, and then once it raised more money, we developed. So we started with something that's really not, it was an okay service, just duct tape and got to the application and got the first thing that actually works.

12:37And we saw what was important. Our competitors were investing a lot in like real -time transcription and all the cool technology that never made like a real difference for customers. So I'm glad in a hindsight that I listened to it alone and we didn't burn our Serious a money on a transcription ended, but then once we started You're right at that time we're ready to rely on things like nuance, right? Which was like very expensive and people don't even know what it is today, but and and not great So we decided to now now we can invest we started our own team. We hired some of the best speech people like in Israel, there are quite a few of them, and developer own, and it was like better quality and a lot cheaper.

13:26So that was a competitive advantage. And now we used anything from some of the Amazon services and Whisperer for some of the things, we still have our own models for some languages where the others don't do a good enough job. But if they do, we'll be happy to switch to service like our transcription is not our core comp, we're very good at this, but we're at the AI application layer. This is a very low level that we do it. It's like a car battery, right? If we can't find something, we'll develop it, but if we can find something better and cheaper, we'll use it. I'm glad you brought up real -time transcription because that was one of those.

14:12I remember, you know, from the technology out, that was one of those, like, of course, going to go in that direction. And I think you like stuck to your conviction on that's not what our customer wants. And so even though that may be the cooler technology, like this is, like, what our customer wants is, you know, actually coaching workflows and the ability to, you know, analyze competitor mentions after the fact. And so, like, that, I think is a great example of thinking customer back versus technology out. Yes, yes, absolutely. We have like, since we started every time, we're always asked about real time, which is like, you know, it makes like total sense, right?

14:43Except that the technology doesn't really work. That these day and the value is questionable, right? So we still have not seen it like in action, like in a great way. I want to talk about generative AI. And so you started this company back in the days of what I would call predictive or analytical AI. And a lot of the value is in that. how many times your competitor is being mentioned, if you talk about this feature first, or do your windwights go up? And I think a lot of the value is in that. Tell me about what you see as the opportunity for GONG in a kind of generative AI technology era. Well, I mean, these are exciting days.

15:24And when we started a company in 2016, in 2015, we started to think about it. We, the model that we use is a self -driving car, right? Autonomous system and self -driving car of like five levels. Level five being like fully self -driving and level one is kind of like alerts of sorts, right? And we envision an autonomous application that runs a revenue organization that you tell what my revenue goals is and any to understand how to break it down to goals and how to help from the leaders to the individual sellers how to run the things. That was like total science fiction. The technology didn't exist back then.

16:06It still doesn't fully exist. There are still components that are missing. But every time we're able to, every time we assess, what's real today? What's available? What's the state of the art? Not the hype. What can really do reliably and create a great customer experience that we can stand behind and integrate into our solution and Generative AI was a game changer. Things like summarizing calls is something that we wanted to do in 2016. I actually hired a TV news editor and I wanted to see because we wanted to take, okay, take like 60 minutes and create like a 30 -second news reel like you do on TV or sports.

16:56right? And I paid someone, I gave a recording, I said, okay, wanted to create something for me. And we did probably like 20 attempts and I gave up because they're not good. And you understand it actually summarizing, you already, it's already biased in some way. So now you could do it like super easy. So it creates like a ton of value. Some of the holy grail apps that people ask for, you know, tell me what's going on in these accounts. So now I could just ask, like, we're at the end of our quarter right now, so I could log in and just say, hey, like, how is do you steal progressing? Like, what could make it go wrong?

17:38Like, can I help with anything? I'll tell you, it's pretty amazing. On the market side, it provided like a huge tailwind for us. So we were grateful, not just for the technology, but what it did in market awareness. So when we started, GONG is an AI company or an AI application company more precisely. But we never spoke about it like it was like third level messaging that this is like AI technology because it was scared the heck out of people. So especially we sell to non -technical people, we sell to CROs and business users that aren't super technical. So if you talk about AI, that seems like scary and science fiction.

18:21So it was like third level messaging. You had to double click like at least a couple times But now we got a lot of people calling us hey my CEO Was asked on an earning call if they Plan to increase or harder going to plan to increase sales productivity with AI Do you have something with AI so there the openness in the excitement in? Chat GPT made it so Accessible at a consumer level that's super exciting. On that note, I think a lot of companies, boards are focused top down on how AI is changing productivity metrics. And I think encoding, for example, it's been pretty exciting to see a lot of companies are now reporting to wanting to 30 % increases in engineering velocity and metrics like that.

19:13What have you seen on the sales side? What has been the impact of using AI for your customers and the revenue orgs? We've seen dramatic changes. So some of our customers have seen as much like 60 % more sales capacity. And just by, you know, you can measure it like lots of different way. Let's say, you know, a person can handle a 50 accounts. Now they can do 100 accounts, right? So that's like just by saving all the time. So a lot of the tasks are just becoming redundant, even like you don't need to talk to as many people anymore. Not there's the fun part of talking to other people, but not like what happened, what I'm going to do, because now everything is visible.

20:00I get on a call with a customer. Gone already prepares me for the meeting. So I don't need to call the county immunie. The way it used to work, you know, I'd get like five people from the account, we'll schedule like a pre -meeting, briefing, they prepare a document, all the questions, and it's probably like five hours of some highly paid people. Now's 30 seconds, right? It just means to read if I have a question I asked. So it's a huge time saver. And we're not even yet at level five driving. So the potential is way bigger. If you remember, 75 % of seller's time is not customer facing. That's the data.

20:43everything else is the opportunity size. What level are we at today and what's the path to getting to level five? It depends. So they're like they're different components like for example at Gauntz, a summer at level two and summer level three and summer at level four. Level four is almost like driving. So if I you know let's say I get on a call with a customer, Gong would all create like the email for me, but I have to click send. So this is like level four. It's almost like I'm in control, which is smart. That's level four. Level five is when it'll send the email on my behalf. It'll take it, it'll take a little bit of time.

21:27Some things are easier than others and something are not as mission critical, right? If I, if you want to do people, everybody talk about like outbound SDRs, right? Coldcoin. This is a pretty poor choice for current state of AR because first it's hard, it's very hard for people to do it. Second, if it goes wrong, all right, it's very damaging to your brand. As we've seen, if you've seen some of the cursor service agent, you know, changing privacy policy or airlines, so I wouldn't let like AI talk to my customers, except for very simple cases, like support tickets that are pretty confined, but not anything that is complex.

22:16So for those, it's still gonna be level four until we get a conviction that technology can be 100 % trusted, but some areas are, if it's not, now we have role -playing, right? AI, you could just practice, do practice call with AI, and it'll give you a feedback. All right. If this thing doesn't work, 100%, it's no big deal. Like, yeah, a little bit annoying, but you're not embarrassed in front of customers. Is that with a virtual avatar? Yes, yes, yes. Oh, super cool. Yes. So the coming on until it's 100 % trusted, do you think people, or I guess in terms of what you've observed with your customers and the users of the product, do people hold AI to a higher standard than they hold themselves?

22:59I would guess most of those reps or BDR's or SDR's are not 100 % all the time. Yes. And no. So the yes, actually fun fact, like when we started going, we actually did not show the transcription. So you couldn't see the transcription was there, but someone advised us like a smart person that they did like another application. I guess don't show the transcript, because it's 70 % accurate, which means like three out of 10 words will be wrong. When someone reads that, it'll just be an awful user experience. So we heat it, we allow them to search, we show some things, and people are super happy. But we couldn't show this thing because it just wasn't good.

23:42So people do expect perfection. With AI, you know, someone, a self -driving, a Tesla makes an accident, right? It's major headlines, right? But, you know, people make so many more accidents like, and nobody talks about. That said, I do see that the reactions to, for example, when people ask questions at GONG or they get the summaries or they get, everybody's very excited. They know it's not perfect. Like, it's, you know, maybe 80, 85 percent correct, but the value, because it's so fast, right? And so easy. I don't see a lot of people complaining. So I think people are able to tolerate somewhat lower quality for ease of use.

24:28So for example, when I was a teenager, we had high -fives systems. That's how you listen to music. And to make people either on headphones or on mono, just one speaker. How can you listen to that? Just poor audio quality. But it's so easy that people are willing to put up with simpler and easier technology. And this is definitely a lot simpler. So we see generally the sentiment is very positive. I have no idea what the hi -fi system is. I bought my wife wanted to kill me, but I bought major speakers in our basement. And she said you did what? Yeah, but it just means that audio quality. Love it.

25:11I mean, what's your take on agents? and how are you guys approaching agents building selling agents at COM? We think agents are real. They're, again, like a lot of other things. There is a bit like inflated expectations right now. It's like it's going to do everything. So everybody like, this is me. Agents aren't taking your job anytime soon. But they can definitely take parts of your job. The parts that you don't like doing and it's a waste of your time. So the approach that we're taking is like task -specific agents. They're not developed like do anything and there's no writing code. So we have, for example, like we spoke about like briefer, right?

25:54So here's like brief me for encounters. There is an agent that trained for these tasks. He does parameters and it's built in the workflow, right? So that's and we have dozens of them. dusty approach that we believe are great for enterprise applications. General purpose are very hard to nail because they could go in all sorts of directions, but it's a real thing. They're also interesting, the line between simple automation and agents like isn't clear. I feel it, some of them, it's the same thing, but it's also very different. So the question, can they make decisions? Again, for non -critical decisions, decisions that aren't like a huge deal, they can definitely do a great job.

26:41But agents can definitely do quite a bit of our work. Have the newer model capabilities, and in particular, the reasoning models have those unlocks new capabilities for GOM? Not yet. I think that there is a line that I'm not sure that even like some small -merc technologies is capable of which is like can you rely 100 % of something that will be true right just truthfulness like not 99 % 100 % right so there is a category of work that has to be you know you send a proposal to a customer, that can't be like 99%. If it's like 1 % of the cases, then you send something as like totally wrong pricing or over discounting or something as like a hallucination.

27:34So even with the reasoning, it's getting better. So I think it increases the classes that have tasks that are that make sense. But if you talk like AGI or things that are absolutely like a must happen, you know, calculate commissions to sellers, right? You get it wrong, it's very bad, right? So I think it's a quantitative improvement, still not a qualitative. Qualitative is 100 % trustee. And that will be some would say, even if it's like, you know, Yan LaCoon or David Doe, they'll say like it's not even with the realm of transformative technology because of the statistical nature of the approach.

28:19Do you agree with Yann's take? Yes, but I also know there's a possibility that maybe he's much smarter than I am that we're not seeing that someone would prove us for on which will be great. But yeah, right now, there's also an issue of accountability and both legal, but also from a work perspective. For example, Gong now has the ability to update all opportunity records, for example. So that's a no -brainer, right? Nobody needs to log in and see a reman update fields anymore, some companies still want like the seller to say, yes, yes, yes, yes, yes, because otherwise they can blame AI for, oh, I didn't say that, right?