Episode Transcript
Kyle James (00:01.074)
Hey, welcome to the AI Chronicles podcast. I'm your host, Kyle James. Today we're going to be diving in head first into how a sales engagement platform called Salesforce is using AI inside of their own business. And we'll share the exact steps that you can take in order to implement AI for yourself. Now, before I dive into that, listen closely. Are you looking to implement AI inside of your own company or maybe just struggling to get your AI to stop hallucinating? Speak to GPT trainer.
GPT Trainer literally builds out and manages your AI agent for you, eliminating hallucinations for good. Go to gpt-trainer.com. I promise you, it'll be the biggest time-saving decision that you've made all year. Trying to set up AI on your own is like trying to build a house from scratch. Sure, you could do it, but the time and frustration is gonna take you to get it finished just isn't worth it. It's a thousand times faster and safer to hire professionals.
schedule a consultation today. Once again, that's gpt-trainer.com. To have with me Frank Saunders, who is the CEO of SalesForge. Frank comes with over a decade of experience in B2B sales, working for companies like Google, SimilarWeb, Black Crow, and WaterGraph. Whether as an individual contributor and all the way leading large sales teams, currently...
Frank is building SalesForge.ai to give companies the ability to blend their current sales efforts with AI agents to maximize pipeline and drive down costs. In less than one year, they have scaled from zero to three million in annual recurring revenue. Awesome, awesome, love this. Frank, welcome to the show, my friend. How are you doing today?
Frank Sondors (01:40.214)
Hey, thanks for having me, Kyle. I'm doing pretty well. Quite a busy week, even though it's the summertime, yeah? And I would expect, you know, things to quiet down, but they're not so, yeah.
Kyle James (01:51.538)
That's always a good thing when you expect like when you expect it to be like nice and chill and it's like actually no It's doing really really well. I think that's that's fun for everyone You know, I mean so tell us a little bit for you know, it's to tell us about yourself and like what exactly is fell sales Forge and how did it how did it come to be?
Frank Sondors (02:08.494)
Yeah, so in my previous gig, I let a sales team of 50 and this is, you it was back in the days when it was growth at all costs. And I thought to myself, damn, you you typically, you know, if you're being the sales or being the sales leader, you know that for every 10 salespeople that you hire, only one person is natural. Two to three people can be trained up to be somewhat good as a person that's natural. And then 60 % is what I call dead wood. And I always thought to myself, damn, like all these people in sales, I bet I could do significantly more output with significantly fewer number of people.
And so I've always thought to myself, how would the process look like? How would the software look like? What would be my stack to do that? And this is where, due to my own personal frustration, due to my own experience being a VP sales, I thought to myself, I have to put an end to this. So I found a couple of really smart engineers through YC Matchmaking platform. And then we started to build Salesforce a couple of years ago with initial product called Salesforce. But we built other six products around that with the main premise of helping companies to hit their pipeline targets, but with the least headcount possible.
So it's quite a complex problem to be solved. So it's not like where you can only solve that through one problem, but you actually need to build these multiple different other softwares that help and sort of make the core software sales which significantly better. So yeah, we scale from zero to three million bucks, profitable, 40 employees right now from three co-founders to 40 employees, still keeping lean and mean. But the reason why we sort of lean and mean and profitable is because apart from having also our own agent, which is our pinnacle product,
that can literally blend with your existing sales team, we also run a multitude of agents internally for our own business. One example could be when you would sign up to our website, salesforge.ai, if you don't end up buying anything from us within seven days, then our own agent, will call you up. Literally, he will call you up and say, hey, Kyle, how's it going? I saw that you signed up a week ago.
What was your experience like? Was anything wrong? Is there any reason why you end up buying? And we'll try to book, essentially, a call with you. So we also do use a lot of AI agents that we don't necessarily sell. And by being always agent first, with every problem that you have in your business, unlocks so much efficiency. And that compounds naturally. So we have one guy right now that's building all of these flows. So we use an automation software called NA10, where we literally use as a base to build a lot of the
Frank Sondors (04:27.63)
automations, AI agents, et cetera. But our own agent, what it does is, so agentfrank.ai, so you go there, you can literally receive a unique email that will be personalized to you. But what it does is essentially, after you outline your value prop, so essentially what is it that you're selling, who is this for, what's the ICP, what's the pain that you're solving, we then go and source leads for you across multitude of databases in the world.
We then will craft a unique email to every single recipient. So there's the typical three emails that we would compute and would deliver. And then if the prospect replies back, we would have a conversation back and forth. And then the agent is optimized towards a particular conversion that we're looking to achieve. And that's typically meeting book. So you can hook up your calendar link or HubSpot link, whatever that is. And then the agent will try and achieve that goal. So today, the average performance of an agent is about 2%, 2.5 % reply rate. And that's email only. And that's outbound. That's not inbound.
And then from that, there's about 10 20 % positive reply rate. So that means one or two in 10 replies are positive when you're getting them back. So that's what an agent could do potentially for you. Naturally, it varies. You could sometimes have 10 % reply rate because the offer is goddamn good. And sometimes the reply rate is also 0 % because you simply either don't have product market fit, you don't have a product channel fit. So there's also reasons why naturally agents don't work. If you think about it, it's just like running a business. So when you hire reps,
Sometimes, you know, it pans out to be very good because you have a strong product market fit and you just apply really good reps and then things just pan out. But sometimes you have a really bad product, right? And then things just don't happen. So that's how you also have to think about these agents. But the beautiful thing is that you can blend the agent together with your existing Salesforce, right, that you have within one software environment. So you can blend these capabilities. So we're big believers that the whole economy, not just in sales, but also in finance and other spaces.
is moving towards augmentation. So you will see reduction in workforce, and you will see a lot of agentic capabilities and automations coming into play. I would say agents are very good on what's called the email front, and also maybe on the social media front. So you could use the agents to communicate via LinkedIn, and that also works fantastically. But then what agents can't do, because it's illegal, they can't do, say, call calling. That's illegal in the US and Europe, for example.
Frank Sondors (06:44.91)
So that's where naturally humans are coming into play. So our sales rep naturally are calling our prospects. Or think of it this way, an AI agent cannot do social selling, let's say, on LinkedIn or any other platforms. So there's definitely always going to be a space for humans. There's always going to be a need for AI agents because companies have to drive efficiencies these days. They cannot be growing at all costs. But really, the ultimate way forward is we blend these capabilities for the max output.
and for the lowest unit cost. So now, okay, it's unit cost of which are optimizing this meeting. So,
Kyle James (07:18.81)
Yeah. Yeah. Wow. Thanks for sharing all that. And I would say like, really kind of like, got this number in my mind from, from the intro I made, like when the zero to 3 million, annual reoccurring revenue, like that's, that's a big deal. And I think there's a lot of companies who are trying to get off the ground, especially like some startup AI companies and like, do you think that like, if, you didn't have AI in place or like, I guess, how did you get there? That's probably the best way asking it.
How did you get there and like, do you think AI was kind of like the centerpiece of that to get to that next space?
Frank Sondors (07:52.686)
Yeah, so the first software we launched, have six softwares. The first one is Salesforce, the mothership. And when we launched, chat GPT just came out. So it was all about GPT-3 and using the GPT-3 in cool ways. And back then, it was all about sending templates to prospects. But there's an intrinsic problem with actually sending any sort of static piece of communication, let's say, as a cold email, which is that on one hand, folks have a fatigue with templates. You can see it's not really crafted for you as an individual.
And the second problem with any form of templates is that Google is very smart these days. So when Google sees that you get a lot of spam reports and it sees that you have a static piece of content, you will get penalized. So more of your emails will start landing spam. So yes, when we launched Salesforce, the main reason why folks were flocking to us is because we implemented GPT into our core sending engine. And you could craft a unique email to every single person. And yes, it was not perfect. It's not as good as GPT-4 right now, et cetera. But back then, this was
as good as it gets. And people are like, wow, you can compute an email unique to me. And I would send that off. And then folks would just go and buy my stuff just because of the email that they saw. So that's also one of the reasons, right? Because the core value prop is like, we don't just send emails. We send really good emails. And there's also good logic. So AI is all about good input in, and then it gives you a good output. So the way that we have designed it is we took the context about the business. So what is it that you're selling?
And then naturally we combine that and match that with what is it that we know about, say, you, Kyle, based on your LinkedIn activity, based on what's on your profile, on the website. What do we know about Kyle? All the context that's available on the web. So we stitch these two data sets together. We apply GPT on top of that. And voila, we had a email uniquely designed to you. And in Europe, we could do that in French, in German, et cetera. So that's why it took off. that's also one of the reasons why it became super successful.
But back in the days, AI was always like, we are AI powered. Today, AI is like the default of any software. So if you don't use AI, or if the fact that you say that you are AI powered is already old school, because users' contradiction today is that you are AI powered by default. It would be kind of crazy to think that you're not. The only thing that I would just mention that you have to be definitely margin conscious, because naturally, AI
Frank Sondors (10:11.598)
as you scale can be very expensive. So you definitely have to manage a lot of the aspects of how you build this, et cetera. I think that was the biggest risk. naturally, the users are just like I think you mentioned earlier, Kyle, if something is AI based, and that's where the users aggregate today. If you don't do anything on the AI front today, it feels that you're a bit of an old school of a company, unless there's a good reason why you don't use AI. But yeah, so.
Kyle James (10:36.818)
Hmm.
Frank Sondors (10:41.342)
let's say if you're big enterprise and when you are scanning for which of the software you should be buying, you have to have AI by default. It's like a tick box. And so in a lot of cases, yeah, mean, if you don't have AI baked in, it would be just weird. Let's go. It's just weird.
Kyle James (10:55.826)
Yeah, yeah, no, for sure. Yeah, absolutely. I mean, it's, I, it's going to be like the norm, right? Like where like right now it's like, Oh, everyone's trying to get AI, but at some point, many of the businesses are going to adopt and the ones that don't, there's a good chance they may not, they may not move forward, right? They may not be able to hit their scale that they're trying to go for their goals. So walk me through, like, let's just imagine here for a moment, like there's people who are listening and like, Hey, I see this. I'm using it like Chaggbt. I'm not really doing much, but like, how do I
What does it look like step by step? if someone were to come on board with a company and say, yes, I want to embed AI with the email, like walk me through, what does that process look like for them?
Frank Sondors (11:35.118)
Sure, so I wish the process would be very simple. But the reason why the process isn't as simple today is because 99 % of companies have never used an AI agent today. And so because of that, using any form of agenda capability feels a bit like the internet moment where you have to get, the humans have to get onboarded on this. So we do a proper onboarding call. So even though the agent setup in our eyes is super simple,
we always see the problem that unless a human assists you and boards you on the AI agent for like a week, know, hand holds you, et cetera, helps you to monitor, et cetera, things will just not pan out. So that's exactly what we do at Salesforce. So we set up an onboarding call with every single customer that wants to use Agent Frank. We get them set up. We set up the infrastructure. We sort out the leads. We look at a copy. We make sure the contact is correct, et cetera. And then
Off they go. So typically after a few weeks, et cetera, they're good to go. And so the customers keep on using. And they keep seeing results. And typically, customers come back when something doesn't work. But yeah, customers still have to monitor the agent. But yeah, there's a typical checklist that we do with every single customer throughout the onboarding. And then we follow up with the different bits and bobs via Slack channel. So that's something also that we set up for every single customer. But yeah, once the agent is in place,
then there isn't that much that you need to do. The biggest thing is just to monitor the replies and sometimes intervene. So sometimes if the agent doesn't know how to respond with a specific question, then the agent will stop and will wait for the human to intervene. So that's where the human definitely comes into play. But otherwise, the agent can be run literally fully autonomously. Well, fully autonomously, you get meetings booked. If you don't trust the AI for some reason, still there are lot of people that are not still comfortable with the AI. You can run it in what I call the
Kyle James (13:25.042)
Ha, yeah.
Frank Sondors (13:29.55)
So that means when the reply does come back from your prospect, you take over every time.
Kyle James (13:35.314)
Okay, so it gives like an option. initially it's like, it can be completely automatic where the AI is taking over, but if not, you can have it where there's a little bit of manual interjection from the human to keep it running for different. Okay.
Frank Sondors (13:48.62)
Yeah, I think in the future, like majority of stuff will be fully fully autonomous. the reason why that's the case is, is because if the human intervenes in the replies, it slows down the process. So if essentially there's a, there's a stat out there somewhere on the internet that if you reply, if your reply takes longer than two hours to reply to your prospect, then the probability of a meeting happen significantly drops. So you have to be really on the ball.
Like when the reply comes in there has to be that ping pong right there the prospect replied to you and then you want to reply back which within minutes right and because they're still there maybe at the desk the typing like the the it accelerate you know, it accelerates the sale cycle and You're able to get the meeting on the books much faster than the human intervening doing bits and bobs etc. So like I don't see a reason why Maybe right now there's a reason but like as we progress
In a few years, everything will be fully autonomous because the opportunity cost is just too high, leaving it to a human to respond. So yeah, that's why I run everything on autopilot. as I said, the only reasons why you want to do the co-op light mode is because you don't have enough information about your business in a knowledge base, what we call. So typically, you want to upload it like a sales playbook, pricing, FAQ, et cetera. So there's that context that we feed.
so that the agent can reply every single time. And yes, you may see that, wow, the prospects are asking this particular question, or there's a pattern to the type of questions that they ask. Then yes, you can upload more documentation. So essentially, the agent just never stops. So you can just keep on going.
Kyle James (15:30.386)
Yeah, yeah, absolutely. I appreciate you sharing that. like, what would you say, um, like for those, especially with some of your clients you've been working with and, maybe internal, cause I'm sure you're using this for your own team and then also for your clients, but like what types of results have you been seeing so far? Um, maybe that's something that's worth mentioning for those who are like, Hey, I need to, I need to set up AI. I need to set up automations, but I just don't know where, is it going to work? what, what, anything, any key metrics that you think is worth notating from, from the results you've been seeing.
Frank Sondors (16:00.302)
So right now, we only do email and outbound. So that's the space that we operate. So typically, the metrics there are that companies see about 1 % reply rate just globally. That's called the email reply rate. About 1%, that's typically best case. Our agent has about 2 to 2.5 % reply rate on average and about 10 to 20 % positive reply rate. So like 1 or 2 and 10 are positive that are coming back. And this is pretty good.
Kyle James (16:23.538)
I know is this only ask you is this cold email only is this a mix between OK so just cold email OK gotcha.
Frank Sondors (16:29.006)
For now, we only do cold email. We're looking to massively unlock other channels. So for example, inbound. If we would unlock the agent on inbound, performance would go through the roof, Because naturally, you have warm leads. They're coming through. They're showing intent and stuff like that. The other thing that we're looking to roll out in about two months or so to make the outbound motion a lot more powerful is we're looking to add LinkedIn. So LinkedIn is not something that we offer yet, but it's coming soon where you will be able to connect unlimited LinkedIn profiles.
And you'll be able to essentially hook up your whole team to the agent. the agent will use your profile essentially as a sidekick, essentially, as an assistant. you'll be reaching out. So the agent can reach out to the prospect via email. And then the agent will use your LinkedIn profile to also reach out via LinkedIn, let's say, as a step two. So that will massively increase reply rates. And we're pretty confident in this, because we're already running.
a behind the scenes. So for example, we have about 5 % reply rate right now on email only. So this is called email, about 5 % reply rate. But on LinkedIn right now, we have 30 % reply rate. So a 6x lift. This is statistically significant data. In the last 30 days, we sent about 8,000 connection requests. So it is quite a bit of volume that we able to say whether this is significant or not. So there's definitely a huge lift.
Kyle James (17:31.922)
Wow, that's really high.
Frank Sondors (17:50.27)
And so we expect that if we would apply LinkedIn then to every single customer, and then the customers would take the whole team and connect them to the agent, they will see far superior results. So we expect that to get to about 5 % reply rate. While the positive, I think, reply rates will, the percentage will still probably be roughly the same. But what we will improve is the more replies. Essentially, you will be getting more replies because of LinkedIn, because multi-channel does add more replies. And as a result, you will be generating more pipelines.
Kyle James (18:20.016)
Yeah. Yeah.
Frank Sondors (18:20.046)
So, so yeah, there's definitely major improvements there and then we're looking to add other things like videos later on memes and you name it
Kyle James (18:27.346)
Yeah, that actually brings up and brings up next question that I mean, obviously the LinkedIn is a really cool approach in the video. Like what are some of those maybe upcoming AI initiatives, you know, stepping in these next couple of years and like, where do you see it maybe playing that biggest? You talked about inbound as one option, but obviously outbound is kind of the main goal. Like what, do you see the operations going to next?
Frank Sondors (18:46.51)
So, we have seen the last sort two, three years, we have seen a huge progress on the text front, computing text, et cetera. Then we have seen major, major improvements in voice. So this is where we have unlocked, for example, voice for our own business we call, and it's legal, you can call your inbound leads if they don't end up buying, or you can call them immediately, it's up to you. But the agents are really good at inbound. And,
Then the next wave and the final wave is the video of the wave. And I think some folks have seen this year, LLMs being applied to making movies, et cetera. But guess what? Just taking your picture, off LinkedIn, I can make you say things and do things on the camera whatever I want. And that is possible. And the tech is getting really, really, really good. So this year, and literally over the next year or two, you'll be getting more
Kyle James (19:17.17)
What you mean by that?
Kyle James (19:30.898)
Yeah. Yeah.
Frank Sondors (19:41.71)
videos from folks out there where they haven't done the video. They just tossed the AI to tell you specific thing. And again, the concept will be very similar. I will be talking about you and your business on the video. I'm going to try and explain how I can help and solve potentially a pain that I'm assuming you're having. Or maybe I see something on your LinkedIn profile or something as a LinkedIn post that you mentioned, and I can kind of latch on that. So that will be, I think, the biggest disruption that we're going to see in the world of
sales technology generally, not just sales technology, but also naturally movies and everywhere with the videos, essentially, we're going to see huge, huge impact there. And I'm personally very excited because I can give you any accent, Kyle, whether British language or any American accent, but also I can make you speak any language. I can make you speak French and German, whatever. So all of those capabilities are lux.
huge, huge conversion lift for especially global companies out there. So I'm building big on video. We've got to be building video as well ourselves because we do see a huge potential to embed essentially videos within sequencing essentially or campaigns to drive a significantly superior lift in performance. Because the problem with some companies when we work with them is they have a very tiny TAM.
And so when you have a very tiny time, you need to essentially engage with the prospect that you're working with in multitudes of different ways. But also, what is it that you, how you engage, right? So what are your texts, send them a quick email message, or maybe a LinkedIn DM. But then, what will trigger sometimes people a lot is the video message. So there are a lot of industries where video is not that saturated at all. So when you get a video, you're like, wow, this is really cool. And somebody can explain to you.
in a video far, far better than a piece of text. And so people are susceptible to videos. It's just that some folks digest information via video far better. So folks like me, for example. So I'll give you an example. I really hate reading books. I know it sounds weird, but I really hate reading books. But what I really love to do is watch YouTube videos all day long. All day long, I could watch a YouTube video and I could digest the same book. I'm gonna digest far better.
Kyle James (21:49.952)
Yeah.
Frank Sondors (22:02.126)
and I get to get a dopamine hit just by being on YouTube all day long than me reading a book. So some people are wired like that.
Kyle James (22:10.842)
Yeah, yeah. No, for sure. I think that's, mean, there's so many approaches that, that you can take. And I think the video generation is something that's super fascinating. I imagine even like a movie, right. As they're developing new movies out there, it's like, okay, let's start throwing out like different lines for different people. Like, Hey, let's see if people respond appropriate with this AI generated trailer. Let's see they respond to this. And then based on that, they can create content. And it's like, there's just so many different, and for those like listening in, as we start wrapping this conversation up, Frank.
Where can people learn a little bit more about you and then maybe learn a little bit more about Salesforce that you'd recommend them check out?
Frank Sondors (22:45.646)
Yeah, the final thing want to say, even this podcast could be all done by an LLM. So you never know whether this is even real or not, right? So I just say. But if you want to find out more about me, Frank Saunas on LinkedIn, can find me quite easily. The guy with the bacon emoji comes from the saying, bringing home the bacon. That's literally what we do. If you're not sure what bringing home the bacon means, just go to chat.gpt and check it out. Otherwise, can out our website. You can book a demo if you want to check out how the agent works, cetera.
Kyle James (22:51.051)
Yeah. Yeah. Yeah.
Kyle James (23:07.986)
Yeah
Frank Sondors (23:14.584)
Yeah, otherwise, you can easily find me on the web.
Kyle James (23:17.822)
Thanks so much, I appreciate you having on the show today. And for those listening in, remember, if you're looking to implement AI into your business today, don't try and do it yourself. The time and stress that AI could cause just isn't worth it. Schedule a call with GPT Trainer and let them build out and manage it AI for you. Once again, that's gpt-trainer.com. Signing off for now, have a great rest of your day, Frank. It's a pleasure, my friend. Hopefully we'll see you on the show again in the future. And looking forward to seeing everyone on the next episode of AI Chronicles.
Frank Sondors (23:40.312)
Cheers to you.
