SaaStr AI Live on Wednesdays: Is GTM Really Dead?! with SaaStr CEO Jason Lemkin

And then there we go. And let me let everybody in. Okay, you're live.

Uh, good morning everybody, welcome to Saastr AI Live on Wednesday. Um, I wanted to do something, um, hopefully brief and specific, but then open it up to any questions you guys have on go-to-market, Sales and marketing today in the age of AI. And I really just wanted to hit on one thing. I'll give you, I'll give you a whole bunch of points around it, but one thing, because I just, I, I see a little bit less of this on LinkedIn than I did 6 months ago, but man, I see way too much that go-to-market is dead, that, uh, outbound doesn't work anymore, that SEO is dead. Uh, woe is me that all this stuff just doesn't work in the old go-to-market toolkit. Um, and you know, uh, I would say last year was a little confusing, um, it— for a variety of reasons that we'll get into. But here's the funny thing, guys. If you look at who is running go-to-market at some of the hottest AI companies, from OpenAI to Claude, uh, I mean, to Anthropic, um, on down, you'll see folks that spoke at SaaStr in the past few years. You will see the exact same people essentially running the same playbooks they ran before the AI explosion. Now, not identical, there are differences, and we're going to chat about it, but literally they're the same people running somewhat upgraded versions of what worked. So here's my tough love message: a lot of things are not working quite as well as they did. Some things are working better. It's not that simple, we'll go through it. But if you think things are dead, or if folks on your team are saying it's dead, or you hear mumbles in the hallway that it's dead, I think it's you. I think it's you. Because just about everybody that is tapping into AI budget is reaccelerating now. Uh, you can see it everywhere. Notion just said they're reaccelerating at $500 million in ARR as they launch their agents. Dialpad and Talkdesk have added 9 figures of incremental ARR from their AI agents this year. Um, this— all this NVIDIA OpenAI money is trickling down to vendors that supply them. Datadog has accelerated. Because it sells to all those leaders because it does all of DevOps for, for, for OpenAI. There's just so much trickle-down from this insane spend in AI. The problem is you're not getting any of that spend. You're not getting any of the spend. Others are, and we'll go through it. So is it dead? You know, Halloween's coming up, or is it you? It's you. Okay. So a few examples. And if you guys have questions, put them in the chat or save them up because I want to spend more time on questions than slides. but I wanted to present a few things. We've done a bunch of Workshop Wednesdays and others, and, we've got another post on this on SaaS for This Week on how we're using our AI SDR. Okay. And here's some data. Amelia put this together. Feel free to chime in, Amelia. You don't have to if there's anything I'm missing, but it almost doesn't matter what this says. This is us tuning up our AI SDR. Yes, we replaced our human SDR team with an AI. Yes, we had to train it for a month. Yes, we have to train it for every day, but once we do it, It works. 1,206 leads to Saastr from this, from 500— 5,216 messages. That's pretty good. 136 responses. So that's a pretty high direct connect rate and 52 positive. Well, I guess some were negative or neutral, but this works. This is sending 5,200 messages out and getting 52 folks that want to buy from us. Pretty good out of 1,200 leads. Really, 52 out of 1,200 is pretty high. Are there challenges? You know, first, we all, you know, if we, let's be honest, a couple years ago we were loading up a bunch of random purchase names into Salesloft and Outreach and friends and just spamming everybody in the same cadences. And we all saw that decline well before the age of AI. With AI, it's worse because we're getting more of it. In fact, AI is making more of it mediocre rather than terrible. It makes more of it mediocre. There's fewer typos. But it's still not very good. And yes, is that stuff performing worse? Yes. But send an email and add value to the customer. They're going to open it. It's not dead. Outbound is not dead. It's you and your team. Two, just a fun one. I popped into Maggie Hott, who's also coming to SaaStr AI London in December. Be there. It's going to share how OpenAI does its go-to-market. I asked, I asked, I asked an AI version of her, what are the top go-to-market things that you do? Top go-to-market tips and It's the same stuff. I like this one. A lot of folks think pilots don't work anymore. Customers aren't paying. Our pilots aren't converting, blah, blah, blah. OpenAI does pilots now. OpenAI, the hottest company on planet Earth, are you going to other than NVIDIA? To close big deals, they have to do pilots. In fact, here's the thing, folks. If you listen to this, if you listen to the recent 20VC Harry, Rory, and I did with Mark Benioff, he wants to do more pilots. Everyone wants— is talking about forward-deployed engineers. We could do a whole Workshop Wednesday on that. But Mark had a good point. I asked him what he thought about— Mark Benioff— what he thought about forward-deployed engineers. And he's like, I love it. I want to have my team before a customer even signs a contract be in production with a great AI agent. I want it already working. The old school version of this is sort of try a pilot, sign the contract, and maybe over 2 years we'll roll it out to everybody. Even Mark wants to go hard on day 1, but they want to do it. Pilots still work, folks. So folks are saying, it's not working, it's because they don't really want your product anymore, or at least they don't want more of it, which we'll chat about. Okay. 3. What about old school stuff like webinars, um, and case studies, man? Those can't work. Well, here's Anthropic and Cursor partnering on a webinar to produce case studies. I mean, this stuff works. Um, people discover you, they want social proof, they wanna know who else is using your app. In fact, in many ways, all this stuff in my experience works better today than it did 18 to 24 months ago. Why? Well, we've seen a vendor explosion. I, I, Amelia and I, I can't tell you how many times a day I get questions, which AI SDR should I use again? And I, I'm always like, well, Google it. We did like 11 webinars on this, but I get it every day. Okay. Every day. And what does that tell— it's just, it's a birth of new discovery. Like for years our vendors were kind of fixed. Oh, should I do Salesforce or HubSpot for CRM? And Should I use ZoomInfo or Apollo? Now there's 100 AI-fueled vendors and a lot of them are new. What should I use for video? Should I use Higgsfield? Should I use Re: for images? Companies you didn't even hear of 12 months ago. So people are doing more discovery than they have in our work lifetimes. And that's why this stuff helps because I wanna know who else is using that app, who else is using Gamma, who else is using Artisan. So webinars and case studies aren't dated. Check all the leading AI guys. They've got it all over their website and they're doing it with each other. Okay, here's a subtle one that's not so subtle. Brand. Why are all the AI leaders everywhere? Why are Greg Jensen and Sam on the left here doing this this week? Why does Amjad from Replit on Joe Rogan? I mean, you know, craziness and Austiness aside, You know, Replit is in a brutally category, you know, Replit's at $1 to $150 million this year. I mean, pick your jaw off the table, but Lovable's almost as fast. Bolt's at $60. Believe it or not, Wix bought Base44 and they're growing almost as quickly. Now, this is a brutally competitive space. I just saw a new one, a new competitor launched on YC today. I just asked it to redo the Saastr London site better today. It did an okay job in about 5 minutes. There's so much competition. And again, when I'm deciding which product to use, I want constant reinforcement of who the winners are. Brand still matters. This stuff works. This is the go-to, go-to-market playbook. It may even be more critical today when PLG and there's so much AI demand that we need more help making that decision quickly. Brand may be more important than it's ever been because there just aren't the same defaults. Okay, another one here is, here's Figma. Figma can bring thousands and thousands of its customers together. Events still work. In fact, if you talk to a lot of marketers at faster growing tech companies today, they're actually quietly doubling down on events if they can get their teams to do them because they know it works being in person with prospects and customers. They know it works in an era of so many vendors and so much noise and so much confusion. That connecting with people in person works better than ever. Some marketers are struggling that their team doesn't really want to do the work anymore. They don't want to build the booth. They don't want to show up. They don't want to spend the time. But it works better than ever. And, you know, on the tech side, look at how many hackathons everyone's doing. Lovable, Bolt, RevenueCat, everybody. You just— I mean, you can't walk 60 feet in San Francisco and not stumble into a hackathon morning, noon, or night. I mean, they're everywhere. And there's a lot of things going on here, but it's also marketing. It's also marketing. Get folks building. The more folks you can get building on Lovable, the more folks are going to stay with Lovable. Events and IRL, they still work. Okay, here's one that a lot of folks are complaining about on the LinkedIn too. Oh, SEO is dead. You know, I got all my customers from SEO. Well, let's slow down, guys. Okay. First of all, here's SaaS itself. I honestly don't know why. But our SEO is up depending on where I look, 3x to 5x in the last 12 months. Okay. It's just up. It is up. Now, other— now, is it perfect? Is ChatGPT and others also taking traffic away? Yes, both are happening. Both are happening. And there's no question many leaders, if you look at what G2 is saying, we just did— Amelia and Guillaume just did a SaaStr session with their CMO, Sydney Sloan, the other day. Their SEO is down, like parts of their SEO are down, but it's not zero, guys. I don't know anyone that had like great SEO that wrote some of the most canonical best content teaching you how to do something really important who a year ago millions of people are reading it and today it's zero. I just don't know anybody. So one, it may not be down.

Okay.

Two, sure, the LLMs and ChatGPT are taking some of that. Adjust. The world changes. But three, it still works. This— it still works. So if somebody's telling you, woe is me, SEO is dead, it's just another excuse. It's you. It's you. You're not producing enough high-quality content that the internet wants to hear what you have to say. Produce great content. It works. And I actually think that's what's happening with SaaS around a variety of reasons. For what it's worth, if you look at this little chart on the right, um, I'd actually never looked at this data until like a month ago. Um, uh, that's what happens with a small team. But our average position in Google SEO is only 28. Like, that's pretty bad if you talk to you know, an SEO marketer, they want you to be number 1 or number 2 or number 3. And we are number 1 or number 2 or 3 for some things, but overall we're 28. But actually, I think if you look at what everyone says out there, SEMrush and others, actually ChatGPT and friends benefit you if you have high authority but you're lower, lower in Google because they don't care. They're just looking for the best piece of content to answer a question. So we've probably benefited from that, not because in the old days we had lots of We were number 1 for how to hire a VP of sales, but Saastr's got 10,000 posts on just about everything, everything you could possibly imagine in GTM. So we're going up. So I'm sorry if your outsourced agency content farm strategy from 2023 doesn't work today, but that doesn't mean SEO is dead, right? But let me summarize it and then dig into a few things that maybe are working differently today. One thing is clear, and I have a lot of data I just shared on outbound and SEO and others. I don't have the perfect data here, but I can tell you from just about every startup I work with or invest in or our own stuff is like crappy is working less good. Crappy is clearly working less good. Crappy outbound, we don't want it, right? We're blocking folks. Um, crappy PR. If you guys want— if you guys recognize that PR has changed like dramatically, but you want it, you want to be at SaaS or you want to be on Joe Rogan, you want to be profiled It's worse than ever because these AI tools have actually made mediocre PR so easy. I don't know. I used to get like a— now I probably get 20 emails a day with stuff that's embargoed and mediocre and some AI that writes a pretty good email saying that someone I've never heard of wants to be on our podcast. Like it's easier than ever to do crap and it's just not working, right? The third thing, what I see, this is where I see so many, especially marketers struggle, but CROs too. The low energy crap that still kind of worked a few years ago, the cloned boring campaigns that there's no reason I would go. Low energy stuff, incredibly boring events with recycled speakers, boring digital events. It just doesn't work. Like we don't— we don't— we've lost patience for the 11,000 thing or the 10,000 ROI calculators aren't good enough. So I definitely think running the same in many ways, running the same playbook from 2 or 3 years ago with less energy, you would think nothing works because that stuff don't work, at least not in my experience. So here's a couple challenges to think about. Some of this sounds Captain Obvious, but I got to tell you, I don't see enough people doing it. If you think something's not working, instead of complaining or giving up on it or investing less energy, what about making it awesome? What about making it awesome? Audit. Just like we've said on SaaStr for the better part of a decade, you have to read your sales team's emails because they're probably terrible. Even today, they're terrible, even with AI. They're probably terrible. Slow down and audit everything. Okay? Make it awesome. So for example, we've done a bunch of sessions on our AI agents. Amelia has done a bunch with Gamma, which we're super fans of. And one of the things that Gamma lets us do is take our own sales collateral that we use for SaaS events, and we have to close 10+ million of SaaS sponsors just to keep the lights on because our events are so expensive. SaaS annual, 10,000 people. We got to— we do have to close some sales. And we, you know, we used to have a terrible prospectus that we would finish like 4 months late. Then we had something that was okay. Then we made it really, really good and polished it and we had a designer do it. But then it was static for a year. Like we couldn't really change much. And then sometimes someone on their sales team would use an old one or they would move it around or break it. It was terrible. Now we use Gamma, which does dynamic decks with AI. It's and everyone gets a custom deck now. Now, sometimes there are a few downsides.

The—

what we can do with design is not quite as great as we can do with the designer. But if someone wants a booth in this, we instantly tell them, how does the booth work? What's the ROI? Who are similar sponsors? Who are similar people that have been doing it? Everyone, instead of getting an okay piece of collateral, get something awesome. Make your collateral awesome. Make your outbound awesome, right? Ask yourself if this stuff's actually any good. It's probably not any good. Ask yourself if you would take that meeting, if you would go to that webinar. Dust it off. Get the best person from the hottest customer you have to come to you, do your webinar. No one wants to go to the low-energy stuff. Try making it awesome. Okay. Just a few more points. The— it's not though, if you look at the hottest sort of AI B2B and AI startups, It's not that they're running different playbooks, but they are remixed. They are different. And a few obvious points, but it's worth bringing up how they are different. I would say almost if you look at most of the hottest AI-fueled B2B companies, many of them have a sales motion, right? The whole OpenAI presentation from Maggie at Annual and the one she's gonna update in London, it's all about how the go-to-market team and sales team works, right? There are plenty of sales teams at Anthropic, There are plenty of sales folks at OpenAI, but at many of them there's fewer. So we had the chief business officer of Perplexity at SaaStr this year. You can read the post. He had 5 folks on the sales team, 5 folks on the sales team. The sales teams, there are sales teams at Replit and Lovable and Bolt and others, but they're small. We're going to have an upcoming session with Vercel, their chief business officer. They have 1 human SDR and 10 agents. Okay. So the teams are often smaller, but it doesn't mean they're not working. It's just, it's just AI fueled. Um, and the third point, and this is confusing, um, what's going on? I mean, you're like, you turn around and you're like, oh my God, how did, how did Repliq go from $1 to $150 million? And how did Higgsville go from, you know, $0 to $50 million in 5 months and Gamma $60 million this year from nothing? It's like, they must be these great marketers or great— no, I— what they have, and I know this is obvious, is this is just the strongest demand in PLG we've ever had. They're just running PLG on steroids. They're trying to do everything, create some virality, create brand awareness, create this and that to keep the engine going. But if you can provide insane value with AI and B2B and you can get the word out, the demand may be insane. So that is different and fueling that demand is different, but it's ultimately deep down, it's just PLG on steroids. Steroids. And again, my last challenge to you, when you, when you scratch your head and you look around at these folks that are, that are running circles around you, just realize it's like, do it. Go on LinkedIn and see who's, who's their head of sales, who's their head of marketing. You're going to see folks that worked at wherever, Brax and Rippling and Ramp and Datadog and Mongo. You're not going to see folks that have been doing AI since they were in grammar school or anything like that. You're going to be seeing folks again You're gonna— if you look at half of the top SaaS executives that has presented at annual and online over the last decade, most of the best ones are still in the game, but many of them are at the hottest AI companies. So it's not like they've created an LLM on their own, right? It's not, it's not none of the stuff. They're running versions of the same playbook, but this PLG pull is insane. Okay. So just a couple points to summarize and then, then I'll take questions on So that's what's the same, but on steroids, what's not working and what's different. Um, I know a lot of you know this, but it's worth slowing this down. Um, one, we just talked about it, so I won't spend more time, but the demand for these high ROI products is insane. It is insane. And I can just tell you with some of the ones we use, um, they're because we can do something incredibly valuable we could not do before. So we spend a couple hundred bucks a month on Replit now. What do we have? A new SaaStr AI, a new homepage, a new SaaStr valuation calculator that anyone can use to value startups. We've had over 300,000 people use it in 5 weeks. 300,000 valuations, maybe not, maybe fewer people, but 300,000 valuations. We just launched an AI that reviews your VC pitch decks. It's free. It, it's pretty awesome. We should cross 1,000 folks have used it by the weekend. So, you know, the, like Gamma at first when Gamma came out for, for decks, I'm like, I said to Amelia, oh, this is like a cool AI tool for decks, right? I didn't totally get it, but it's not just a better PowerPoint or Google Slides or Canva. No, it lets us create dynamic decks for sponsors, for sales we could never do before. That's a new use case. It's like super powerful or the videos we make on Higgsfield. So this is why, and I have this at the end, this is why copilots are like a failure. No one wants to pay for the Office copilot that does a little bit extra for another $20 a month. They want something that does something incredibly powerful and very meaningful with almost instant ROI. But if you have that, the demand is like we've never seen before. This is sort of end user demand. The second point, which I know everybody knows, but I see too many people still banging their head against the wall in the enterprise, or once you're budgeted, once you, once you're at the CIO or even in a, in a, in a, in an SME, a bigger company, there's budget. And listen to anybody on the internet, any CIO out there, they're gonna tell you the same thing. I'm actually, there's too much demand for AI tools across my organization. There's too much demand. I, I don't even know what to do today, but I can't turn it off today. Everyone wants more AI tools across my company, but I'm still cutting the efficiency tools. I've still got 200 classic SaaS tools for pipeline management and this and that. I want to end the year with 200 or less. And even worse, Salesforce just raised prices again the other day. Everyone's raising prices. Zoom just raised prices again. Everyone's raising prices. So they want to hold the line on the efficiency budget in software. Um, but even worse, even if there is any extra budget, it's all being taken up by price increases, which— what does that mean for you? Again, if all your app is is an ROI calculator, if all it is is you're trying to show how to make the sales team 10% more efficient. That was the game in 2021. No one has any budget or interest in that today, no matter how great your app is. There just isn't any budget at the CIO office. It's all being consumed by AI and price increases.

It—

listen, if you break out and do something no one's ever done before, you'll get budget, but that's really no different than AI. There's still always experimentation and new budget, but there just isn't any budget for another 20 sales tools today. There just isn't. Okay, and the last point why so many folks aren't getting AI budget out there, if I look across startups I've worked with or invested in, is it just has to work. It just has to work. And this is why so many mediocre Copilots, so many sort of pseudo AI, or hey, we added a little bit to our app, it just doesn't work. Like the, you know, the classic B2B was Hey, you know, this is going to create efficiency in your company. Try it a little bit. And then over the— over a year, you'll see it roll out across the org. Now, there's still some truth in that in AI apps, but we're expecting magic up front. And going back to the whole forward deployed engineer, Mark Benioff's point, we're expecting the vendor to do most of the work to train this app and to get huge ROI up front. I mean, we were— again, we, we essentially replaced 2 human SDRs with our AI SDR, but it's not that simple. It's not just paying money and walking away. It's a lot of work and it has to work or there's no point in it. So you could kind of, a sales team and a little bit of marketing fluff could promise a lot of efficiency benefits in the old days. No longer. You've gotta deliver this ROI almost instantly. That's what everyone's expecting. So no lame copilots. If you, if you're wondering why your product isn't growing faster in the age of AI, you have a lame copilot that doesn't change the game almost on day zero. Even Marc Benioff wants to do that. He doesn't know how yet today. but even he wants to do that. So my last, I know, slightly annoying kick in the arse, and then we can take questions if, if we have any, Amelia, but be honest. Everyone that has something good today in B2B that can tap into AI budgets, okay, one way or another, either they can tap direct, like at that prosumer level where there's just end user demand, or they can tap into it at the CIO's budget because it provides that value. They're all getting lifts. And I think if you talk to any investor, um, and I just had this conversation at YC Demo Day, you'll hear the same thing. They'll tell you, look, half of my companies, my pre-AI B2B companies, are accelerating now. Half of them are accelerating and half aren't. And, um, it's not just, again, a lame copilot or adding, uh, you know, a little, a little AI calculator. It's fundamentally adding insane more value than you can do before that taps into AI budget. So if that's not you, Hit, hit. It's almost too late, but it's not. Hit stop, hit pause and figure out what, what's changing in your industry and tap into this budget. And look at this fun one on the right. This company Filevine CEO has been part of the SaaS community for 10 years. So over a 10-year-old company doing case management for law firms, always had product market fit. He reached out to me over years. The numbers were always strong, but, but it was an ordinary vertical SaaS company. And then boom, they just leaned in and legal deep, deep on how to change case management with AI. Now it's worth $3 billion and it's accelerating. And I think the founder is going to come to London in December and share how they really did it for real, how they really AI-natified a 10-plus-year-old product. But if you do it, the budget's there. You do it, the budget's there. It's there for Notion. It's there for Filevine. Even stuff that are hard, you know, Intercom said Fin as a standalone product has already crossed $50 million. Okay. It'd be nice if it was $500 million. But that's from $0 to $50 million by leaning into where the budget is, right? Dialpad said they're accelerating earlier this year, $300 million, right? It does work. But don't pretend if you're not seeing reacceleration now in the end of 2025, it is you. You are not tapping into where the budgets and customers are in the age of AI. So go figure it out before it's too late. So that's just my analysis of why it's all still working. So I don't know if we have any questions, Amelia, but we could open it up if we we do.

Yes, there's one to start out with. Let me disable your screen share. Uh, we can come back to your slides if needed. Um, and then for folks who do have questions, if you can drop them in the chat first, just because we do have also live streaming questions. Um, just if you're on the Zoom with us. Um, but for— yeah, first question I know we've covered it a bit, but I think it's an interesting nuance. So the question from Cheryl is about our SaaS or AI SDR agents that we use. We've got a few. Can you share more about what you trained it on? Was it only past successful emails? Did it go through an exercise to identify what good looked like? I'll give my take on it real quick, just having trained the agents a bit, and then Jason can share his learnings on it too. I think the interesting nuance here, even though we've covered this, is that You actually don't, you counterintuitively don't upload your good emails to any of these AI SDR platforms. Like literally none of them that I've looked at, none of the ones that we use, even some of our AI agents that do like meetings bookings with our inbound didn't ask us what like good outbound or what good inbound looked like. So I think that's kind of maybe the counterintuitive learning. Is that they actually— none of these really do that when you train them. Like, that's almost a, like, prior to 2025 way of thinking about, you know, how if I'm going to train a person, right, I'm going to look at, okay, what are all the best sales calls, right? If it's a— if it's an outbound SDR, I'm going to train them on all the best sales calls. I'm going to get all the best gone calls, whatever it is. I'm going to find all the best emails. From our team, train this person on that so that they can duplicate that. And where the AI is fundamentally different, it doesn't necessarily need to know those things or want to know these things to bias the AI because it's gonna make it a lot more personalized to whatever's going on with that person today to add value. So the way we've trained our AI SDRs specifically for outbound is I never loaded into the platform Okay, here's some of our past very good outbound emails. Honestly, they're fine. And I think the thing is, once you look at your past outbound emails that like eventually led to Close One deals from a human SDR, I'm like, they were always just okay. Like they were, they were okay. And then the AI SDR, um, what we trained it on instead was not what a good look like. We just trained it on, okay, Here are different personas because we have different personas at Saastr. Here are our core personas. We train it on our core personas. We train it on, okay, based on past knowledge, what do we think would be the best way to add value to them? Train the AI on that. The AI will scrape— in any of these tools will scrape what they're doing now, right? It's gonna look at their LinkedIn profile. It's gonna look at their website. It's gonna look at any funding news. It's gonna look at any announcements they've done. It's gonna look at their Company X account. It's gonna look at their personal LinkedIn. It's gonna look at their professional company LinkedIn. And so when you train it on based on adding value, that's where the bulk of the training is. It's not what good looks like necessarily. Now you could argue that, you know, you're training it on because you're training it so much on adding value, that is what good looks like. Sure. But it's different than traditional, like, hey, let me upload to the AI platform a little screenshot you know, that Mikey did on the team to book a deal. It's training it on adding value to the customer and then letting it do its magic, which is it can scrape all these data signals in a way that no human can, right? Like that I can't, that Jason can't, that no, no human SDR that you ever hire will ever be able to do because they don't have access to all these data and insights and ChatGPT and Claude and whatever to go see what this person is doing, right? And so that's where it will then formulate a sequence for you. To reach out to this person on a way more personalized level. And I think again, the nuance there is that good looks different to every single person, right? And versus training it on like what good may look like for 5 personas, you're basically teaching the AI how to get it good for the thousands of people versus just like 10 emails are pretty good and the rest were crap.

That's a good— that is a good insight. The one thing I mean, I don't know what your experience with all of our stuff is. I think layered on top of that, one, those great emails are the past, right? They're helpful, but they're a relatively smaller data point than you might think, right? Two, you're ingesting lots of signals, right? Almost in real time. But three, all these platforms— here's the thing that I didn't get until the other day, until a new platform was demoed to me. It's Captain Obvious, but they're all running multivariate tests. So the thing is, one, it's great that you have one email, but they're running 3, 4, 5, 6 different tests and then they're iterating and then they're attempting to personalize it. Right. You can put multiple— you can run multiple tests in a way it's no human's ever going to do. Right. I've never met a human SDR that can run 5 multivariate versions of an email and iterate it and make an improvement and see how it interacts with signals. They're just cutting and pasting often in 3 different colors and 2 different fonts. Right. So the, the training is as much about the iteration of the data that comes out of it as it is all the history of what's worked in the past. That's a pretty, that's a non-intuitive and pretty small part of it. Would you agree with that?

Yeah, I would say so. Right. Like, again, I think when you, when you start to evaluate any of these tools for yourselves, there's a lot of different factors on which ones you should ultimately pick. But on the backend, a lot of them function the same way. And it's, There's a, again, there's a lot of nuances, but I think thinking about it, not, it's a combination of thinking about it as it's gonna take probably the same amount of time as it would to train a human being, but the training is not the same, right? Like in the way that you would train a human SDR on good emails, good calls, you know, what brand, right? Like words, how, how does your company add value, pricing. All of that. Again, it's much different on how you're telling an AI to go about and do this because it has just so much data. Oftentimes, I find myself telling our AIs to keep it simple, right? Like, add value and like, here's my laundry list of things you could say. And you do wanna put in all these variables because, again, good will look different and custom to each every single person versus like buckets. But at the end of the day, like, you wanna arm it with those things, but you also wanna tell it like it will decide for you, right? A lot of these AI platforms, especially for outbound, you can put in like, I think we have like 20 different things in Artisan right now that it could pull up on. If I wanna try and book a meeting for sales, there's like 20 different data points from SaaS responders literally in the last like 2 months. But I'm not gonna tell it, hey, use that one when you're outbounding Dialpad, use this one when you're outbounding Zoom. I let it decide it and it will decide it based on what they're talking about, right? So that's just another nuance than like We, we're overtly explicit to all of our AIs that we always want it to add value. And so we've set it up to be very conservative on how much it talks about SaaS and we keep it simple. And we typically talk about them a bit more. And then it surfaces some of the stats of how we can help them. Okay. Cool. Now there's a next question, which is related. A few folks have asked, what are you— what are we driving towards with the AISDR? Is it a meeting with an AE? We started with one AISDR, if you will. Now we have 3 or 4 different agents on the same platform. They're all— they're trained a little bit differently. We covered a bit more in the deep dive, but we do have one that is working towards meetings with an AE. So I'll say like, if more of you are getting started with this, um, than maybe are further along than Zaster, you shouldn't expect it to work instantly either. I think a lot of founders I've talked to were like, oh, I loaded up an AI SDR. I, I trained it and tuned it and that took 2 or 3 weeks. And then in like week 4, I was ready for it to book me a meeting instantly. I also don't know of like a lot of human AI SDRs that would, or a lot of human SDRs that would book you a meeting their first day on their first email. Like, yes, it's happened and it's happened for our AI too. But most of the time it's gonna take multiple emails, even though the emails are a lot better for us. We've seen, again, much to a lot of Jason's points earlier, like we've kept a lot of the same things that work cuz they're not dead. And plugged it into our AASDR. So we told our AASDR specifically for booking meetings, like usually for us it's a 3-step process to book a meeting, right? Like it's 1, it's 1, 2, 3 really good outbound emails all sequenced out, you know, different days, all custom. And so we've told our AI to do the same thing. We've replicated that. And so now for us, it typically still takes those 3 emails even though it's our AI, to get a meetings booked, right? And then from there, there's a lot of variables that right now we don't have like an AI BDR. So they're getting on the phone with an actual human being. So if the, you know, the close rates are up or down from AI right now, it's about the same. 'Cause again, there's a lot of other variables. So keep that in mind too. Like I think I've talked to a lot of folks lately who are like, okay, I'm, if I'm gonna fire up Argovive, like I expect to see instant results like Soundstrasz. Like we saw instant results with our ticket agent, but that's a lot different. That's a lower ACV. That's a very highly targeted, like warm outbound audience that we used it for. And then for the one that was our more traditional, like, AISR meetings booker, that one took a few weeks to get going. But again, that's true of any human being either that you would hire. Yeah.

One small point out, there's so much we could chat about here. It is interesting. I don't, you know, the— on the qualified side, right, on essentially the inbound, right. What I really like— and again, there are a lot of things you, you could do this pre-AIR with other tools. It's just much better now is that now the— it's not perfect, but the AI will qualify the prospect on its own. Right. So there's nothing I hate more than a qualification step, which is not with the seller. Right. It's just awful. So we don't do that anymore. And then as it does it without, without seeing obnoxious, it then does create the appointment and sets up the meeting and puts it in the calendar of the rep. Right. Could you do that with Calendly and Chili Piper and all that? Of course you could, but you couldn't automate all of this. You couldn't have someone come to our website, talk about what sponsor package they want, be recommended a package based on what they need, being shown other data and said, hey, do you want to talk to David or Amelia or whomever? All in one elegant step. Um, it— that's a, that's a huge positive rather than, you know, even today in 2025, I can't tell you how many times I inbound to a vendor and it's like a week before I can get a meeting set or someone tries to qualify me out. It's just unacceptable today, right? So not exactly the initial topic, but this is like if you smell any— I mean, literally there's a vendor, there's a, there's a pretty cool tool, a niche tool that I literally love. I hadn't seen anyone automate this before, so I reached out to them on Twitter and I said, listen, if you can get this to work for me, I'll be, I'll be on your homepage. Like, I will tell everyone to buy your product. It's so great. And then a week later I get a DM from the CEO. What happened? I mean, what do you mean what happened? You did nothing. He's like, oh, we emailed you. I'm like, I don't know who on your team said they emailed you, but like, I got no emails. This is a week later. And he's like, and then finally the CEO sends me the email himself. I'm like, I love you, but that last week, I'm too busy. I— you lost your window, right? So if you—

if that—

That, that tragedy you can eliminate instantly with AI. Like, that should not happen. The, the, the, the stealth qualification, you can pull all the data, you can pull all the signals, you can, you can, you can, you can, it'll hook up all the data on the API. I don't really know who you are, right? For the, for the company, for us, right? That meeting should be set in real time. Like, they just lost me. I might do this in a month or two, but like, I, I'm pretty busy until after Halloween now. So they lost their, like super fan customer. Like I would have been there, you know, and if we really back somebody at SaaS, we're not perfect, but we do have reach and we do it because we believe in it, not, not for any other reason. When we back someone hard, it, you know, it works. So a missed opportunity that AI can instantly solve. So a niche point. But that's one of the things I love about it, right? Is no cut, no prospects left behind.

Yep.

No prospect is left behind. And I can't tell you how many sales motions I see participate in watch. Where anything other than the easiest low-hanging leads, even in 2025, are left behind, and there's just no excuse.

Yep. There's a few related questions, and then there's a few others about other things in the presentation. So I'll try and hit the related ones. So it's not so jumpy. Were all of your AI SD— or sorry, were all of your AI agents trained on the data you aggregated from over the years or was it new data? That's an interesting question. Again, nuanced answer, right? So like for outbound, and we're starting to get questions in the chat on like how we train our outbound versus like inbound agents. Different, differently. Like again, you have to remember inbound is much higher intent. They're coming to your website. They probably already have a sense of what you do when people come to Saastr's free websites. Or 3 core, 4 core ones, right? They're there for different use cases. So we've trained them much differently on one versus the other. So I'm seeing that question come up now. But just to the point on the data of what we trained our AI SDRs on, our inbound agent is trained on a lot more than our outbound, which again, maybe sounds counterintuitive for folks in the room and on the chat because you would, you'd think, okay, don't I need, don't I need it to know more to do outbound because it's cold? They don't know me as well. And maybe, but like, honestly, what we found is keeping it simple and adding value. The historical data has nothing to do with that. Like adding value to a person. Sure. Is it nice to know that they sponsored before or that they came to SaaStr? Yes. And some of our outbound does, does do that. But honestly, I found people switch companies so often, especially in B2B and SaaS. That a lot of the context we give our AI is almost useless because we have 13 years of data. But this person may have just started at Cursor or Anthropic, and they don't know that they spoke at SaaStr Annual this year. They may not know all these things. And so for us, we try and again, we've tuned ours like all the way to the max on the slider of adding value to the person versus it talking about SaaStr from like historic context or what we've done with them. On outbound. Inbound's a lot different. Like if they're coming to us, we've determined that as a signal of, okay, they're coming to us, they're having a conversation with us. In that case, because they're coming to us, maybe it is actually more beneficial to have our AI fill them in on other things their company's doing. And typically it'll be like, okay, your company's attended in the past, you've sponsored in the past, but also, hey, did you know your company's like been on our website a lot lately? Those kinds of signals are a lot better on an inbound where they already kind of know you and you're filling in the blanks versus like trying to educate them more so on the outbound side.

Yeah, that's a good— that's a great— that's the great answer right there. Very minor, but it was interesting. We, you know, we do use this app called Delphi as our general AI agent, right? It's the first one we did. It is trained on 20 million words of Saastr content, it's updated almost in real time every day. So this will be in— this will be automatically ingested into that general AI later today. Everything I wrote today will be ingested. So it is updated. But it was interesting. I wrote up this week, the CEO came and gave a presentation on a million chats between my AI, Brian Halligan from HubSpot, Keith Rabois, and Lenny's. And the, the, the captain obvious— one of the many captain obvious learnings was the value radically declines after X months. Like the old stuff that Brian did, the old stuff that I did. There is value there, right? But you almost want to repromote it and edit it, which we do on Saastr. I edit old content and update it because the value in that agent of old content is very, very low. So just like Amelia said, you don't need a trillion great emails that worked in the past. Just 1 or 2 is enough to get this multivariate engine going, right? You may need a smaller body of higher quality recent content. To make this work than as large a body as you think. And a lot of folks have come to our events in the past and said, oh, well, we could never catch up with SaaS. You have 20 million words of content. Well, the Delphi example is a perfect— you don't really need 20 million, right? You might need 20,000 great words of content that are recent on your product. It might be enough to make the AI great. You don't really need what— and we got this wrong in the beginning because we were very early here and we had our general agent everywhere. It was qualifying accidentally. It was doing customer support. It was answering founder questions and it was so good and it was calling back some old stuff. But I think we missed the fact that it, it wasn't, it was a little bit great because it has 20 million words of content, but it was especially good just because we were keeping it constantly updated every day with 3 to 4 new pieces of relevant content. I think that was more powerful than the long tail and I got that wrong.

Yeah, there's a, there's a lot of nuances. There's. A related question on when our AIs scrape data on the web or in our data, is it simply scraping or is it more dynamic? And again, depends on how you tune it, right? Like, most of these platforms on the backend will start to look very similar the more that you use them than different, quite frankly. And it's just for us, you know, again, on the slider of how simple we want it to be, we try and keep it simple. Like we have literally written in instructions to our agents to, you know, use no more than one value-add prompt ad, no more than one statistic. Here's the list of statistics. And again, we let it decide. And then it uses, again, either contextual data, if that's on the inbound agent from what they've done in the past, what they're doing now, or if it's the outbound data, it's based on what they're doing. In their websites, on their LinkedIn. So we let it decide. And granted, all of these have oversight, right? So whether it's our inbound, outbound agent, you can go into any of these platforms and see what it's doing. In the beginning, we used to spend a lot more time on this than we do now, because you have to refine the training and tuning. You might have to tell it, hey, I don't like the way you kind of present this, maybe present it that way. And you do have to keep it up to date, right? Like our agents, are on saster.com infinitely. They're, they're all on there. So like Delphi is sitting on there, it's recrawling the website every day. Our two other apps are on the website seeing what people are doing, what pages are they going to, literally in real time, right? And so keeping it up to date is a big part of it. And I think part of where I've seen people maybe get stuck a little bit with our AI SDRs recently, or AI for inbound is they try and, again, much to Jason's point earlier, set it and forget it doesn't work, right? Like, it's 2025, that doesn't work. I don't know that setting and forgetting an outreach sequence ever worked that well either. So that's not going to work here for the simple fact of you should want to keep your AIs as much up to date as possible. So like, anytime you have a new statistic, you should want to tell it that so that it has it in its knowledge bank. Anytime you make a website for the update, you should tell it to recrawl it, right? If it hasn't done so already. Anytime you've got a cool stat, a case study, like, again, these are things that are not dead. Like, all of our AI agents love our case study. We don't even, like, we have a few. We should have more. It's faster. But our AIs love it because they can reference it, right? So, like, to the point of GTM not being dead, anything you do that produces an either static output or dynamic output, whether it's a video, a case study, a PDF, stats, feed that all back into your AI. That's what we do because then it, again, it keeps feeding it and it can keep adding value to who it's reaching out to in your different formats if you keep it up to date.

And maybe just, Amelia knows this better than me, and we can move on to the next session, but maybe to summarize my high-level learnings, it's sort of all these tools are sort of scrape and augment. So they all have some sort of ability to scrape and some sort of RSS ingestion combination. And that, that's great. Like, set those up, right? Have it scrape your website, give it any RSS feeds, give it your YouTubes if it supports it. Like, have it ingest all of it. And they have similar ingestion engines. Some are better than others, some are more frequent than others, but they'll all ingest this stuff using third-party APIs, right? And then every day, augment it. Every day, augment it. Like, I know this is Captain Obvious, but in the early days when we just had one agent, which was Delphi, it was pretty good, but it made a lot of mistakes. And I know, guys, this sounds so obvious, but I realized we'd never added our prospectus. So I uploaded our prospectus. It was 10 times better with that. And then we learned every day augment, every day augment. And, and in the early days it's training, right? It's training and then it's iterating, then it's, then it's watching. It's like observing. But even if things feel good, right, then you got to think about augmenting. What could I augment today to make this better, right? And make sure whether it's been auto-ingested and augmented or you have to do, but more often than not, every week there's something good you can augment with the AI and have a material impact, like augment it. And so that, so, but, and that's why it's a, it's, it's not a full-time job, but it's a forever job.

Yeah. Last question related on this topic, and then there's a couple more on like SEO and Delphi and other things that we can move on to. Question from Kelly was how much time do we spend monitoring the AI? SDRs, interactions with customers and prospects? Is it more or less than humans?

It's more. Well, I'm gonna hear why, Amelia, why it's, it's much more, but I'm gonna say one thing and then Amelia will give you the, the deeper answer. It's more because it's actionable. We can, like the problem with us, like some folks have, are better managers and better teams, but our humans were a black box. It was a struggle just to know what they were doing, right? There's so much resistance to sharing data, inputting data. Amelia may see it differently, that because it's all measurable and we can see it, we spend much more time just because it's, it's, it's visible.

Yep. Yeah, I would say, um, it's, it's more time. So again, I, I think in the, in the onus of, are you going to get back time from hiring or bringing on an AI SDR, let's say, versus hiring a human being, you're not going to get any time back. The time is just gonna be spent much differently, right? Like, I spend more time with our AI SDRs than I did our human SDRs, because I'm using that time a lot differently. You've gotta think about too, I know Howard Lerman from Roam says this a lot, but he stopped doing meetings because they became therapy sessions. Sometimes they frankly are, especially when you have new hires or you've got seasoned reps who want more leads and that's all they talk about. And I, or even if they talk about that for 10 minutes out of 30 minutes of your one-on-one, your AI's not gonna talk about that. It's also not gonna really talk to you. Like the time I spend with our various AI agents is usually seeing what the responses are, right? Like, okay, how is it talking to folks? Like I go through our, as much as I can, I keep an eye on our Qualify, which is our inbound agent in real time, because I can literally see on the backend what are people saying across all of our websites is faster, what are they putting in the chat, and it'll flag to me, like, it'll slack me if it's not if it's struggling, but if it thinks I should jump in, if it's either like a lapsed customer because it's like, hey, do you want to talk to IBM yourself? Like, you, you could probably add more value as the human than the agent. But most of the time it's on autopilot and I, and I let it, like, I'll watch it talk to people. And I'll see if it'll struggle. And honestly, most times it doesn't because of how much we've trained it on. But again, the time I'm spending with that versus like a customer success rep, very different. Like I'm seeing what customers and prospects are seeing in real time. It shows the meeting booker instantly. So then I spend my time differently with our human sales team, right? To say, okay, hey, you got a meeting from Qualify or you got a meeting from Artisan. And they don't have access to maybe these tools in the backends the way I do. So I show them like, okay, hey, this was the conversation that our AI agent had that booked the meeting so that you have it for reference, that you can talk about it. They had specific questions they already brought up, like, let's make sure you cover them in your sales call. Again, you're just spending time much differently. I would say almost much more effectively because again, you're just cutting out all the time of, okay, hey, I need to talk to you about this new segment. Maybe I want to go after, as you know, for outbound. Okay. But for, for human, you've got to give them all that context. And yes, you have to give your AI all that context of, okay, who are you trying to go after? What are the contacts? But honestly, but it's not going to say, okay, but why are we going after these people? It's just going to do it, right? So the time is just spent a lot differently. And I would say I spent a lot of time I probably, again, I spend more time in as a whole with sales and our agents. But most of the time it's because I want the data to be really good. I want the responses to be really good. And these are things I just never even looked at before. So like it always, it, it does also just open up these new avenues for you of things you probably had on autopilot in the business that you have to be actually more hands-on now in the business. Because there's also a lot more of it, right? Like You guys saw the stats Jason threw up from our artisan. You are qualified is literally on 24/7 on our website. That is 24/7, which we didn't have. We didn't have like a chatbot before, so, or maybe we did in the old days, but like we now have conversation, like our agent can have conversations in theory 24/7 around the clock. So it's also just gonna output a lot more. So there's a lot more to review. So Yes, it's more time and it's spent a lot differently, but yeah, hopefully that's a helpful context on where the time actually goes when you bring on tools like this.

One other Captain Obvious comment. I was watching a demo the other day of a next generation AI SDR product that is designed for new founders that haven't done sales to take none of their time up, right? This is a next generation idea. It's interesting, but the, the, the, and what the reason I bring that up is, if I think about, if I think about one of the reminders from AI SDRs is founders especially vastly underestimate the amount of management time it takes for human sales reps. They, they don't understand how much time a good VP of sales and my God, a VP of outbound or SDR development spends with their team. It's a mammoth amount of time. And so in that founder-run stage, founder-led stage, very few folks can survive in that environment without enough oversight and coaching and therapy from— they just don't. You just don't have enough time. And you can do a lot of this with AI. In fact, you could probably do almost 100% of SDR, but you got to put in more time because you weren't putting in enough. Wait until you have a human great SDR, I mean, sales leader. You'll see they're spending so much time coaching, fixing, backfilling, reading, reviewing, listening to their team. and no founders have the time to put into it. So you're gonna have to make more time to Amelia's point to get this right, not less. This is, you'll get more benefits, but you do not get time back. You lose time. You lose time with these agents.

Yeah. Um, a few more questions if you have some more time, Jason. Um, we can do a few more. Um, a lot of the tools and insights from the presentation you mentioned were maybe later stage. Any advice for what still works in GTM in the early days? What do you see as core?

Well, I don't see— I think all these tools work fine early stage. I don't disagree. I don't see— I mean, if you have no revenue, go find your customers one after the other, right? But let's— I, I think all of this at some level works. Um, if you could— listen, if you're an early-stage company and no one wants to buy your product, we can't fix that, okay? But to use an old SaaS-ism, if you have 10 paying customers, your job's to get to 100, and 100's to get to 1,000, or 10 to 20 to 40 to 80 to 160. And so you just want to use these tools to attempt to improve and reproduce what already worked. It works better than you think. I mean, I know just folks think this stuff don't work, but let's imagine you have 10 paying customers. Use a little research on Claude and then pop it into your ASR. I just want to find 200 prospects that are like that 10. Okay. And then slow it down and figure out why the hell those 10 actually bought. We often screw this up as founders and just try to slowly reproduce that, right? Just reproduce that piece. It will work just as well at your scale as it will at mass scale. And again, Amelia made this point that I half missed until today. SaaS has massive amounts of data. We have so much more data than you have that you will— almost anyone will ever get. But that's not why we're so successful. We could be as successful with a tenth of the data. It's the volume of recent data that we have and the amount of training and constant updating that we do. The 10 years before that, real value, but limited value, right? You can collect as much data as we can if you go— if you— if you're serious about it in 12 months.

Yep. There's, okay, curious to get your take on this question. Tracy has 2,000 customer success calls in MP4 format from Google Drive, Zoom, Gong. He has a Delphi. He's getting started with Synthesia. What do I do with all these calls? In the age of AI?

Wow, that's a big question, right?

I know, I was like, that's a—

yeah, for CS, right?

Yeah.

Well, look, I mean, honestly, I, I know this is going to sound brutal. I would fi— I would use them to fire half my team. And what I mean is, uh, listen, what do you want out of CE? Like, do you want your customers happy? Do you want upsells? Do you whatever? I mean, I have had in the last I— if you— if you're old SaaS, you know, I was the longest superfan of CS since the beginning, telling everyone like, stop focusing on sales, make your customers happy. Customer— like, earliest stuff, we— at an early Gainsight event, we had an entire SaaS day on the SaaS playbook to customer success. And then I lost confidence in the last 18 to 24 months when companies that were slowing growth all wanted to dig the knife in and do upsell. And then when our core vendors, vendors that little SaaS spends 6 $60,000, $100,000 a year wouldn't help us solve any problems. And every CS call was an upsell from someone that didn't even know how to sell SaaS. Or I've lost all— I haven't had a single good customer success interaction in a year and a half. Okay. But I've had incredible interactions with solution architects, with onboarding specialists, with forward-deployed engineers. Okay. I had a— listen, we're at mass scale on Replit now, and I had a bump yesterday with Replit. It doesn't really matter why it was on the back end. and the team came in and just friggin fixed it at 6 PM. Okay. If I asked a typical customer success manager, there's just, I wouldn't even, all I would get was, do you want to buy the team edition? So, um, I would use it to figure out who's either truly adding value or who's upselling. And I would get rid of everybody and restructure your team. That's what I would do.

Yep. There's a, there's more of a comment. Let's throw a question. Um, in the chat and then maybe we can break on any final thoughts. But Felicia has said some of the problem too with training humans versus AIs is that you'll train them, you'll invest in them, and then a few months they'll leave you for a quote unquote better company. Yes. Yes, they will. Which my, my funny comment back was, unless you cancel your contract for your AI agents, it's not gonna leave you willingly.

So, yeah, I don't know. Yeah, this is one of the things I talk about that I think people miss. We talk about it. We want AI to like, we want to press a button and have it do things we can't find people for, right? We want to do that and we want it to produce massive uplifts in revenue. You know, the reason we got started on the SaaStr journey and I think people are missing is Amelia and I didn't want to use AI to make SaaStr better. We wanted to use AI because people kept quitting that didn't want to do the work. We, the reason we replaced our human SDRs with AI is not because we didn't, we asked them all to stay, but after a couple years they didn't want to do it. Um, they all went off and wanted to be VPs of sales and typically failed almost immediately or other things, but they just didn't want to do the job and it was exhausting to replace them. And so we just started replacing different humans. And then our content team like got tired of reviewing 2,000 speaker sessions each year, so they quit on us. And so we replaced that with an AI. And so we didn't do this to increase revenue or to save money. We did it because we were burnt out with people, just, just with change and especially post-2022 change. And so that's a start. But what's really the next wave next year is there just aren't enough people that want to do the type of work we need them to do. It's hard. People don't want to do it anymore. Young, medium, old, retired. It's just the honest truth. And so we're gonna be using more. We are, we're early on this on SaaStr. We're all gonna be using these AI tools, not so much because they're better, but because they are there, because they work, because they don't quit. Because you, you can't find the humans. You can't find anyone that's good enough to be worth giving those precious leads to. Leads are precious for many, many, many years. And it's better to give a precious lead to an, to a B+ AI than to give it to someone in sales that's going to quit, not show up, or screw up the deal. That's what's happening. We're going to use AI to replace the roles that people don't want to do, even if they pretend they want to do on LinkedIn, they don't want to do it. And that's why it's going to radically accelerate the next 24 months. So thanks, everybody, for all the time. We'll do another one of these very soon.