What’s Working Now: AI’s Real Impact on Sales with SaaStr's CEO and Co-Founder, and SVP & GM

So an interesting thing from all of these things, whether it's slides, whether it's QA on the, the SDR, which now can do 5,000 interactions instead of 50. But if you add it all up, it's more work for your overall GTM team. It's more work. It's more work. So I think what a lot of folks, especially CMOs that I talk to, or sometimes CROs that don't really want to get in the weeds, they're looking for solutions that are less work. They want AI to be less work for them, right? Why can't I just turn this? But it's more work. It's more work managing these AIs, right? You get higher output, you get higher quality, but you, but there's no less work trade-off here. That does not seem to exist in high-quality GTM AI today. More work for, for better output. Right? You might need less humans so you could have less humans. Yeah. But the ones that you do have are gonna be working harder. Hey everybody, the best.coms are taken or overpriced. I was literally looking for a domain name the other day, it was $10,000 just to make an offer. So you settle on a workaround domain for your website. Don't compromise, get a clean, sharp.tech domain that instantly says it's a tech startup. Grab yours at get.tech/saasster. That's get.tech/saasr or on domain registrars like GoDaddy. Stop compromising, get the domain you need,.tech. Hey everybody, SaaS Annual will be back May 2026, the world's largest SaaS and AI gathering for executives. Just as last May, we hosted 10,000 attendees with 68 VP-level and above attendees, 36% CEOs and founders, and 25% were AI-first professionals. It's the very best of S-tier attendees and decision makers that come to SaaS Annual and AI Summit each and every year. But here's the reality, folks. The longer you wait, the higher ticket prices get. They're cheap now. They're cheap. So just get them early. Lock in your spot today. Use my code JASON100 for exclusive savings. Get your tickets at podcast.saastrannual.com or just use code JASON100 when you check out. See you there. Saastr Annual and AI Summit 2026. It will rock.

Welcome back to Saastr AI Live. We thought we would do something fun today. Um, Since we've been using AI ourselves a lot at Dafter, I know you guys have seen Jason's tweets, you've seen some of the content we've been putting out. Seems like you guys have an insatiable appetite for some of this stuff. So we thought we'd share a little bit about what's working now. I've been using it a lot, specifically for sales, outbound calls, a little bit marketing. But we thought we'd share with you some of our results, what we're seeing, some of the pitfalls. And just to preface, like, we, even though it's faster and we do use a lot of AI, I would still consider us early on our journey. You know, we've used different AIs now probably for a year or so, if not more, in different ways. But I would say specifically in sales, we're still, and marketing, we're still early on, but seeing some good initial results that we thought we would share with you guys. So with that, we'll go and get started. Let's start with outbound because I feel like this is where we have— I just got the bad outbound email this morning that was from— that was, I hope, from an AI. Maybe it was actually a person. I assumed it was from an AI because it was so bad. So let's talk about outbound. What's actually working? What's actually closing deals using AI for real? So in the last 2 weeks, and that was at the time the only time we had been on this platform that we're now using, In just 2 weeks, little old Zaster became the highest response rate on the platform that we use for our AI SDR. We opened pipelines, we booked meetings, we touched lots of accounts. It both seems fast and slow, but it's better than historical human averages we had with any human SDR we've had, maybe even some of our human AEs that we've had in the mix at Zaster. These are screenshots here. If you can see how many leads it's touched, right? Like, there's no way I could send 4,495 messages myself in 2 weeks. And to do it at a hyper-personalized, hyper scale that actually is adding value. So you'll hear me talk about it a lot, but I think the overall consensus you'll see in my key takeaway is your AI has to be good enough to also add value if you want to see these kinds of results. Like, If you're doing, you know, AI that's just paid names, you on our website, you're probably not going to get these types of results we're getting at Zaster, even though it's early on and even though it's already doing better than any historical human averages. There's a lot that goes into that, which I'll talk about next. But just to preface there, it does take one, a lot of tuning, and it does take a lot of data and a lot of conscious effort to get these AIs to work this well. And have this good of an overall response rate, positive response rate, and also to make the emails seem like we actually wrote them. There are screenshots in this deck of people asking me two things. Like, one, is this a real person? I'm like, yeah, here's my LinkedIn. I'm real. I'm at Vassar. I'm on this webinar. I'm not an AI avatar. But also half the time they don't care. I think it's just a big learning too. Like, if your AI adds value to the sales cycle, Why does it matter if it's an AI versus a human?

Yeah, maybe just add— let me add one context while you're doing this just for folks. This is great. And well, these are good results. A few caveats for folks depending on where you are in the journey. One is we have a pretty big brand. Okay. So that helps. People have heard of us. We've been— Saastr has been around since— my God, since 2012. I've been writing content. We've been doing events since 2015. We've sold over 100,000 sponsorships. People know us and these 4,495 messages that Amelia's AI sent. And that is important. It's so many. Like, this is— we could never send this many good emails. We could send this many emails. Just, just, just send them. We could never send this many good emails. But we have a brand. And two, this is all to our base. You have to do a two-by-two. Do you have a brand? So is there some reason people will interact with your email? And two, is this, is this just raw folks you're emailing out of the blue? That have never heard of you, did not opt in, or are these folks that you already have some sort of relationship with? And I think, Amelia, all 449 of these folks are folks in our database that have participated in SaaStr, that have opted in in some scenario. So it's a pretty good set of folks to be hitting up. It's just, there's no way we could do this at scale with humans. Yeah.

Some of them are website visitors, right? So there's— and there's a lot of AIs that do this now. So there are folks that are on your website.

They know us, but they know you. So that's the key. So I just think these are some good learnings just to provide the context to you. You know, if no one's ever heard of you and you have no base to email from, no matter how well you do, you're not going to see these results. You may see— you should see results. But I think that's one of the many reasons. There's probably 3 reasons we've done so well, like off the charts well. Brand, our own base, and lots of training and lots of iteration, not just turning on and walking away like most people do. So just, just for some context, those 3 points. But yeah, keep going. Nope.

Super helpful. You talked about it. I wanna talk about it a bit more in depth. It's like how to train it. I already see some comments in the chat. Here's, there's a lot of implications here, right? I literally was just emailing with somebody today that was like, I'm coming to the webinar. I have a quick question on what can I do that's quick? I'm like, it's not quick. That's already one. That's maybe like the wrong mindset. Like if you think you're gonna turn on the AI SDR and it's gonna give you the same results that we got, it ain't gonna happen, dude. Like, A lot of training. I would say I have almost dropped a lot of other things I should have been working on. And I think Jason, you were talking about yesterday on Twitter, like you don't spend as much time in some of our AI as you used to. And it's becoming the same now for this particular sales tool we're using for outbound for me, but it still does take a lot of time. Like at the very start, you should expect to spend the most, like basically all your time with it. Like these 2 weeks were also the 2 weeks I spent the most time on the AI SDR. I spend less time with it on the outbound now because it got Right? That's the thing that AI will get better at the outbound. But if you don't have 2 weeks to onboard your AI, just like you would spend 2 weeks onboarding a human being, then don't do it. Like, you literally need to spend the same amount of time, maybe more, on training the AI that you would a human being. Like, you gotta train it. You absolutely gotta train it. That's one maybe trade-off or nuance people don't realize yet, but I would say the nuance of that being you can do it when you want because it's not a real person. Like I would tune my AI first thing in the morning and then last thing in the evening, right? Probably 90 minutes in the morning, maybe an hour in the evening, maybe more. And then I would respond throughout the day pretty much in real time if I could to any responses that our AI was getting in real time. But that's the upside of the AI is not a human being, so I don't have to set 10 to noon Pacific. To do the training that I would a human being, but it's still the same amount of time. So pros and cons. But let me get into one of the biggest things is you got to train it with all your data. And this has been, I would say, even for us, a little bit of a hurdle. We have so much data at SaaS here. Like, you guys have probably talked to Jason's AI on sas4r.com. It's got something like 20 million words in it, I think, and counting, right? Because we post new content on SaaS every day. And that's one data set for us, right? Like if you think about, or if you have multiple product lines, or if you also produce a lot of content, or if you've been around a while, you're going to have a lot of data. That's just one set. That's just saster.com. That's not every data signal that's in our Marketo, which is millions upon millions upon millions. That's not every data signal that's in our Salesforce, which is probably another couple million. Like there's so much data you have to train to work with your AI. And so, you know, I do think you'll start to see more RevOps people become AI RevOps. But I get why now, after spending like almost 2 full weeks with our AI, it takes a lot of time. And like to do it right, you do need to know the business. Don't hire somebody off the street to do this for you. If you hire someone to do this and you don't do it yourself, it's got to be somebody wicked smart that actually knows how to use AI and actually knows your business to feed the right data inputs to the AI.

On our AI, our general purpose AI you can chat with on the homepage, you know, it's trained on almost 20 million words of Saastr content. And to Amelia's point, for the first 2 months up until Saastr annual issue, every morning I would wake up, I'd get my coffee, and I'd spend 30 minutes training it. And training can mean different things. Like, I'm not an expert. What for our AI, for the base AI, not for sales, which Amelia didn't do. She can talk about how she trained it. For our sort of generic AI, which is pretty good. We've done 100,000 chats on it. It had to ingest 20 million words of SaaStr content, all of our SaaStr annual sessions, all my tweets, all of our YouTubes, all of our videos, everything. But they're like very specific questions were really good because it had ingested all that data. Where it would hallucinate or make mistakes is where it just wasn't in that data. For example, if you asked it about a speaker that we had at Saastr that actually spoke, what does Stewart Butterfield say the 2 times he came to Saastr? It'd be great, right? Then if you asked what the CEO of Notion said at Saastr, who never came, it would make something up.

It would make it up, yeah.

And the worst, the most— so those are a little funny. And the most extreme one was we had not published the dates for Saastr 2026 or Saastr AI London this year yet. And so it didn't know the date, so it would make them up. So what I would do is every minute I would just audit the chats, audit the questions for 15 or 30 minutes. And there weren't that many errors, but the ones that were, were egregious like that, right? And then I would go in and I would manually write the question and manually write the answer. Okay. In the training, in the training section, when I did that, it wouldn't make the mistake again. Unlike a human, like a human would make the mistake the next day. Right. But I had to do it literally for almost half an hour for 2 months. Every morning. Every morning. And so now it— now I do it once a week. Okay. And but I think that's a metaphor for even if you have all your data in, even if you have it all in, you may have to QA it every single day for the first couple of months.

Yeah. And I will say there's on every AI SDR tool, all of them have different rule sets. This is trained on most of our data, not all. There's a lot, right? So like some of our AI is trained on more of our data, some AI is trained on like the core of our data. Doesn't mean you have to train on all your data, like more context is good, but at the end of the day, what your AI is looking for and what you really need to train it on is what do you want it to surface to this person? Okay, if you're using it for outbound, but you're using it for, let's say, warm outbound in the way we are, or people who are visiting your website, who are already on the website and kind of know at least a little bit about who you are and what you're doing. And then basically what you want to tell the AI is, okay, if they're on the website, what are they doing on the website? Okay, great. The AI will tell you that right away. It already knows what, what this person's doing on the website. Next, okay, let me go look in maybe your CRM or your marketing automation platform. Maybe I don't necessarily need both to see, okay, is this person, let's call it a VIP account, like have they been to SaaStr? Have they sponsored with you? Are they already a customer? Don't need to give it every single email. You've had with this customer and every closed loss opportunity. Half the time, let me tell you, the people on the website might not know that. They might be employee number 10 at Oracle who doesn't know all the history and context of SaaS and Oracle from back in the day. And like a few words of that context from your AI is nice and handy and it is personalized and it will get them to open and respond to your AI and probably take the call. And that's the level you need. So I think you also need to just think about the mindset of Okay, a lot of data is, but it doesn't need to be perfect. Like it does not need to be perfect. And we started to hyper-segment more and actually like rip out some of the data from this to make it more hyper-segmented and actually a little bit more simple. And that is part of why it's also working so well for us. So just to train your data, if you will, to get this to a working state, I outline the 6 steps you'll need to take in the early days if you're serious about doing this, right? So one, yes, you're going to need to start with your CRM, your marketing automation platform data. Hopefully that is in a working stage in which you can use. So just keep that in mind. If it's not already living somewhere, it'd be hard for your AI to adjust it. Each tool kind of has its own way where you can say, okay, make an artifact and say, okay, hey, by the way, IBM wasn't sponsored this year, it wasn't in Salesforce. And it's like, okay, I've recognized that data kind of like you would in your brain, right? Like the same way a human brain works. You can tell the AI's brain like, hey, this— let me do a brain dump of my human brain to the AI brain of things I know that are in our CRM. You can do that. A lot of these platforms have a way you can do that. Make it good to do that at least once. So it won't know what you don't know. Hook up any of the major— most of the tools hook up to all the major platforms. Some of them are listed there. But yeah, you're going to need to probably do some data cleanup to varying degree based on where you are and how much data integrity you've had historically. Not to say you can't start with zero. It's just to the point Jason was making earlier, it probably won't work as good if you start it from zero. So if you're like, I know I'm behind on AI, I probably need to get started with this, but I don't have a lot of data yet. You might want to spend the next, even like 6 weeks of data might be enough for it to get up and running and going with something, thing. So it has some historical context to help your open rates. I think otherwise it will start to default to being too generic, and that's where you'll see like lower open rates and like lower positive sentiment scores. The other thing I did as part of this journey, a lot of where we spend our time now that we did an additional data cleanup, is enriching our current data and those warm leads, the warm outbound. So like people on our website, people coming to SaaS events with fresh data, right? So There's different, there's like, I think 20 different vendors that do this. Obviously a good way is to scrape your website itself and see what you have. So maybe in your journey you might need to do step 2 more than you need to do step 1 to get all the fields you need to work with the AIs. So the other thing I'll just mention is a lot of these different AI SDR tools have certain requirements of fields that you will need to have in order for it to send the really good emails. I mean, each of them have kind of weird quirky certain fields they need, and that's because it wants to 100% make sure it finds the right person on the internet to say, okay, this is the person that you are trying to prospect. Like beyond a certain of a doubt, you know, that is what the AI is trying to do. So there's a good reason it needs those kind of quirky certain fields to do your prospecting and outbound for you. It wants to make sure it's doing it to the right person, not like, you know, I typed in Amelia Larue, it's faster. There's only, I think, one of me on LinkedIn. But if I've got a more common name like Joe Smith and there's 10 of them, well, this is why the AI needs all those fields so it can make sure it's expecting the right Joe Smith at the right company. The last thing you want is for it to send an email to the right person but wrong data points of, oh, hey, I saw you're at Fox, and actually that person works at Dropbox. That's what the AI is actually trying to avoid. So that's part 2. And then yeah, step 3 would be actually thinking through before you start a structure for the AI. 2 and 3 go hand in hand, right? You may need to even take it a step further to structure your AI SDR. I'd also start small. So it's not something where, you know, you want to necessarily start blasting everybody that comes to your website in real time because you can. Doesn't mean you should. What are some of the things that perform the best is, yes, what do they do on our website? What are they doing on SaaS? Or have they been to an annual? Have they been to Europa? Have they been to all these workshops? Um, what stage are they in? All those things like that. Recent funding news, job title changes, pain points, like some of this that will scrape for you automatically. All this to say on points 1 through 3, you're trying to get your AI to a point where it knows enough about the exact person you are trying to reach, that it will write a really good email that you would've otherwise had to spend probably, let's say, 10 to 15 minutes writing yourself. That is the whole point of training the data and getting the structure right, is to write the email good or better than you would've yourself, obviously at a much faster rate, at a much higher scale, right? Like I couldn't email those 4,000 people myself in 2 weeks. So we talked about it a bit, so I'll keep going. But yeah, just for hyper-segmentation, you know, it's been true of marketing since before. The more you can hyper-segment it, the better it will work. Ask yourself, what have you— ask yourself, what has this person done with our company lately? Have they been to a customer dinner? Did they go to the website? What did they read on the website? So ask yourself, what have they done with you lately? And how does that tie to them? Maybe showing intent to buy, or could you get them, you know, down a path where they have an intent to buy? Jason mentioned it, it's one of the caveats. I have it here as number 5. Prioritizing what I call, what we call warm inbound always works, right? This is always going to work better than any like old list you could scrape off the internet and say, hey, AI SDR, I just got this list of 100 companies that are the top 100 companies start emailing them. That's probably going to work as good as you giving that list to an SDR. That's a human. Say, hey, start emailing them, which is probably not that good. Don't know you or you haven't added value to their company. Again, at the end of the day, you want your AI SDR to know exactly who you're targeting, have some context with all your data of what they've maybe done with you and what your value add to them is so that it can write an email that will start to add value to the person. You can call them a prospect, you can call them an account, call them whatever you want. At the end of the day, if your AI is adding value, people will not care it's an AI.

Amelia, maybe a question on this. With your AI, you're basically trying to do two things today, if I have it right. One is, um, reengage folks to buy a ticket and come to our SaaStr events, right? In London in December and May next year. And two, reactivate folks to buy sponsorships, right? I think we're running, are we running both?

We're running both. Yeah.

How do you do both at the same time? Is it two different instances, two different orgs, two different iterations? And so, like, how many variants of this AI SGR are multitasking, are running in parallel?

Yeah, it's a good question. I would say in the first iteration, right, because this was an experiment for us. I know we're all— you're all in on AI, I'm all in. We all agreed we should try this. Before this webinar, listen, there's not a lot of content out there on how to do this for real. So I just started it. I was like, okay, let me use the marketing principles and sales principles I know since the dawn of time, but also let me just try different things. So I started it on 3 different experiments, one for each bucket, one which was a totally cold outbound, literally to a list of people. It was a mix. Some people knew SaaSers, some people did not know SaaSers, because I wanted to see how the AI would do. Then I did lapsed accounts, right? So these are folks that have known SAFSTER, but maybe haven't engaged with us in a while for sponsorships. And then I did one, then I did one for tickets of people who have been to a previous SAFSTER event, haven't been to one in a while, see if that would convert. And then fourth was I had the website one on in a version that I don't have on anymore, but it was like basically doing some light qualifications. Of folks on the website to see what would happen with the AI. Yeah, I would say across the board, results are better or worse in each of those 4 buckets. Probably most relevant for folks on this call, as different product lines, you do have to kind of force rank what product lines and different campaigns you want to run with your AI, because already, like, my inbox is out of sense. Like I have, you know, I think there's like 30 personas I have on our tool and I'm already like maxed out. And so I literally have David from our team joining now who is now, I'm segmenting some of the sends through his inboxes because I'm like, okay, I'm going to use mine for certain campaigns. I'm going to use his for certain other campaigns to work because even the AI, as fast as it is, as much as it can send, even for us, was not enough. I was like, I already need more. So there's that. Before we move on to things like other ways in the sales cycle than, than pure outbound is don't expect the AI to do it all for you. I think when folks turn on, or even if they make a conscious decision to buy something like an Artisan or something similar, they're like, okay, I'm going to buy this, I'm going to use this. I'm all in on AI, like it's going to work magic for us. Even at that stage, I would say don't expect it to do everything, right? Everybody has different data signals. You have different campaigns you're trying to run. It doesn't mean it's a genie and it's going to grant your wish right away of like more pipeline, more closed won deals, more reactivated accounts, more sales than the way it has for us. Like it may not work that way. So just know that going into it. It will require a lot of tuning and training. The other learning was once we turned this on and people started responding to our AI and we had a human in the loop, which was me, and it for a while, and it worked the best. You have to respond instantly.

This is a top learning, right?

This is a top learning. I was like, oh, because the AI is so good, the expectation is that you literally respond. Like, if you're going to be the human in the loop and respond, you got to do it one, right away, and it's got to be good. Like it's gotta be at the same level as the AI or they're gonna be like, oh, was this just a bad AI email that I got? It's gotta be so good. So I have it hooked up to Slack. And honestly, SaaS is global, right? So sometimes people are in Europe. I'm like, dude, that's like, well, at night my Slack and phone is going off and I'm like, eh, okay, send. It's just, you have to think about it constantly now. So yeah, that's, that's a big learning. I think that's something where like mentally you might want to prepare yourself for because it's a lot, right? It's on top of everything else, on top of the humans on your team slacking you, the actual emails you get from other human beings. Now your AI is like, hey, you should probably respond to this person. They, they need info that maybe I don't have, or you could send a better response than I could, right? And the other thing, Jason, you've talked about this a bit, and we've all been victim to getting a bad AI SDR message. Everyone at this seminar probably got one this morning or in the last 24 to 48 hours. I think folks are like, eh, I don't really want to use something like a tool because I'm just going to get that. There's weird, like, two camps I've kind of seen lately. Either this bucket of the AI is going to do magic for me instantly, which, hello, it won't. It can if you train it. Or the AI is going to suck and like gonna send super generic emails. Hey, have you heard of SaaS? Or like Austin would never send that email. And so I think there's kind of like these two camps, but if you do it the right way and maybe follow some of these steps I've hopefully helped outline, people will get over it. I have noticed, like, I can't tell you, I put some of the screenshots in this deck. But I have many more on my computer, in my inbox, of people being like, I actually don't care if this was an AI or if this was you, or if it was a mix of both, but because it added value, it doesn't even matter. So like, again, if you can get your AI to be decent to the level of you yourself, that's probably the quality you should aim for. They will prefer that to, you know, an AE or SDR that is one day in at your company sending, hey, have you heard of SaaStr? I'm, I'd rather get an email from an AI or me AI that knows Zapster, has been a Zapster, than someone brand new that actually doesn't know the product. Like, Jason's talked about this a lot. The AI knows the product cold, right? Like, it knows the product. That is something it will not hallucinate on. Like, it knows your product. You're telling it your product. You're hooking it up to your product, your pain points, your websites. I'd rather get that email. If it's really good, I'd rather get that email. Then hey, have you heard of SaaStr? The maybe other biggest learning is the AI alone is not enough. So I put it here 3 times. Hopefully it's clear by now, this requires a lot of human involvement. I know a lot of the CEOs from the different AI companies will say that too. Like, yes, we're even at AI companies, but we know like some of this you can't automate away in 2025, maybe in a year or so if we can, But right now, like, still at the end of the day, if you're on the exec team, if you're the CEO or founder, you will still know the product the best, and you can tune the AI to get pretty good at it. And sometimes the human response that I give it is better than what the AI response would've been. And so I override it. Like, you know, the AI will tell you, hey, do you wanna send this response? I'm like, I could send that, or you could send that. Like, and I just send it. So the AI alone is definitely not enough. You need a human in the loop. Which does take some time. So again, lots of things to think about here. As you're implementing this, you gotta commit just like you would a brand new hire. Like, it takes time. Like, you, if you're gonna invest in a human, you gotta invest that time in the AI. Just 'cause it's an AI doesn't mean it doesn't require the same training, respect, and level you would give a human being. Maybe it even needs a little bit more respect because it knows everything. Okay, a few examples here that I redacted. So yes, don't send this email on the left. This is literally one we got today too. Okay. Hey Eve, how's it going at SaaSer? I did a bit of research. I thought you might be familiar with blank, our company. This person worked at that company. This is an example of bad, bad, the worst. Hey, have you heard of SaaSer? Hey, have you heard of my company? Don't let your AI do this. All right. If you, if this is the output your AI gives you, dump it and find another tool. Like even with only a little bit of data, it should not be sending this. I assumed that was an AI SDR email. Hopefully not. I, I shudder to think this guy actually wrote this email because I don't think he read it. So I shudder to think, but don't let your AI send the email on the left. These are screenshots of emails. My AISDR sent on the right. I did 3 different ones. They're kind of across the 3 different, like 3 or 4 different buckets I mentioned earlier that we're using the AISDR for. A few different things here. You can see it is customizing it based on them and our data. One of them's a VC. It talks about the VC programs we're going to have this year. I don't have time to write that email to that dude. Or the second one, hey, congrats on your new role. It saw that on LinkedIn. It also saw from us. That they had already been to SaaS before, but they moved on to a new role and I didn't have their new email, but it found it for me and sent them this email about SaaS in London. These are pretty good. And then from here we did a mix of automated responses and then human response. The other caveat being some of this also worked really well because we also still did marketing air cover. So like this is all to say, don't just turn on the AI SDR. For, you know, outbound accounts or reactivating lapsed accounts. A lot of this was a mix. Like I would go in and say, okay, reactivate these lapsed renewal accounts. But I also still sent an email from me personally to some of these folks. And then I also still sent an email from, you know, Marketo to some of these folks that was a little bit less tailored, but it didn't need to be as tailored because I was like, I already know I have emailed them, my AI has emailed them, and this is a third additional touchpoint. Don't dump this for everything. Again, it's not a magic wand, it's not a genie. Don't say I'm going to turn on the AI SDR and I could dump everything else I'm doing in marketing or fire my whole marketing or fire all my SDRs. It doesn't work that way. These are just 6 kind of combined touchpoints. 3 to 6 for us right now has been the right mix. It may end up being less, I think, as the AI gets better. This is just additive. Like, this has been additive to our journey and what we've done at Zaster. And a good addition, but we didn't dump anything. So just a few other touchpoints. I, in also this AI journey of, you know, our AI did book calls for us, but I was like, okay, where else can I plug in the AI? Because now I have these meetings booked. You sent all these really great emails. Let's say your, your AI got you the call as it has for us. What do you do now? Right. As the humans in the loop, we still have to get on the call with them. And you want to basically know everything your AI knows. And if, and the AI knows everything. So I've come up with this phrase that I call it, instead of discovery, I call it a working theory. So instead of coming to a, a call to do discovery, you come with a working theory, right? You, before the call with your team, you can do this internally. I literally just did this with David before a call we had this morning. We pre-researched them using AI. We made a working assumption. On who they are and what they do and why they might be interested in you, right? You will need to do this working theory more and more the more the AI sets up these calls and meetings for you, because you yourself may not know as much about this company as your AI does. But if it got the meeting for you, the person who comes to the Zoom will assume you know what your AI knows. And so to match that, you do need a working theory. So, um, I called it the 4 C's. It's to do your research before the call. Okay. In chat, like, did they inbound? Did they chat with our AI? Like, what is the discussion, man? That's the first thing. Claude, use Claude, use ChatGPT, figure out as much as you can about this company. Literally ask Claude or ask ChatGPT right now, like, hey, I've got a meeting upcoming with XYZ company. Why should they buy my product? It will tell you. It will tell you in 20 seconds. Great. Now go to your CRM. If you could talk to your CRM, great. But like, look into your CRM and say, okay, what has this person done with us previously? So just a, it just is a way to have better calls, but also it's a way to just keep up with your AI. Like if it's booking the meetings and it already knows all this about the company, but you don't, you might have to do a little bit of human catch-up. If you also come to the call with a working theory, you will need to validate or refine that, right? Like you're going to have a working theory about why they maybe want to buy your product and what they might need it for. Maybe you know some of their competitors that are already using it. Maybe your AI knows that, but obviously you need to validate those, right? Like no company ever is gonna be like, oh man, so like you came to this call like knowing too much about our company. No, no one's gonna say that. Your AI already knows everything. You should know everything. Have a working theory. Think about why they might wanna use your product. And also you can ask chat, you know, you may need to use things like Apollo or Seamless or ZoomInfo to say, hey, also, who's on their buyer committee? Who's on the call? Let me do, you know, 5 minutes of research on them, ask a few questions and understand all that before they get to the call and then validate that. It's so much more productive, even the calls we've had recently in the last like 2 weeks, to do a working theory and just validate those points versus waste 10 to 20 minutes on asking the human beings questions that, one, your AI already knows, or two, they don't want to answer because they're like, you should already know this. So catch up to the AI.

Okay.

Last couple other things you can use it for. Maybe you guys are like overwhelmed because I spent so much time talking about data and you're like, okay, but now each new call I have, each new like interaction the AI has, what do I do with that? Push it either back to your AI, push it into your CRM. There's a lot of different tools that do this. I listed them here. Some of them do it automatically. We use some of these tools. I will say each incremental tool needs its own training. So the more you add, like, start with one. Do one thing at a time. If you're like, okay, maybe I'm not ready to go outbound, but if you want to start small, you can be like, all right, I'm going to start to use the 4 Cs. On any calls we have upcoming, I'm going to make sure we have it recorded. And at the very least, I'm going to start using the AI to do research on this person before they come into the call. That's a good stepping stone. I would say too, the AI is pretty good at following up. Not only does it know your product goals, it will not forget anything. These AIs, man, they're so good. I'll be like, hey, um, write me an email. Like, the ones we use do it automatically, but you can also do it manually if you need to, if you're starting small and say, hey, Listen to this call, write me an email to this prospect. It will do it for you. Even if you're like, okay, I'm a ChatGPT stan. If you use the record mode, which is on the desktop, right? So it's not on the web. If you use the desktop version of ChatGPT to listen to your calls, one, make sure you have a present recorder just because that one is invisible. So you might want to tell or, and/or tell the person you're recording them. But then at the end of that, you can ask chat, like, again, this would be a really easy step that you say, Hey ChatGPT, you just listened to that call. What should the follow-up email be? It will write you a really good one. The other thing we've started doing is I don't send people a generic, like, perspective. You might have a sales proposal you send at the end of each call. They might ask you for a proposal and pricing at the end of each call. You're probably like, oh yeah, here's the slide. I'll send you the slide. I don't do that anymore. Now I have AI and I'm like, okay, this is the company I just talked to. This is the pricing I want to give them. Make me a customized proposal. One, it could be a one-sheeter, it could be a full-on flight deck. I'll show you two examples we've used recently. I've been using Gamma and GenSpark for these kind of as a mix with other things, like with other things we have here too. But this is what, this is my level expectation of like, if I were being sold to, this is what I would want at the end of my call. Like, AI can make you a fully custom Boost meeting sales deck to send to this person literally in minutes. It does take a little bit longer than like ChatGPT, they have a profession, but it does just take a couple minutes. Literally in minutes, it can make you this. I sent it to them. They're like, oh, this is great. I'm going to use this with my internal team. I'm like, yeah, why wouldn't you? It's literally a fully custom deck that I did not have to go ask our designer for and wait 2 weeks to get back. A fully custom deck on what we just talked about. Who wouldn't want this? Literally, this person on the right was somebody who was known to Zapster, had some interactions with our AI. I did a human meeting with them and then asked me for like custom booth options. I didn't want to. I was like, "Ugh, I have to go through like my photos, ask our designer." I was like, "Nope, I'm just going to ask the AI." I'm gonna, I literally uploaded a few photos, which took a couple minutes. I uploaded our current options of sponsorships, and then I said, hey, this is what this person is trying to achieve. Can you make them a custom slide deck of each of the different levels, the pros, the cons, the design element? It did this. I didn't write design elements, natural element. Like I uploaded the, the image to Google Cloud and that was it. And I told it what company. Was interested in custom sponsorships. That was all I did.

Bigger organization, your average sales rep is just going to download whatever piece of collateral they get.

I know.

Sometimes months old if there's no controls. How would you, if you want to make that collateral literally an order of magnitude better using your workflow, how would you scale it? Would you, you, would you have a process break where I can't imagine most sales reps can do this. That I've worked with. So how would you implement it with some sort of marketing ops or sales ops support? Because this could be— this does— like, you just generated this with AI. Yeah, but you have the assets and you have the domain expertise to QA it and make it correct, right? Even if it took you 10 minutes.

Sure. Yeah.

How do you product— maybe you don't have the answers today because we're learning, but how would you roll this out if we had 20 reps?

Yeah, there are certain AIs you can use now. It's a Slacky. Every time you've had a sales call, It will give you the insights. It'll give you what they talked about. It'll give you what they asked for. Like if they ask for pricing, if they ask for custom booth options, and then I would just make it for them. Like I would just take all those notes from the AI, right? Pop it into either Gamma or Jonesboro. Could do it with your own thing and send it back to the AE in that same Slack thread so that they have it.

Okay, but this might take you 15 minutes to do a good job of today. Would that be accurate? Okay, so if your sales team is sending out 50 of these a day at scale, that's a lot, right? So it's a person. So I think you need an AI marketing ops person to do this. And I also think the other learning, I think, yeah, from dude, I don't think a library is good enough because a library is static. That's the problem today. Everyone has libraries and there's ways to control it. No. The point Emily's making is every single prospectus is dynamic and customized to that human. So I don't know, maybe we're at, I don't know, I'm sure people will challenge this, but in our experience, I don't think in July of 2025, we're at the point where if you want to have super controlled, high quality communication with prospects, you can just let sales reps go free and wild here. I don't believe that exists today. So an interesting thing from all of these things, whether it's slides, whether it's QA on the SDR, which now can do 5,000 interactions instead of 50. But if you add it all up, it's more work for your overall GTM team. It's more work. It's more work. So I think what a lot of folks, especially CMOs that I talk to, or sometimes CROs that don't really want to get in the weeds, they're looking for solutions that are less work. They want AI to be less work for them, right? Why can't I just turn this on? Even if I need a day, a couple hours of training that I hire an agency for, they think it's less work, maybe, but our experience is it's 10 times better, right?

It's more.

Maybe 10 times the product, the output, quality output, not just crappy output, quality output, but it's more work. It's more work managing these AIs, right? You get higher output, you get higher quality, but you, but there's no less work trade-off here. That does not seem to exist in high-quality GTM AI today. More work for, for better output. Right? You might need less humans, so you could have less humans, but the ones that you do have are going to be working harder. But if you want to do what I think we're doing, which is S-tier, everything's going to be hyper-customized. And I don't know that the component approach is going to get it to that level of excellence. Um, but it certainly could improve it from a static level, right? It's classic static content. It's just a challenge. It's just a learning.

Okay. I think they actually get you most of the way there. Um, but that's why you still need— it's not enough, right? Like it is good, but you still need. Humans in the loop to just do even a better of a job with it.

I know we're all using these buzzwords. I don't think it's humans in the loop. It's human orchestration. This requires human orchestration. It does not require some crappy human or agency that reviews the deck for 5 seconds and says, great, and they flip through it and don't even pay attention. It requires orchestration, high-quality orchestration, even with no other humans in the process. To create S-tier output, you need S-tier orchestration to create S-tier output. That's the learning. So ask yourself, if you're not happy with your AI, do you have top-tier S-tier human orchestration? I bet you don't. I bet you don't have top-tier orchestration. That's what we've learned. You need top-tier orchestration. Then pick a tool that you like. At some level, I don't think it matters. Like, there are— we love the tools we use, but honestly, you could pick a slightly worse tool, but if you invest the time and effort, it's going to be dramatically better than what you have today, right? The key is investing in the orchestration. And the QA, right?

Yeah.

And no, no grab-and-go libraries handed off to sales reps that don't know how to spell product. It's not going to get you there. It's just not going to get you there. Right. I mean, really, I was on a demo the other day with the head of revenue at an AI leader coming up on $100 billion in revenue who did not know what cloud was and do not know what MCP was. Okay. I don't care what sort of like dynamic content you create for this, for him and his team, they're not going to know anything. You have to actually go further and do it for them with this orchestration so that the presentation is great.

I am going through the chat real quick and see if there's any rapid fire questions we can answer in our last few minutes. Any last things you want to wrap on, Jason M. Lemkin?

I think you hit it. If I summarize all the learnings from our own AI for chat to SDR to collateral generation to BDR and other things we're adding now is pick a platform. It, right? Um, and then the first, like, you got, you got to make sure you trained it properly. And at a most basic level, audit it. Like I say, but audit it. Don't get frustrated or say it doesn't work. Like, every day spend an hour and read the email, look at the collateral. And then sometimes it can be frustrating, but slow down. Why is it not working? Why is it not working? Like, we had an issue on our little team. One person on our team just did not hook up their SSO to one of the apps, no matter how many times we asked. Every day, said it was working. And then I made him do it live on a Zoom and it finally worked. But like, that's a QA thing. Like, why, why is it not flowing in? Well, he actually never actually hooked it— said he hooked it up, never really hooked it up. So like, every day— but read these emails. Like, when we started at the beginning, Amelia showed a really dumb AI-generated email. You read them and you don't have to read every one, but just do a statistical sample. Do what you can get through in 30 to 45 minutes each day. And then Take a pause and figure out how to fix it. Learn whatever tool you're using. It's early. All these tools have limitations, right? Some, they can't work on multiple domains. They can't work on multiple platforms. Almost all of them can only do one or two workflows. They can't do everything you want. You might hack it, but learn your tool's limitations and figure out the next day how to make that workflow better. What was wrong? What was wrong with, why was that communication dumb? And then just research it. You may not want to, you thought, maybe you thought AI would do it all for you. AI is magic, but it doesn't do all the work for you. And then every day make your AI better. If you do that after 60 days, you're going to be happy with what you have. Every day make it better, right? But you can't, you can't just assume it's going to do it on your own or assume you can hire an AI dude to do it for you. That's my summary.

And we'll see you guys next week. Thanks for coming. Thanks for watching.