6 Months Later, How Our AI SDRs Actually Work as AI Runs GTM with SaaStr's CEO and Chief AI Officer

Welcome to the official Saastr podcast where you can hear some of the best Saastr speakers. This is where the cloud meets. Up today on the Saastr podcast. Don't fire anyone good to replace them with AI if you haven't learned anything. If someone's failed, if they can't close anything, or if your AI SDR hasn't set a single appointment, I mean, your human SDR, maybe just replace that budget with a, you know, if your SDR can't do anything, you know, you can't do worse than zero. But we'll track it. Many of you, if you're early stage folks watching this, you'll think that's expensive. And if you don't believe in the vendor, it will seem very expensive. It will, because you'll get quoted by a sales rep a lot of money and it will feel risky to you if the deployment doesn't work. You know, we, that's why we're not trying to be walking billboards, but we do share the vendors that we use. Others are good too. But if you get the right people, it's going to work. But it can feel risky. It's tough to start at $500 a month today, but you got to take a little risk in life. Hey everybody, it's Saster. Connect data, automate busywork, and empower teams like nobody's business with the one platform that grows with you every step of the way. Learn how Salesforce works for startups at salesforce.com/smv. That's salesforce.com/smv.

Hey SaaSer, imagine having agents for every support task. One that triages tickets, another that catches duplicates, one that spots churn risk. That'd be pretty amazing, right? HappyFox just made it real with Autopilot. These pre-built AI agents deploy in about 60 seconds and run for as low as 2 cents per successful action. All of it sits inside the HappyFox omnichannel, AI-first support stack. Chatbot, Copilot, and Autopilot working as one. Check them out at happyfox.com. All right. Hello, everyone. Welcome. All right. Welcome to everyone in today listening or your agents that are also listening and you'll be reading the recap later. We thought it would be fun to do a deep dive into where we're at today in our AI journey. It's kind of crazy to think before SAS for Annual this last May, we really only had one agent that we had sort of just deployed. And now we have about 20 or so core agents, which we've gone through, but I'll go through a little bit as well for context. So it's crazy to see, you know, just kind of from May till November, you know, post-Annual, now that we're fully in all the platforms. We've added a lot more use cases across go-to-market, which you'll see, just to give an update of where we're at, what's working. There's some nuances on what hasn't worked, and there's some maybe unexpected learnings there that we'll also go through. But hopefully it's helpful for folks just to listen and hear and see where we're at, share our learnings and our findings. Hopefully this helps you as well. But If you have any questions as we go along, put them into the chat. We'll try and do a bunch of questions at the end, help answer you guys' any deep dive questions too. So with that, I'm gonna try and keep admitting you guys. Let's go through it. So what do I mean by 6 months of AI running our GTM? Now it's not on full autopilot, so it is not just that the AI is running amok with our GTM on its own. It does require a lot of oversight. Insight and time and management, which you'll see here today, how we do that, how we think about it, and also how we are thinking about doing that going forward. Just because right now it does take literally the majority, I would say, of both mine and Jason's time to run all these agents and use them successfully, right? We could run more agents and they would fail if we didn't put in as much time. But since we do devote a lot of time to them, they've become quite time consuming. Now, none of that is to scare you. I don't think, you know, 20 is the right amount for everybody. So if you haven't seen, this is sasser.ai/agents. You can see all the agents we use. There's a mix of ones we've, I've coded that Jason's talked about previously. And then there's ones that we are using third-party tools for. So that's all listed there. You'll see most of those here today too. We'll go through the metrics of them. Just like in full disclosure, I've got screenshots of all of our metrics. And then also, we're going to go through how we think about them in our day to day. So Jason, anything you want to add there?

No, I think that's great. I think, and I think we'll go through all of it. I think most importantly, we'll give you an update on how all our AI SDRs/BDRs have worked, outbound, inbound, the actual data, which I think will be super helpful. Emily put this together, which is great.

Yep. And then, so why 20, right? So I'll kick this off to say 20 is not the right amount for most people. I think even like a few is maybe the right amount for most people, or even right now, if you're like, I have maybe let's say one agent kind of fully deployed. I, it's scary to me to go from 1 to 20. I don't think you should take this as like the gold standard of everyone needs to be on 20 apps or you're behind. Now, I do think you need to be. We've talked about this before. I do think you need to be on some version of this stack in GTM to not fall behind, and I'll show you why. But I don't think 20 is the right number for everyone. I feel like between myself and Jason, we're fairly technical, right? Some of the stuff we've live coded, we've been thinking about for a long time. Some of the stuff you'll see that we've deployed across our go-to-market org is where we had gaps and we had a need to fill them, right? So. That's why we have 20. I just put a fun image of, you know, this is a real image of myself that then Reeves added a digital clone to. But I think there is a lot of talk and I want to get your thoughts quick, Jason M. Lemkin, on this before we move on. You know, how much can AI clone yourself? How much can it do? I would say now, 6 months in, it is for us and for me, it's become a clone of all the best things we can do when we devote the time to it, right? Like it can do such a better, like a, such a massively scale of output more than I could physically do or that Jason could physically do. There's just so much scale in getting these systems to actually work and go to market. That is a scale that no, no single human could achieve, right? And so that's kind of the magic of it. I think 6 months in, it's still does not cease to amaze me how much we can do with our AI together across all of go-to-market. And I would say too, in the last 6 months, we didn't start with all of go-to-market, right? We started with basically a support use case and then we added in the— we layered in the other places where people either left the org or we had gaps or we wanted to improve how we were doing these things in go-to-market. And so that's kind of how we landed on Twaini. That's why we have it. All across the org. But I think this concept now, it's cloning the best parts of your best players, right? Maybe your org isn't as small internally as Samstra is, and you've got more folks. But I think there is now, I think, a clear path to where I'm seeing in these platforms. You can clone all your best A players, right? Take all the best A players on your team across marketing, sales, CS, sales, rev ops, and make them S tier with AI. I truly believe that's possible and you'll see it from our results. But yeah, what do you— any thoughts out there, Jason?

I think that's a really great insight. Thinking about what we have learned since we've had some success, a lot of folks want agents, specifically AI SDRs, AI BDRs, AI marketing agents, others. They want them to do magical work for them. They want to spend $20 grand, $50 grand, $100 grand, and all of a sudden get leads. That is I didn't fully realize that till just now. Till I said, that's the wrong way to think about it. What your agents can do, and it is so powerful, is they can take your best practices and scale them out almost infinitely. Figure out what works, figure out what campaigns work, what messaging, whatever works is already working. You have to have something that's working. An agent can't today figure out, make something that isn't working work. But if you know what's working, and then this is important, you train the agent with it, then you get 24/7 infinite firepower backing up your best practices and your best practices will change and you'll run A/B tests and ABCDEF tests and multivariate tests, but you've got to understand what is working in GDM before your agent can scale it up. And I think that is a fundamental mistake that confuses folks. It's not about buying a tool, but once you have something that works, which we have, and Amelia will show you the data, then you can scale in a way you can't with humans.

Yep. Yep. I would say that's my biggest learning 6 months in. Like we, we took things that we're already working, or some processes were broken, right? Like our RevOps process was pretty broken before we did this, but we figured out how to fix it. And then we basically put that on acid with the AI and the agents. So I do think it's a, it's an interesting nuance of how to approach it as, you know, people are in different states. Okay. This is just a quick fun one before I go into our data. This is from Jason's co-host Rory Hitzgamp. This is their state of GTM AI. It's pretty good. It's actually fairly short but concise. So I like it. You can just go to their site, you can download it. It said a lot of interesting things that I used to kind of anchor how I wanted to present a lot of our data since we have a lot, right? So I think we're in this right now, sort of paradoxical world in November of 2025, where a lot of folks have adopted at least one sort of AI or agent. But maybe to Jason's point that he just made, a lot of them aren't seeing the impact because either they're expecting too much or they're not putting in the time. So I think that's kind of where I see the paradox lying. In this report, you'll see— I'll write it up too on just some of my learnings. But, you know, you'll see across go-to-market, which ties into this most, most, most deeply, the adoption has been in marketing. But if you read the whole report, it's on kind of what I consider base-level adoption, right? Like they're using cloud to create content. Yes, so are we, but you'll see our more kind of like sophisticated 2.0 stack at the end of this. Sales dev, same thing. They're using it for sales messaging. Like the kind of middle ground is using it for starting to do things like RevOps or more intelligent emails. Again, consider that kind of baseline. I feel like you can, again, if you take the best people, give them an agent, they get used to and they know how to train it, I do think you can get them to blow by that fairly quickly. So I think across the board, even what they said in sales and CS is a lot of folks are still early, right? So a lot of still early adopters there across the board, but some good stuff there. I'll try and tie into what we're going to show. I think just a quick one to tie into that data, plus what Jason M. Lemkin was saying as well. Here are my quick biggest learnings and then literally all the next slides are the data that we'll run through. I think again, in, in a lot of early days, and as you add more platform, I think the expectation that it gets easier with each platform you add is false. So that I wouldn't expect to have that expectation. I do think a lot of the platforms, whether you use what, like one specialized one for each function the way we are, or you end up on one platform that does kind of like a little bit of everything pretty good. Maybe in that case it gets a little bit easier as time goes on, but it's never set it and forget it. Like the biggest thing is that our agents will ebb and flow oddly depending on how much time the humans put in, right? So the more time I put in with our agents, wow, shocker, the better the output is. And sometimes that ebbs and flows, right? Like we get busy, we have SaaStr London coming up, so I can't always spend the amount of time I would like to with the agents, right? And so that obviously does translate to the results. And so I think that's one of the biggest things. I think Jason's point on like immediate ROI and headcount is a hugely false expectation that I hope nobody has on this call. But if you had it maybe going into this, can kind of course correct on where you're going to see lift. And there, there have been cases in our data, you'll see where we did get immediate lift. But again, it took our processes It took us recreating some of our best processes that were already working to get an immediate lift in these 6 months. You know, and then I think in the expectation on needing to staff like an AI expert, I've talked to a bunch of CMOs and CROs now, as well as founders that are either at SaaS or participating in SaaS, or where, you know, they say, hey, Amelia, you know, what's your advice on my biggest problem right now is we're starting to onboard all these tools, but I don't have anybody to run it. I don't have a chief AI officer like I am now to help us figure all this out. What do I do? And most folks, I will say that, that have figured this out and started to deploy AI more across the GTM org have told me that their best success has been, again, taking your A-player, take somebody that is the best person on your sales ops team, your rev ops team, your best marketer, your best SDR, your best AE, sit down with them, figure out what tools you want to use together. Don't just deploy it to them from, you know, the castle. And then go through that process with them together. Because it's very important too that you need to do the AI yourself in order to figure out how it works. You know, you can't expect the, even your best SDR to know how to instantly use this. You got to help them out and help them figure it out a little bit. But everyone I talk to that says they go this route of using the best A-tier players, and injecting AI and figuring it out with them have now— they have now changed their job roles. Like in the way that my job role has changed into now Chief AI Officer, they're saying the same thing. This is like the best path I've seen work for people across. And these are, you know, founders and CMOs at different stages, different ops. Some of them are even AI-native companies and they need to figure out AI too. They're, they're just taking all their best people, figuring out— and these are the founders figuring out AI with them in the trenches. And then you become this S-tier team together with AI that can do this, again, great outputs that you'll see like ours with everything AI. I think this note, like somebody else asked me yesterday, hey, I, you know, I'm rolling out a few platforms. Some of the, some of them are ones that SaaS uses. Do you like recommend anybody to any, either anybody or like an agency that can help me deploy all this? I'm like, that's the wrong way to go about it. One, you gotta figure it out yourself. To, you know, no, I don't care who, where anybody went in life or school or whatever. No single person on planet Earth will know how to do this better. There's no agency that already knows how to do this. There's no, maybe there's a few like S-tier people at orgs that you can try and steal. But outside of that, I wouldn't, you know, buy the snake oil of, oh, we're an AI-native agency and we know how to do GTM across the entire org. I'm like, they probably haven't done it themselves. I know maybe that's a little disheartening that you have to figure out yourself, but that is the best path forward in the meantime until others catch up. And then I would say too, don't get too hung up on this data governance thing. A lot of folks I talk to now, especially at bigger orgs, get very concerned on, okay, if I add so many agents, I'm going to need governance as a layer, or I'm going to need to clean up all my data. And cleaning up all my data or getting all the processes right is going to take 2 years before I can even think about implementing. I think too much data is too much data. And if you have things that are working, again, I would focus there first. So I wouldn't try and tackle, you know, bite off more than you can chew. It just, it didn't work pre-AI and it's not gonna work with AI. And then I think the last one, because the other ones we'll get into per department, is I hear this panic a lot from folks now. I don't know if it's because of how much content we post about AI or because of all the tweets and all the LinkedIn messages that they have to keep adding. Like, oh, I gotta catch up. I gotta— Saastr has 20. I gotta get to 20. You don't, like, actually we have 20. I'm only adding one more tool for the rest of the year. That's literally it. I'm adding more use cases to our current agents. I'm adding a 2.0 agent-to-agent force. I'm adding a 2.0 thing to qualify, like, I'm going deeper on our current tools. I'm adding one more 'cause I think it's cool. And that's probably it for the foreseeable, like Q1. I don't really know what I'm gonna add yet actually. I feel like if our agents can just go a little deeper, I may not add anything new. I might add, you know, just deeper use cases on our current tools. Okay, let's start with outbound. I'll try not to spend all the time here. We, I know outbound is near and dear to a lot of our hearts and we talk about it a lot, but I'll start here and Jason, I'll get your input too. So for outbound, 6 months in now, we have sent now almost 20,000 messages. Actually, it would be 20,000 again, if the human me had a little bit more time to spend with my agent, it'd be over 20,000, but still 20,000 is like a lot. Overall, our outbound with the AI is almost a 7% overall response rate, which is, you know, kind of double just overall averages across anything you'll see from all any, you know, tool that does sequences and outreach. A 4% positive response rate, which is higher than most. Now keep that in mind. This is like higher than most folks in the platform. But interestingly too, now that it's been 6 months, 10% of ticket revenue for SaaS or AI in London only on this agent. So this is an interesting thing where I have seen our outbound agent. It has different goals. We've explained this previously, but I'll just give you guys some context. It has different goals and so it has like different agents in one platform. Because the training is different on all those agents. I have an agent basically for lapsed sponsors. I've got one for current sponsors. I've got one for people who previously attended Soundster. Now I, now I have one for people who are opening our emails, but maybe not taking any actions with us. And then I have one purely for cold outbound. So all of those agents are trained and tuned in a slightly different way because those audiences are different, right? The messaging is different. The collateral of what we want the agent to do. And also the goal for the, what we want the agent to do is very different across those 5. And so we've got about 5 core ones just within Artism that are set up differently. And this is, I think, a surprising learning because at the start I was like, okay, you know, it's an outbound tool. They were a sponsor of SaaStr this last May. Like, I'm going to use it for sponsorships. But then I quickly saw, okay, like, it's not bad at booking meetings. Like, it's fine at booking meetings and that's great. And then, you know, capture that intent. And it's a great way to do follow-ups again to like current or past or lapsed customers. It's a great way to do that using AI again at scale in a more consistent way that sometimes we can't physically do. But then, you know, there, there was just this interesting learning of because tickets for London and then now for Annual are, you know, less than $1,000 depending on which event you go to, the AI got pretty good at selling the tickets itself. Like, it got pretty good. Like, at first I was nervous. Now, 6 months in, I'm like, it's— look, let it go. Let it run free. Like, it's gotten pretty good now with all the time we've put in that it is empowered to sell tickets to people. Again, it's a lower ASP, so maybe that's the most interesting learning. And it is still booking meetings for us on like the higher ASP things. But, you know, these people know us and it's just, it's giving us this scale and personalization in a way we couldn't pre, you know, pre 6 months ago. Like every event after we have this after event, I struggle to, you know, send, I'm like, okay, we should send out hyper-personalized follow-ups to everyone, to all the sponsors, to all the attendees, to all the speakers, everybody. Think about London, that's 2,000 people annual. That's 10,000 people. Like, we could never do it right. Like we had to literally cherry-pick the people who got like an actual human customized email, then who would the sales team send like a somewhat customized email to, and then who just got like a templated email because we didn't have time to do the rest, right? We just can't send that hyper-personalization at scale before this. Now, you know, we've sent 20,000 post-London. It'll easily send a highly customized email to all 2,500 people that come. That's crazy if you think, just stop and think about that for a second. It's doing all this for us. And this also was something that took not just me, it took, you know, different members of the sales team. Like everything that we did is now consolidated into this in a way where the output is obviously fairly, for us, positive at least, right? We've seen some good gains here. But it's not without time, right? So our— how we think about using Arzon specifically for outbound, and you'll see here, that's why I put the data, like the positive response rates do vary, right? And so this is varied on agent. This is showing just the— so one's a website one. You'll see a couple of these are attendee ones, but different years. And so they have slightly different trainings, slightly different things that they're referencing here. But these are ones where, you know, these positive response rates do vary depending on how maybe recent they interacted with us. Again, that was probably true pre-AI and it's still true now, but because our outbound emails sent per rep, and this was a crazy stat I just looked into last night, our outbound emails per rep across the board. So across the things I just mentioned for us, you know, tickets, sponsorships, speakers, whatever, on the average bad 6 months ago was anywhere between 75, you know, let's say at the low end to 285 at the high end, depending on the person. Some people are, were faster than others. Now our AI is blowing that scale out of the water. It does 3,000 on its own per month in one platform. This is just one platform. I have another one that's always— that's, I have actually two more, two more that's doing outbound. This is just one platform that's giving us this leverage. And so I think again, what's working for us is using this to do hyper-specialized personalized, you know, messaging at scale has been working. There is a, which I fully believe it. I've talked to Jaspreet, the CEO of Artisan. There is a 2 to 3 week warmup period with Artisan so that they can get your deliverability as close to perfect as possible. You want to do this. You don't want to skip this step. There's a reason they make everybody do it. At first I was kind of annoyed. I'll be honest, like previous webinars, they'd be like, it's kind of annoying. It takes 2 weeks or anytime I need to add new domains, it takes another 2 weeks. I get a little frustrated, but then you wait and then you see why, because you're like, okay, you're not only do your emails in this hit the inbox, they don't go to promotions. This is a great— I can't even solve that in Marketo. So many of our newsletters end up in promotions. These actually go to the inbox. And so Again, just another point of like differentiation there of things that could do that we couldn't do 6 months ago. And at the start we had to review all— I was nervous. I wanted to review all the emails myself, but now I just spot check, right? I spot check it, but I do mainly let the AI in order to draft a response. I don't let it send it fully yet, even 6 months in. I think it's still, and some of my other agents I do empower to do full responses without me looking, but this one in particular, because of where SaaStr is, and maybe if you have multiple products, you might be the same way. Sometimes people ask multi-threaded questions, right? And so I think that's where the AI is pretty good. If, you know, let's just say in this scenario, it's to book a meeting and that person's like, yeah, I want to book a meeting. I'm like, cool, here's the calendar. The AI can do that. But if someone starts to ask about speaking and sponsorship, that's a common question we get. It's multi-threaded. It's not as good. And I think because there's just some human nuance and okay, do I feel like they're trying to ask me about speaking or free tickets? Because they want us, they actually want to sponsor or do they just want free tickets and they're kind of wasting my AI's time. And so there's still a little bit of that, you know, we do constantly monitor the human attention and responses specifically in this. Platform that we're using. So again, I do think it— I think the unexpected learning is for Artisan in particular, and probably any other outbound tool you use for AI for outbound, it relies heavily on human consistency for training and iteration and quality of contacts. Like, it's not 100% of— not 100, okay, 90% of the contacts we've reach out to with our Artisan agents are contacts we have. They're not contacts I'm going and finding in Lucio or Apollo or Clay or Artisan has a way you can do it natively in the platform. I just, you know, I trust our contacts more. And so putting in all that data does require time and it requires consistency on our end. So when I've had more time, you know, the sequences and everything works better. When I've had less time, it still works on autopilot, but I see a drop-off in the amount of responses we get. And that's because I just can't keep up with the AI. 6 months in, I have to try and give it fresh context. It used to be I would do it once a week. Now I do it twice a week if I can, ideally, because it's gotten better. Jiu, right? Like, I've empowered our agents. It's gotten better. And so ideally, twice a week would be better. So that's, I think, the, again, unexpected learning there. Anything you want to add on outbound, Jason?

No, I think I, I, I think that's a lot here is it's interesting. I think there's a, this is plenty. I think the question, I mean, the conclusion obviously is that moving from the 2 human SDRs we had, the way we were doing it to AI, there was no, no downside in net, right?

Yep.

Actually not necessarily better. It looks like the open rates and response rates were roughly similar in the end, but we have 10 times the scale if I had to summarize it, right?

Correct. Yep.

I guess the one question that I might have from the audience is that's great, but what about like super high value prospects? The ones you would want to reach out to personally? How do you, how did we address that? How do you measure that? How do you think about that versus this is a broad swath of a lot of potential prospects?

Yeah, it's a great question. Yeah, for super high-value folks, I still put our AIs into draft mode across our agents, right? So there's different use cases where I want to hit different folks, different parties with a message that is maybe hyper-personalized, or it's somebody who's actually met me and would know that You know, my amelia@saastr.com is one of my Artisan emails and not amelia@saastrinc.com, literally. And so for those folks, like no matter what platform I use, Artisan does some of them, some other platforms do others. I pretty much put the agent into draft mode and I see what it says, right? But that still saves me time, like having it draft that next email, either next email or first email of you know, whatever I'm trying to action with them as like the end goal is helpful versus me writing everything from scratch. Definitely helpful. I think I still spend probably the most time making sure the contacts are the right people. I think where there's a gap right now is for a lot of these folks, especially our high-value ones, I always have to find new people to CC, which is manual, like that no AI can do this yet. Trust me, I've looked like, no way I can do this yet. We're like, because we have so many CEOs and founders in our database and maybe you sell to a lot of C-suite executives too. The biggest issue I have right now in the AI is not that it's not working or that we have high-value contacts that it can't draft an email to pretty good and fairly customized. It's that you don't want to send it just to CEO, right? So if I'm going to email the CEO of Databricks, like I want to copy his press team, his chief of staff, his CMO, if it's for something speaking, probably the CMO too, for visibility. Like, I don't want to just send it to Ollie. Like, even though I have his email and he knows Aster, and there's no way to really do that in the— like, there's no good way right now in any AI to find those additional contacts and also include them in a sequence. Like, right now, across all— across the board, all of our platforms are one-to-one, right? So I think that's something where we're 6 months in. I do think that will get better and the platforms will let you do that more so down the road. But that part takes a lot of time right now. That's kind of like one of my biggest pain points. Okay, let's flip to inbound because I want to try and go through all of go-to-market, but I'm already 30 minutes. I'll pick it up. All right. This is the inbound. So there's some interesting numbers here on the right. My Zoom's a little bit in the way. We actually have not had this for 6 months. We've only had it since August. So August, September, October, a couple weeks in November, 3 and a half months. Let's say we have not had this one as long because I added outbound first and then I tackled inbound. Again, you got to stair-step your use cases here on AI. I wouldn't try and do everything in GTM all at once. Do one, get it working, then add another one if you want. So then I added inbound. We've had, you could see it's had a lot of, it's had almost 700,000 sessions with people, which we could never do physically. And it's had 1,000 conversations. So like sessions versus conversation in this realm is that the sessions are either, I think he counts it where it can answer itself, whereas a conversation is something where the AI agent is actually conversing with the person more fully. It's looking, you know, it's looking at its training, it's answering it a bit more. I think that's the new ones, but don't take my word for it. But it's still a lot of sessions and conversations. And then when I took the screenshot, it was 91 meeting books. And then this morning it was close to 100. So we've had a lot of interactions with inbound. Now that's always going to be true, right? This is inbound. But this is the data I wanted to share because I will say I wasn't as hesitant as adding an outbound agent. There's alternatives to Artisan if you don't like that one. I like that one because again, it's highly specialized on outbound. So I like that one because everything it does is to get the like similar results to what we're seeing. Like that's literally its whole manifesto. So I like to go specialized. I know other people like to try and find one platform that can do like multiple agents and try and consolidate for a lot of reasons. But I like to go specialized because I do feel like the specialized ones for now still work a lot better because they go deeper. But anyway, so then when I added inbound, we've only had this, let's say, 3 months. The inbound agent has been responsible for already a million dollars of revenue. Now you could say, okay, some of that revenue would've closed anyway, right? It's inbound. People are reaching out to SaaS, or in this case, particular for sponsorships. You could say, okay, you know, one that would in the old days, 6 months ago, that would've been routed to a rep. Somebody would've responded eventually, gotten the meeting on the books, and then, you know, hopefully, I don't know. I mean, my data does show our win rates are better now with AI on inbound than 6 months ago, but assuming it was similar-ish and still fairly high, let's say, you know, 750 of that would've still closed. The onus here is that it closes— not only does it close a lot faster is what we're saying in our data, it's also closing at a higher rate than previous to us 6 months ago. So even though it is inbound and you could say, okay, there's obviously inherent demand because they're coming inbound, this has given context to both the prospect and our sales team in a way that it's giving us speed. So our inbound agent right now is definitely picking up speed. A crazy stat I just saw too, like 70% of our October close won came from the AI. So you can see, right, there was a little bit of lag in August, September, and then most of that million came last month. But then we have like another $2.5 million in pipe that is literally attributed to meetings and deals that came from the qualified agent. And that's magical in a couple ways. Now, I know some people like to describe it as, okay, it's just a meetings booker. I could do the same thing with just a meetings booker. That's like round robinning. And if I show that on a website with a little AI, maybe that's good enough. And maybe that is good enough for you, but I'll say The difference here of why I've become a fan of Qualified for what we use on this tool is that it gives you so much context more than a meetings booker, right? Like even if you have an AI meetings booker that's doing round robin for inbound, our agent is having conversations with people. Like it's okay, why, you know, let me feel like, let me book the meeting, but first, okay, it books the meeting instantly. It used to be 6 months ago. You had to fill out a contact form to talk to Saster. Then I would— it would go to me. I would round robin it. So there's a delay, like if it's overnight or on the East Coast or a time I'm not awake, there's a delay for me to round robin it. Then I send the round robin to the rep. Then the rep has to respond to that person. And so that process is anywhere from within the same hour to a day later. That's highly inconsistent 6 months ago. Now our agent is empowered to book meetings on its own and have a conversation. So it's send— it's not only booking the meeting instantly where this used to take up to a day at the worst case scenario, the meeting is instant, but it's having a conversation with this person so that when I give it to the sales team, like literally me and David will go through, we'll be like, okay, what did they say to the agent? Did they say anything interesting? The agent can also show you other people who have been on the website. You could say, oh, hey, cool. Like, this person actually booked the meeting with me, but I saw their CEO was on the site, or I saw a lot of contacts from the same company were on the site. Let me see. And also what they were looking at, it'll show you directly. Okay, this is what they were reading on the website. This is what they were probably interested in. And so that context just gives us a lot more like highly pre-qualified contextualized meetings, right? We don't really have to do discovery anymore because the AI has already done it. The AI already knows what websites you are on. It already knows who was on the site, who was there, how many times you visited the site. And because we have ours on multiple domains, it can also say, hey, they were also looking at XYZ thing on saster.com. Like, again, I, we don't do discovery anymore. We're like, okay, our agent already did it and already has the data. We just use that to do like a theory of what we think. We're like, hey, we saw you were looking at XYZ thing. Is that actually what you want to talk about today? You know, part of sales is still learning what their goals are, just, but we have, it's so nice now to get on a call where we have that baseline with folks. They appreciate it, right? That you're not doing, you're not wasting 10 of their minutes on discovery. You already kind of know what they're about. You know what they've been doing, or sometimes they're like, oh, I didn't know our CEO was on the site looking at speaking. That's good to know. Maybe I should look at a sponsorship package that includes speaking. The conversations of quality is anecdotally a lot better. I don't have data around that, but maybe we will now, but it's just nice that it can do that. So that's an interesting context on selling something that is more of a higher ASP, right? Because for us, again, tickets are like sub-$1,000 sponsorships. The most popular one is, let's say between the $50K to $100K range, there are more custom ones than that, but this is helping us with something that's at a higher price point, close one a lot faster and we're seeing a lot more of. I will say two, two interesting things before I move on to the next one. And then anything you want to hit, Jason, on this one, but I see a lot of folks who use Qualified that limit their agent, right? Like they limit it to talk to support or talk to sales and there's two buttons and that's kind of it. And then when I talk to other people's Qualifies, I've seen it sometimes not as good. I'll be honest, it's not as good. And that's not a fault on the technology. That's a fault on the training. Again, if you take something that is A-tier and make it even better, like we have, like I've empowered our qualified agent to ingest all of our websites. So across Saastr.com, which is, you know, 20 million words, Saastr AI London, Saastr Annual, our YouTube channel, things I upload to it. I'm like, these meetings, I upload these meetings. I upload some of my sponsor meetings. I'll upload calls to it, so we kind of know like what we're saying on calls. Because it can ingest all that, it does so much more than I think most agents do on Qualified. And so ours in particular too is I think pretty good, but it's also pretty good because we've empowered it to be that way. There's a level of trust I have with my Amelia AI now that we've also, you know, I talked a little bit about the outbound piece of that's booking meetings for sponsorships, but it's also selling a lot of tickets. The inbound agent previously, right, you would fill out a form on the website for sponsorships. And then if you wanted to buy a ticket, there was kind of nothing. There was like, you could email events@saastrink.com 6 months ago. It wasn't great. And it's, you know, the number one question our chat gets now related to tickets versus sponsorship is, can I have a discount? And so quickly, oh, I think a weekend when I saw that in the data with Qualified, We empowered the agent to sell tickets directly. So what it does now is if you ask it that question or if you prompt it for that question that you're looking, you're like seeking a discount, the agent's goal is to not only give you that discount, it will remind you because the agent lives on our website. It will know if you come back and if you don't after a couple of days, it will actually email you like, hey Jason, I gave you the code. We had this nice conversation as Amelia AI. You didn't use it. And then it reset you the code. Hey, are you still interested? And people then use that code. Like, again, I can't do that at scale, just me. And it's something we didn't have prior to this. So between our two agents, it's now 20% of our ticket revenue for the upcoming London event. We'll see where we net out for London and then where we net out for Annual, because it will have a lot more time. But I think our AI just totally crushes it on this kind of stuff. If you can empower it, if you can get it to work in the same way, man, does it crush it. So I love our Amelia AI. I actually went to go film with Qualified, and I think it's called Tavis, to make our AI into a full-blown seller and supporter where it's gonna have my video and voice. And so that will hopefully roll out by London and y'all can play with it there in person too. But I'm just letting it do more.

It's good. Yeah, I want to keep moving because this data is so good. We've said it before, but when we look, you know, we're, we're 100 days in here, we're 125 days into using Replit, we're 6 or 7, 7 or 8 months into using Delphi. If you're hands-off with your agent, not only will it not perform, not only will you not train it properly or iterate, you won't realize all the things it can do. It can do so— all of our agents, when you get good at it, if you just are really hands-on with your agent, you work with it every day, you watch results, you're as quantitative and data-driven as Amelia is here, if you get there, there is a pot of gold after that because you will find your agent can do so much more than you thought it could. That is the magic. But you got to push on. You got to get really good at what you think it does and then open your eyes for other use cases. All these AIs get so powerful, they know all your data, they ingest everything out of you, And the underlying LLMs are so powerful, they can't help but do more.

Yeah, I, it's a, you'll see, like, I see some, I'm reading some comments in the chat too while you're going. The, you know, our, again, my slash our theory is I like to go deep now with like specialized tools where I can just keep adding use cases and add agents within certain tools. I don't use an all-in-one tool, but I feel like the results are worth it. And so that's why it's worth, you know, sometimes I have to put stuff in Artisan and Qualified separately. And is it a little annoying? Of course it's a little bit annoying. It's time-consuming, but am I going to do it because the response rates I think will be better in the long run for inbound versus outbound? Yeah. So I'm still going to do that for the foreseeable future. It's going to be a little bit annoying. And painful. That's true of anything. Like outbound's always a little bit annoying and painful. Like, it's not like the AI was gonna magically wave its wand and like Glinda-fy it. Like it's less painful and I have more scale, but I'm just still outbound. If you didn't like doing it before AI, you're probably not gonna like doing it now. Oh shit. Okay, a quick one here, 'cause I do think, I don't know if the hologram, I have like more use cases, but maybe I'll have to do a part 2. AI and GTM, the— there's this middle ground, and I do want to spend some time on this because I feel like most people have a misconception about the next product, which is Asian Force. And I do feel like when I talk to people about it, they're always shocked. They're like, what do you mean? I've— either they've never seen it, so I put like more screenshots, or they're like not sure how to use it. Because I don't know, I think because it's from a, you know, an existing company versus a, it's an incumbent versus a startup. I don't know. But anyways, I get some mixed reaction, but I'm like, yeah, part of our core stack now is AsianForce. So I'll explain my AsianForce. I, I've gotten to know their team very well. They're all fantastic humans. Let me say they are the nicest people on the planet. They're like some of the best people. And I now love also You know, I love the qualified team. I've gotten to know Jasper and Artisan. Part of this is I do trust the humans that are on the backside. I think that's, if you're gonna pick an AI vendor, I do think that's part of it. Like you should trust the humans, like you should wanna work with their human counterparts. Like I call Replit's team Replit's humans, like Replit's humans. Like we went there this week and I was like, oh, we're gonna go see Replit's humans. I'm like, but I like those guys and I trust them. So same thing with any vendor you pick, but we have this middle ground where I've gotten to know their team a lot and they, you know, I think publicly they say their number one, their number one use case for Agentforce is support. And I'm building a 2.0 agent that lives somewhere between support and like personalization. But our first agent that we've deployed, and this was recently, so this was just right at Dreamforce, which was like a month ago. So it's only been live a month. This is our newest agent, right? I had ours and I added Qualified and then added a bunch of other stuff you'll see in marketing. Then I added Agentforce. We had a gap where it wasn't inbound in the way that like Qualified is. It wasn't outbound in the way that Artisan mostly is. I have this gap where embarrassingly post-SaasRay annual, I was like, we literally just from this SaasRay annual have about 1,000 people, our sales team, we never followed up with. These are people who literally filled out that form I was talking about earlier, said they wanted information about SaaStr Annual. I routed it to a rep and then the rep did nothing. I found 1,000 of these people, which is all the blog. Like, I can't kill— I'm not gonna go— I should email those people an apology one by one and email them. But that, again, that would take me too much time. I need the scale. And so I thought, okay, this is probably the best first use case. Born AgentForce agent because these people raised their hand, meaning they're already in our Salesforce because that's how our forums used to work. And I already have info about them in Salesforce, whereas like Artisan does a lot of personalization based on what they're posting. This, the magic I would say of AgentForce is that it knows everything your Salesforce knows, which sometimes I say that to people and they're surprised. I'm like, what? That's the beauty of it. Like, it has all your Salesforce data if you like click a button and you want it to. And then you can use that to then sense the context of the email. So I'll show you what I mean next. But this is our first use case. We had 1,000 leads nobody wanted to follow up with. So it's now sending outreach to mainly those people. Like, it's still working. Again, it's newer. So It's still working that list of 1,000. I also capped it, right? Whereas like Artisan's doing 300,000 a month, it's doing 3,000 a month on month 6. Like, agent force is very much on, I would say, month 1, at the end of month 1, going into month 2. So I've capped it so it's still emailing those 1,000 people that we ghosted. But then I also started adding people where if I have a conversation with them for a deal cycle, it starts to email them a follow-up because sometimes I'm really bad at that. Between those two use cases, what we've seen, and this is early days, is a higher response rate so far and a huge open rate. Like, again, the open rate is 72%. It's freaking high. You can't get that in Marketo. You can't get that with a human that doesn't send that email. The open rate is zero. It equals zero for an email that was never sent. And even though it's early days, I'm fairly bullish on this one because of the early data. Is pretty good. So I have liked it for that. I think it's a good use case that more folks are getting into. Okay, first, you know, again, support is their number one kind of like agents, number one, most people deploy, find whatever, but are qualified in support. So like I qualified one that did sales and you'll see I put two sample emails here. This was like a follow-up email to again, somebody we ghosted. And for this agent, because it's early days, I haven't empowered it yet to do bookings itself. So that's why it's asking for a meeting versus like just showing a direct booker. And Kyle responded literally right away and said, you know, no thanks, but I'll attend. Again, never would have happened 6 months ago because our sales team, nothing against them, they just decided not to follow up with these 1,000 people. And Yeah, you could say maybe I started with the use case that is low-hanging fruit. Sure. But why also, why wouldn't you? I already have so many use cases on the other agents that we use, but I wanted one that was near and dear to me because I was like, if I inbounded to somebody and they never followed up, I salty. I'm like, maybe because the AI has data of, you know, them being in our sales force and what their company has done with us. Maybe they could send like a nice note. Like, you know, so I don't know. That's where I went into it. The other thing I'll say about agent-first, which was the learning, is there's two slides here. So if you're flipping ahead on the deck, because somebody asked this earlier actually about the setup of it, like it is within the UX of Salesforce, right? My biggest learning was like, I am, I will admit, I am not the greatest Salesforce user on planet Earth. I was not the greatest user 6 months ago. I've definitely become a better user now in this process. But I'm by no means like, you know, I'm not even certifi— not even certified. Like, I know that a lot of people— I'm not Marketo certified either. So I, but I would say like in my rankings of tools in our stack 6 months ago, I would say I was a lot better at Marketo than Salesforce. And I pushed a lot of things from Marketo to Salesforce just because that's what I was trained on in the early days as a marketer versus Salesforce, which was a sales tool only back in the day, they just wouldn't give you access. And so I didn't know all the like nooks and crannies. I learned a lot of them in this process. And so I think that's, I think there's a misconception on it's super technical and nobody can get it to work. That's like the number one thing I hear, but honestly, one, I had help with all these platforms. Artism helped us help. Set up Artisan. Qualified helped us set up Qualified. Salesforce helped us set up Agentforth. Like all of these, they help. It wasn't just me. Like we did a lot of the training and tuning once we knew how it worked, but they helped us with the setup. That's not vendor agnostic right now. And I think it's true of any good vendor in AI right now. They do have to help you and they should want to help you get successful. They want you to have the same numbers and results we have. I think this misconception of, you know, oh, I can't use Agent 4.0 in Salesforce because it's probably too technical or it's going to be doomed to fail, or I probably can't get, you know, I'll need either a third-party agency or like a Salesforce admin to come do this for me. I'm not a Salesforce admin. I figured this out with their help. So again, I think it's true of any tool where you'll need their help at the start for sure. Like now I need less help. Now I get how this works, but you'll see too, like a lot of things, even though they don't look the same. Are kind of the same concepts. So once you learn one tool, because I learned Artisan first, like this is for reference, the Artisan setup where this is only two screenshots of a very long page and it's great, but you're seeing like I'm putting in, what is this one for? This one's for London tickets. I can tell because of what I put in my proof points. And then there's coaching right here that I put in. Then there's other things you have to put into the setup. This is not dissimilar to that where I'm putting in instructions on the prompt builder similar to, or actually copied my cold agent force. I actually copied all the instructions I put into Artisan. I just put them into agent force and rewrote them for this use case of email people who haven't been reached out to. And then it worked. So like some of the training you do is again, maybe not going to look physically the same, But some of the same concepts I think are core across any AI tool you use, where again, between the ones I've shared so far, Artisan, Qualified, now Salesforce/Agentforce, like this logic that I had to learn 6 months ago, okay, this is how you tune an agent. This is how you train it. This is how you consistently update it. Does have this kind of universality to it, which I know seems a little funky, but once you kind of learn how to do it in one, it is easier to do the others. I don't think I could have gotten successful on having all of these agents and now learning how to vibe code in Replit had I not taken the time to learn the foundations of one of them and go really deep so that I could basically speed through the other ones and say, okay, I kind of know a version of this. I kind of know the concepts. You just start to learn how to talk to AI and talk to the agents and the models with your data in a way that, again, no one can teach you. Like it's just something you've got to go through yourself of, okay, I've learned now how to talk to Reppli. I've learned how to talk to our agent, birth agent. I know how to talk to qualified. I know how to talk to Reppli. Like again, certain universalities are there actually 6 months in where if you kind of, if you learn one, you can figure out the others because they're not dissimilar to one another. Okay. I had, there's a lot of questions. There's 8 minutes, but I haven't really gotten to all these. What do you recommend, Jason? Should I keep going or should I just do questions and do a part 2? You're on mute.

Sorry, let's do questions. I think the SDR outbound inbound is so important. Let's take questions and then let's break this up into a part 2 on the rest of GTM and AI agency.

Okay. Okay. I can do part 2 next week, guys. It will not be long. I can do part 2, but I haven't even touched on RevOps, CS, and like marketing. So. There's a lot there. Let me go back to question. Oh, some people are like, keep going. Like, do you want to be here for this? Well, let me go back to question. Go back to a few key questions and then I'll do a quick overview and we'll do a part 2. So I'll meet you guys in the middle if that's okay. Go back to the start. Let me go back to the start. Someone has asked, what are the tools you are using with success? Yeah, they're in this deck. And then I know, sorry, a lot of you asked for a link to the deck. So I'm going to put that in now. It wouldn't let me do it while I was talking. I think I have to refresh. Share. Hold on. I sent it to Jason M. Lemkin, probably. Hold on, I do think this is important, because if we're gonna do part 2, you guys can also tell me what you wanna see in the next one. Okay, here it is. Let me get you guys the link. Oops, Megan. It's not everything you can be told by AI. Some things you have to do yourself. Like finding this link in a very roundabout way, but I got it, don't worry. And also you guys can keep asking questions. I will— I'll stay on. I can go over. So if you guys want to stay on and get your question answered, or if you want to watch the replay later. That is good. 5B. So copy link address. Okay. Slides. There we go. Okay. There's one. First major question I thought. Somebody else was asking what's our tools and stack because they joined late. So at the beginning of this presentation, but if you also go to faster.ai/agents, we talk a lot about these tools. So this has all the tools. You can see I really didn't make it far. I only made it through 3. The ones I was going to go through today, but we have a few here. So you can go to saster.ai/agents. There's a little bit more on all of them here. And then I'd link this deck where you can— we'll throw it on saster.com too, so that you can also see it there. There's a few good questions here, I think, on the earlier part of the session that was about inbound and outbound. Some folks were asking what is our cost for each of these tools. I think it's on how you should budget for it. It's a good question. I want to get your thoughts on this as well, Jason M. Lemkin, mainly on the budget side. I would say for cost, one, it depends on what you're going to use it for. And I'll give a quick anecdote. I literally spoke to a $50 million ARR CMO this week, and he was like, oh, I have 10 betas going. And I was like, why do you have 10 betas going right now? He was like, I'm in a bake-off with, you know, 3 different outbound AI tools, like 2 for inbound, a couple more for all-in-one tools. I was like, maybe don't do that. Maybe my advice is to stop all this. I was like, why? He's like, you know, I just wanna see which one works the best before I spend money. I'm like, I think you gotta make a bet here, dude. Like, I think in trying to save money, you are actually spending a lot more time than you normally would just deploying these and getting the outputs of AI by doing these 10 different betas. I was like, just pick one. And if you feel like the cost is prohibitive, work with the vendor on the cost, see what they can do for you. If not, go to— fine, go to their competitor if you need to. But I was like, don't do 10 trials because you think it's going to magically save you money in the end on AI. This is like the craziest thing I've ever heard.

I think, and to be— it is too. First of all, one, I don't think you— I mean, it would be nice to have a bake-off, but unless you're going to invest the effort to truly train all of them, It'd be hard to do a bake-off with more than 2 vendors. You should do a bake-off for a variety of reasons. There's different flavors, different UI/UX, different limit— they all have limitations. They all have limitations. So I would bake, but I wouldn't do 10 because you won't put the energy into training them. The bake-off will fail with 10. I would do 2. Just like in the old days. The one thing I would say, Amelia, you could add to it is what's the pricing they ask, right? You can talk to the vendors.

Yeah.

Here's the tough part. Basically all the vendors that we're describing for this, for GTM sales type motions or marketing BDR, whatever you want to call it. They all require training. They all have a fair amount of data and really they're mostly optimized kind of around $100K price point. Now sometimes it's $60K or $70K annual and $30K to pay you for the onboarding and training. Some of the vendors absorb some of the onboarding costs, some don't. There are lower-end versions you could, but it is all in the tens of thousands of dollars a year or more, and probably more like $20,000, $30,000, $40,000. Artisan that we happen to use is launching a low-end version, right? I think everybody will, and we will be on top of that. We will be on top of that, and we will test out more of these cheaper versions and see how the more automated training works. I'm not skeptical with where LLMs and AIs are going today. I think I can guarantee you that the low-end cheap versions just, and this is true and support, which is for the long— the low-end cheap versions aren't as good. And it's not that the software isn't as good. I just did a deep dive with G2 on the CEO of Zendesk and we talked about it at Zendesk and support. The low-end agents work. They just don't ingest as much data. They might ingest your wiki and a little bit of information. And so the low-end Zendesk agent is still good, but it maybe only has 20% of the power of the enterprise version. And I suspect that's what's going to happen instead of training this. On every bit of data we have for a decade, every interaction, every customer action, it's just going to simplify what it trains. Maybe Amelia feels it differently. And so it won't be quote as good, but it may be plenty good. It may be plenty good for $299 a month or $10,000 or what. We don't know. We're going to be objective. And I think if we do nothing else, we're going to pilot the low, some low-end versions of these and we'll compare and contrast for you. But I am not aware of in right now as this is recorded, I'm not aware of any cheap AI SDR and BDR tool that works because of the training ingestion and data, but I wouldn't rule it out for 2026, but I would budget $50,000 to $80,000 or more. And if you can't afford that, don't expect much. Just don't expect.

Okay. No, it's a good point. And I think, yeah, the Bake Off thing was interesting because that was something I was just kind of shocked when I heard this. I was like, what? I know you're trying to save money, but I don't think this is the way to do it. It's just too many. I was like, it's just too many and too much. Like, you're either gonna say that they all suck or you're gonna kind of pick one and it's gonna take you too long to actually implement it. And you're— this was the CMO I was talking to— like, your CEO's gonna get frustrated with you. There's a lot of problems in doing too many bake-offs for all these AI tools in the vein of saving money, right? There's other reasons you should do it, to Jason's point, but I think in the vein of saving money, I was like, this is not the way to save money. The other thing I'll say about budget, I think the $100K range is right. Some are right below it, some are maybe right above it. Depends on what you use it for. Pricing does fluctuate based on, you know, uses. But the other thing I'll say about budget that I've come to learn from folks too, and then like trials. Yeah, a lot of these are going to be rolling out self-serve versions, right? Where it's more of a pay-to-go or You know, you can try it for a few months and see how it goes and see if you get some of the same results we do. But I would say too, like, the thing that you should budget that is maybe more important than the direct cost is gonna be time. So we also were able to reallocate budget, right? There's certain things you could do where this was not new budget I had for all these tools that I went to Jason M. Lemkin and said, hey, can I get approval for Asian Forest Qualified Artisan? But also around the time of our event in May, a few people left faster after. And so I was like, okay, instead of replacing that headcount, which was already budgeted, I will replace it with the tool. And so that's another way to do it. Like, I'm not saying to fire someone and then hire this tool instead, but if somebody—

No, but we replaced two human— the budget from two humans to support our agents. And we didn't fire anybody.

It—

I really think natural attrition is going to create your budget more than, you know, if you're going to fire someone because they don't perform, just fire someone. Like you don't fire anyone good to replace them with AI if you haven't learned anything. If someone's failed, if they can't close anything, or if your AI SDR hasn't set a single appointment, I mean, your human SDR, maybe just replace that budget with it. You know, if your SDR can't do anything, you know, you can't do worse than zero. But we'll track it. The many of you, if you're early stage folks watching this, you'll think that's expensive. And if you don't believe in the vendor, it will seem very expensive. It will, because you'll get quoted by a sales rep a lot of money and it will feel risky to you if the deployment doesn't work. You know, we, that's why we're not trying to be walking billboards, but we do share the vendors that we use. Others are good too, but if you get the right people, it's going to work, but it can feel risky. It's tough to start at $500 a month today, but you got to take a little risk in life.

Yep. Uh, there's another question for Bo, I think related to outbound, for what this question came in.

Oh, let me ask, can I add one more thing on pricing? It's interesting. All of the vendors have too much demand, including Salesforce. They have too much demand. And so you're gonna hear that manifested with some of them where they might not take your business. And The main reason— Amelia could maybe share other angles she learned. I've seen it just a few ways. She's seen it more often because she interacts more with them. If you don't have enough data, if they don't think you have a rich enough data source to train these agents, they might just— even if you have the money, they might not take your business. We've sent a lot of business to our— the vendors we use directly, or just by doing this, we generate millions— we've generated millions of dollars of revenue for all these guys just by sharing what we use. That's the nature of of the beast, but they've turned away a lot of business we've sent to them. A lot of business. So don't get flustered, but just be aware, not only is it not dirt cheap, but if you're not an appropriate candidate, it's a, it's bad that they have too much business and that it's like everything in AI, folks are overwhelmed with demand. But also if they tell you they can't support you, ask why and listen, and they're probably right.

Yep. I agree. I think that biggest other thing I'd seen is like, I either through this webinar or content we post, right? Folks will then reach out to these vendors or they'll ask me and I'll, I'll say, hey, I'll just make like an intro to you to the team. 'Cause now I know them and you know, I'll be like, hey, did that ever work out? Did this person that I referred you actually sign? Some of them do, some of them don't. So it's, they do have a lot of demand, which shouldn't turn you off, but also they're going to go through with you and the way they did with us. SaaStr is not unique to this because I saw another question in the chat. Are we getting, you know, special upgrades from some of our vendors to help on pricing that aren't public? Some, okay, now those 3 vendors I just focused on today have become sponsors of SaaStr. Sure. If Artisan hadn't, Artisan actually sponsored SaaStr before I became a customer. So if they hadn't not sponsored, I maybe I would be on a different AI SDR tool talking about it. But because they were a sponsor, I did the classic thing. If I just went to the founder, Jasper, and said, hey, I want to use this post-annual, can you help me get like the best person on your team to help me set it up? Because I didn't know as much at the time. And can you help me figure out like, yeah, what is your pricing? How does all this work? And so we were, I just went to the CEO, but again, I think if you know the CEO of a company, you're going to do that anyway. That hasn't changed in my life. And then I put it in the chat. There are a few features specifically, like beyond pricing, we're in a lot of beta features, I would say, with all of these vendors across the board. So we spent the most of the time today on sales, but also in marketing. Most of the things you'll see next week, because I'll come back and do a part 2, we're in beta features. And I think that's partially because one, this is not like a hubris thing, but a lot of vendors have told us because we have so many agents, we're a lot faster now at adopting them than than just maybe a typical SaaS company/off-the-street company. I think too, because we spend so much time with them, I think of more use cases, right? I'm like, okay, it's already using it for one thing, I want to use it for another thing. And then I'm like, wait, can it— I'm like, maybe can it do this or that? Or maybe can it do XYZ thing? And so I just email, you know, the founders or CS rep, whoever it is. I'm like, hey, Can the agent maybe do this if I have this use case as like the next use case? And sometimes the answer is yeah, that's usually the answer I would say is yeah, that's already on the roadmap, but we can give you access. Like we can have you guys test it, break it. And so I, we do get access to things, but also because we want to, I want to test these things. I have now comfortability with the AI where I want all those beta features so I can be like, okay, if I break it or if it hallucinates, I understand the risk there of testing something that's not fully released. But I will say that's my disclosure. We do have some features on each of those tools that are not publicly available yet, but some of them are like either just launching in the new year or soon. But I don't think that's unique to— maybe that's unique to Saastr, but I also think it's— there's a want there for— there's a— like, I personally have a want to do some of those things, and they have a want to test some of those things.

Yeah, but let me just— I think it's a good question because we do have a lot of success. Some folks think it's because we have so much data. We thought in the beginning it turns out to not to be true. You just need enough data for these tools to work. You don't need 10 years of data. 6 months is enough. You need a volume of actionable data. You don't need to know who came to SaaStr annual first meetup in 2012. That data really doesn't help. Okay, so it's not so much data. Do we get special treatment? I think we do. We get the best people at these vendors helping us, the best FDEs, the best onboarding folks. And so I want to tell you a counterstory. Just to help you qualify vendors. There's another vendor on this list that we thought about using. Great founders, like them. We were routed to a very mediocre sales/onboarding person who really didn't want our business. Told us we couldn't use certain features. Was a pain to work with. And so we passed them by and we never used them. Do that yourself. Talk to these vendors. If you have a bad experience, don't use that one. And don't— also, in the age of AI, this is where I'm getting frustrated. Don't get bamboozled by a sales rep that doesn't know the product or isn't technical. Ask to talk to an FTE or a solution architect or an onboarding specialist. Don't waste your time with an idiot sales rep. The bet— there is the— I will tell you some of the very best AI companies across the globe, not just any ones we're talking about, the best ones I know have really mediocre sales teams. That do not understand the products they're selling. They just don't understand how AI coding tools work and how AI support tools are trained or how any of these tools are trained. Don't stand for that in the age of AI, because a rep will tell you something that's just wrong or they won't understand it. Bypass it and say, listen, if you want my real money, I want to talk to the person that's going to be onboarding me and own it. And if some blocker that last worked at You know, a dated B2B company 5 years ago and doesn't know AI won't let you talk to that person. Find another vendor. You deserve it. You deserve it. Assume the sales rep, unless they're technical or know their CRAP or have done a deployment, if they hesitate on an answer, fuck, talk to somebody else. And so we just didn't deploy one of these agents. It's probably as good as the ones on this list. It's probably just as good because we had the idiot sales guy. He lost all this PR, revenue, promotion, friends, and just ran the deal into the ground. And you deserve to talk to an expert, an FTE or whatever it is before you sign, before you sign. And we did with Harry and Rory. We had Mark Benioff on about a month and a half ago. And we talked about what's the number one thing you wanted at Salesforce. He not only do they want, does he want to have thousands more forward deployed engineers on agent force and AI. He said he wished, obviously it's not practical at $44 billion in ARR, he wished every company could deploy their agent and have value before they sign their contract.

Okay.

And it's not practical for Salesforce. It's not even practical for the folks, the startups on this list. But you should demand as close to that as you can. You should demand that you at least talk to an expert who in 20 minutes can tell you exactly how successful your tool will be. They'll go— that he can just look— he or she can just look at your Salesforce data, look at your HubSpot data, look at your Marketo data, whatever data, pop up a little bit and say, yeah, listen, I did this with 20 other customers and clients. This will work or no, it won't work. You deserve that. You deserve that. And if you have to pay $20,000 for it, for the training, it's going to be your best $20,000 that you're going to pay this year.

Yeah. The other thing I'll say to add on to it, and then we can answer a few more questions, we're going to do part 2 next week. Somewhat related to cost, like I, like I said, we, we replaced headcount spend. So I reallocated, I shouldn't say replaced, I reallocated headcount spend into some of these tools, that will start to happen at your org. As people naturally leave for no matter what the reason, you can start to say, okay, can it— is it a role where I can start to earmark that budget and move it into a bucket that would then be an AI tool? That again, I'm not saying you have to replace your entire SDR team with AI and get these free agents that we have. You can, we've done it, but we still have humans here too. Like I'm saying the magic of this is getting the best A-tier players on your team that are still there to become S-tier with AI. So I think there's some of that. And I think a lot of folks I talk to, Jim, more common, you probably do too, Jason, just say they're not going to add headcount so that they can empower their teams either Q4 2025 or Q1 2026 with these tools. Like they're just again reallocating budget that they would have to say, okay, instead of growing the sales team from 5 to 10 humans next year, we're going to keep it at 5, but we reallocate some of that spend to the tool so that those 5 people can be S-tier. I think that's a good way to think about it. The other thing I'll say, we're not immune from people roasting our AI. And so I think how that kind of plays into cost, if you decide to roast an AI in the beta, because maybe it's not very good. I don't know all— listen, I don't know all the tools. I know a lot of them. I've seen a lot of them now. I can't say I know them all. There's so many of them. I do try— if we can't use it for ourself, I do try and at least learn all the tools and their capabilities so that when folks ask me for recommendations, I can make at least an informed answer. So I don't know all the tools, but I will say there are instances where people roast our AI, and they've roasted all of our inherent vendors because they roast our AI. I think to some degree, that's a little bit okay. And your team should be maybe a little bit skeptical and maybe the outputs— listen, the outputs are never going to be perfect, but the outputs with the human are never perfect either. So maybe if there's one final takeaway before we answer the last questions is if it wasn't working pre-AI, it probably still won't work now. If it was working or it was working kind of good and AI can get you that leverage, then as we've seen in our results, then it's even better. But it, you know, if you hated outbound pre-AI, you're still going to hate outbound. It's— hate it. If you didn't have the best support or you still don't have the best support documentation for your AI to ingest, then your support is still probably going to be B-tier versus S-tier. There are times where our agent can't answer a question and it's a support question. And I'm like, that's my failing. That's— and they should— I know people roast the AI in those times when, you know, It's happened, I put it somewhere on a slide, I think less than 10% of the time, or I think 3% of the time. Well, we'll, yeah, we'll touch on it next week. Less than 3% of the time. But when they do roast it, I'm like, you know what? It's not even a failing on the AI's fault. It's actually a failing on my fault. Like it's something they're asking the AI that I didn't train it on. I didn't tune it on, even though it has 20 million words. There are certain use cases and scenarios where people ask certain questions that our 20 million words do not have the answer to. And They were not going to have the answer to pre-AI and they don't have the answer to now. Last couple questions, then we'll wrap and we'll come back for part 2 next week. There's a few specific questions that are fairly technical, but I think they're good ones just if you guys are thinking through this as well. So the first one's on Artisan. So I'll come back to here, which we use for outbound. There's two questions on it. One, someone has said, have I played around with the lead gen contacts that it has? Related question is, have I seen Artisan engaging with duplicate contacts in a smarter way than maybe a human would? Is there anything built in to prevent bombarding inevitable dupes? So on the first part, have I experimented with their lead gen contacts? Yes, I commented on the chat. I just started. So like, to this point, if we're 6 months in, I have 99% of the contacts are ours. I just started using it to see how I, I wanted to learn it. I wanted to see how it does on completely cold outbound contacts, but that's a new experiment I'm running. So it's too early for me to say yet if the data was, if the contact data was good enough for us to keep adding. But again, I don't think you— with the beauty of Artisan, it's like, even though I used to train like the sub— you can call it sub-agents, you can call it sub-campaigns, whatever— separately, the— it's too early for me to see on, you know, outside contacts, but I do want to test it because there's other things we use when, for instance, I need to find a new email for a contact that existed in our ecosystem, which I think is a lot of people, but it's not there anymore. We have a lot of contact. We've got 12 years of SaaS renewal data, like 6 Europas now. People are not all at the same job. They may still be in SaaS and they still may be at a company I want to reach out to. And that's where I use contact enrichment to get their new contact. A lot of people, we have Gmails. I will say we've been very careful since the start to get Gmails of folks. So you're not going to change your Gmail. So I have those. But for context, then move on, which I think is a common occurrence for most folks. That's where I'm testing it. So it'll be interesting to see how that pans out. Now, on the question on dupes, a lot of the— I will say I don't get dupes within a single agent. So Artisan has— Artisan will dedupe your list across, you know, multiple campaigns, multiple agents. Qualified does the same thing. Salesforce does the same thing. Where there's a gap is because I have Curry. I have to be super careful in saying, okay, this bucket of contacts going to Artisan, this bucket going to Qualify, that bucket going to Salesforce. And then inevitably all contacts— yeah, all contacts are off the Salesforce because we use Marketo, which syncs to Salesforce. And so they're kind of all already in Salesforce anyway. And so I don't really see a lot of due for one vendor. So if you only use one vendor, You won't have the issue I do, but my issue right now is I have to be super careful in manual oversight to make sure those contacts are not getting hit by 3 different agents because those 3 agents as of right now mostly do not talk to one another. They're getting there though. Like I know Artisan just pushed, Artisan just pushed an update where they, you could turn on a toggle and you can pick what Salesforce campaigns you want it to exclude. So that's great. So it's starting to solve for this problem of they know some of their customers like us have multiple agents that don't all live there. Qualified already syncs natively to Salesforce. So it kind of has that toggle by default. So again, because the backbone of all these three is that Salesforce is the common denominator, it works. But I do think it's a, it's an interesting nuance there. Like you may not have that problem if you just go deep on one, but I've got that problem now where trying to get data across the different agents is sometimes right now a little bit manual, pushing it to a few different places. That's inevitable. But again, because we're getting good results, I think it's worth it for the interim. Last question, and then anything you want to hit, Jason M. Lemkin, in the end, let's do it. Somebody asked me about how to trigger all these. Which I think is a good question. It's fairly technical, right? Let me stay here. So they're all, again, the underlying foundations and training are all a little bit the same, but, you know, particular to the vendor that you use. And so the way you can trigger in, I'll just go through them for you real quick. The way you can trigger in Arden is you can upload a list. That's the easiest thing you can do. You could just export that. That's what I do, honestly. I just export our— again, because I don't want the contacts to be on the other two agents. I export the contacts, I upload them. That's the way I do it. That's the easiest way to do it. Now you can't— there's other ways that you can do it. You can, like I said, now you can kind of like cherry-pick from certain Salesforce campaigns or contacts that you have. You can do just like a search, like an intent search on there and do it that way. I use CSV. It's the easiest trigger. Also because I don't, I've seen this, I don't know if this is like a, again, a platform thing or a me thing, but I've tried different sizes in ours and really different campaigns and different audiences, but I see 800 to 1,000 as a sweet spot. So the other reason why I do CSV is I want to keep it in that band because it seems to perform better for whatever reason. I keep it in that band. Triggering on qualified is more so automatic. So Also, the user can just talk to the agent. So if you guys go on SaaS or London right now, you could talk to my Amelia AI. That was something where in the early days I used to, I still somewhat do it, but I used to monitor our qualified like a hawk and be like, okay, what are people saying to Amelia AI? Is she giving the right response? Is she saying the right thing? Now I've obviously built some like trust and comfortability with it and I let it run and it will tell me if it needs help. Like it'll raise its hand and say, someone's interacting and needs your help. Then I'll jump in. But now if she doesn't say that, then I just let it go full autopilot. So that's how that's triggered. I'll talk next week in part 2, how I trigger some of the emails that it's doing. I made the earlier case of if I, if it, if our agent gives you a code, it sends you an email follow-up in a few days. I'll show that more so next week on the marketing side, because there's different ways you can think about using that use case that's relevant for you. And then in Agentforce, The trigger is, I don't have a screenshot for it, such a big question, but the trigger is based on either the, you can do it on a lead level, contact level, or like a campaign level. It's, I will say that portion of it, I think, I think they just pushed it. I think, you know, as we were rolling this out, we did it very manually just to make sure it wasn't like spamming 1,000 people all at once. But then since then I've been able to say, okay, I've got like a list of contacts already in Salesforce. I can put them into a campaign and then I can attach this. I, then I could just have the entire campaign go live on the agent and it'll parse it out the way it thinks it should, right? Okay, you've got 1,000 people. I'm not going to hit them all at once. I'm going to see what time zone they're in. I'll look at the times you want me to email those folks and I'll just start to queue them up. You could do it either way. It's actually gotten easier. So I commend them on that. It's gotten easier on how to In our use case for this sales, this sales motion that we use it for, it's gotten easier to trigger this particular flow. Anything else you want to wrap on? Sorry, we have to do part 2, guys. Good. Okay, cool.

Thanks, Amelia. Great job. Appreciate it.

Yeah, you guys can keep— you can— I'm just, like I said earlier, I'm just Amelia's string. We try and answer any additional questions. I'm not on LinkedIn that often, but I will add you in a week when I check it. And then the— we will do part 2 next week, so I'll send that as a follow-up to everyone so we can cover the rest of GTM. And then I'll do a much shorter version of this at Sastra AI in London. If there's anything you want to see in particular for part 2, I've already gotten some good ideas from this. Just let me know. I will put it into part 2, and then we'll see you next Wednesday. Hi everyone.

You didn't create a startup to run a small business. Let Salesforce help you connect data, automate busywork, and empower employees on the only platform you will ever need, no matter how big you get. With smarter AI and built-in collaboration tools like Slack, which we use and you use, the sky is the limit. Learn how Salesforce works for startups at salesforce.com/startups. That's salesforce.com/smb.

Hey, SaaSer, imagine having agents for every support task. One that triages tickets, another that catches duplicates, one that spots churn risk. That'd be pretty amazing, right? HappyFox just made it real with Autopilot. These pre-built AI agents deploy in about 60 seconds and run for as low as $0.02 per successful action. All of it sits inside the HappyFox omnichannel, AI-first support stack. Chatbot, Copilot, and Autopilot working as one. Check them out at happyfox.com/daster.