How to Use AI to Hyper-Customize Go-To-Market at Scale with SaaStr's CEO and Chief AI Officer

So I know a lot of folks, again, going to Service Company want to turn on an AI SDR and have it magically create revenue. Another way is understand calmly which parts of your funnel, which parts of your go-to-market process is not even happening. Is it even remotely optimized? Where are the leads where it's just not like your team says they want to do them, but they don't want to do the small leads. They want to do the, the low-scored leads, you know, and that's what that 15% is. And everyone is like, you're going to say they don't do it. Every salesperson is managing their time. They're going to, they're going to force rank their leads in their head. They're going to put all their effort into that one big deal that's going to close this quarter, a little bit of effort into the small ones and nothing to the bottom. So if you can put AI into that scenario or the ones that wasn't worth their time, 70% open rate, even if you grow 15 or 20% faster in 2026 because of AI, it's a gift from heaven, right?

Hey, Sasser, 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/saster.

Hey everybody, Saster Annual will be back May 2026, the world's largest SaaS and AI gathering for executives. Just this 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 Saster Annual and AI Summit each and every year. But here's the reality, folks, the long 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. All right. Good morning, everybody. Thanks for coming back to Saastr AI London 2025. We really appreciate it. So we have a full day, so we're just going to keep going. I know some folks are going to stream in, so folks are going to have coffee. This first session, we will go deep on our agents, our 20 agents in particular, the core ones we use for outbound, inbound, leftbound, rightbound, and sales. Amelia is going to drive this. We will do this for about 45 minutes. But one thing I want to give folks context on. We have— if you guys have followed us, maybe you've even seen a version of this content. We're updating it all with all the latest data, all the emails we've sent, all the performance, everything that our AI SDRs and agents do. So you'll get all the latest data that Amelia has aggregated. The one thing I want to bring up, because we've heard it online and then I've heard it at this event already, is a lot of folks are like, well, that's great. This stuff works for SaaS or it doesn't, but it won't work for me. I'm not as big. I don't have your scale. I don't have your data. I don't have your history. Amelia will dig in here. It's not true. All of this will work for you to some, to some degree. Why? All of you have data. All of you have leads. All of you have a database. Very few people here at this event have zero in revenue. Very few people here have zero customers and zero prospects. Those guys go to Web Summit. Okay, this is SaaStr.ai. You guys have customers and revenue. So agents will work for you whether you have to train them or tune them differently. Emily Degan, maybe so. And maybe you don't have 10 years of data, but you don't need as much data as you think, and you don't need as much trailing data as you think for agents to work. Here's the simple aha moment, and I know a lot of you know this, so it sounds Captain Obvious, and then I will let Amelia drive this. To simplify a lot of complexity, the mistake that everyone made in AI SDRs and AI sales agents in 2024 and into this year, and even through today, and even conversations we have today, but especially is they would buy a product, do nothing, and expect it to get you millions of revenue. It don't work that way. Amelia will dive into it. It didn't work that way before Q4 when none of these products work. It didn't work that way after Q1 this year when they started to get good. And it don't work that way today. The way an AI agent works for sales, for GTM, is you figure out something that works with your team, with humans. You figure out an email, an outreach, a script, a set of objections that you overcome, a set of questions. You, you nail it yourself, just like it has always been. It's mastered since the beginning. Humans figure it out, and then you take what worked and you give it to the agent and you train it for a month, and then every day after, and you take what worked and you do it at scale. That's all these products do. Like, look, listen, there's a lot of AI, there's a lot of LLMs, a lot of sophistication. You're gonna learn about it all day. But at the end of the day, if you wanna simplify, 'cause all these products are good, take what humans have nailed, document it, and then train an agent with what works. If you're expecting an agent to sell when you can't sell, that's never worked. You gotta go back to the old days of founder-led sales, but instead of giving it to that first human you hired, you give it to the first agent you hired. But basically the same thing. So with that, Amelia, we'll, we'll dive into our data.

Okay, well, let's do it. Well, I wanted to focus today just for some of you that have been to like our podcast or watch Workshop Wednesday. I aggregated a lot of our data. I put in some new data. and then I'm going to focus a lot on hyper-customizing at scale. If you haven't seen this yet, these are our 20+ agents we've deployed. You can go to saster.ai/agents. A lot of these agents I'm going to talk about today. Because this whole guide is up, I'm not going to mention too many of the vendors because I aggregated a lot of the data. I'll talk about some of them, but they're all in depth here. So saster.ai. /agents. A lot of them are third party. Some are ones that we vibe coded. Some of these agents have subagents, which I'll talk about in a little bit. But these are our 20+ agents that we have currently right now at Saastr. We do also have more agents than humans. So I literally just saw someone we know yesterday and was like, oh, Saastr London looks so great. This is my first time here. How many Saastr employees did this? I was like, well, there's 2 on this stage there. One more full-time employee. We've got a few contractors trusted that we brought here to London, and then we have all of our agents. So we have more agents than humans. But on today's topic of hyper-customizing GTM at scale. So this is just in the last 6-ish months. So after SaaS Tour in May, so basically post-annual for us, we really started going deep on this AI journey, deploying our agents. And we wanted to see what would work in different use cases. As we did that, we deployed more agents and more use cases, more subagents, and that's how we ended up with this hyper-customized AI emails and messages at scale in the last 6 months. So our— just to give you context, our personalized outbound 6 months ago when we still had people before all the agents, The average SDR we had here as a human would send, you know, on the low end, maybe 75 emails a month per rep. And at the high end, maybe somewhere like 280s, 300s, right, per rep per month. So if you look at this total over the last 6 months, almost, you know, 60,000. So let's say $10K a month, that's 32x the max output that those human SDRs were doing in a very short amount of time. I'll talk about the ramp time of these agents, which was a little bit different than humans. But if you think about this, that means we were also only doing about 3% hyper-personalized emails at scale before this. Because the other thing I'll show you right now is the consistency between our customization of our agents versus on humans. Very hit or miss, right? You have one SDR that's maybe really good. They send really good, high-quality emails. You get another SDR, they're just not getting it, and the email's terrible. Like, you wouldn't buy from them. You're embarrassed that they sent this email, and you wish you could unsend it. So this is how you can get hyper-customized at scale. And a majority of these target accounts are receiving, you know, highly customized messages and reply on the first send. So that's something I think that is also a big learning is if you do this right and train the agents the right way, Not only is it different from a human AI, like a human SDR, but I think in most human sequences it takes a few messages, whereas on ours, because it's so custom, usually on the first email or first message we get a reply. So a lot of change to encapsulate here.

This is really good. First of all, these, the timing here is really interesting. Maybe we're going to get to it, but this, if I see this, so these are 3.

These are our 3 main outbound.

3 agents, Artisan Qualified Agent Force.

Correct.

And if I'm looking at, and that's the The sequence, that's the, we would Artisan first, then Qualified, then AgentForce, right?

Correct. So in sequence of time, we did Artisan first, then Qualified, then AgentForce, but actually Qualified's done more because it's on the website. So it's cheating a little. And then Artisan's in the middle and then AgentForce is our newest at the bottom. Got it.

Okay. I thought, it's a good chart. I thought some of the things you're saying is by the time you get to your third agent, you can also roll it out more quickly. Maybe I wasn't reading that into the chart, but you can.

Yeah.

The other thing, and then Emilio, keep going. The other thing, Emilio will show you examples of the emails and communications our agents are selling. Another sort of criticism that I hear from folks, I put criticism in air quotes, is, oh, those emails are only pretty good. Like you talk about hyper-customization, but this isn't going back to the 1980s when you bought this and you did this. Like they're good, but they're not the bar. This is important when folks say, oh, I can't do this. The bar is as good or better than like your average human SDR.

Yep.

So you will see these emails are customized depending on that. We do them, we have more data to pull from. If you were at SaaStr 3 years ago, we know about you. They get better, but, but your jaw's not going to fall on the floor when you see these emails. You're going to see they're pretty good. They get as good or a higher or equivalent response rate to humans.

Correct.

But they're not the most— the customization on a scale of 1 to 10, I'd say varies from a 3 to a 6.

Yeah, and I'll actually show you on—

it's not in the 12. This isn't an email that someone spent equivalent to spending a month of deep Claude research crafting. They're just like, here's a— they're just pretty good. But the average human and then a lot of folks on the internet are like, well, I could do better than this if I hired 30 top-tier SDRs. Maybe. Where are you going to find 30 top-tier Oxford-graduated SDRs that want to carefully craft an email each day? And if they're that good, they want to be promoted to AE in like 3 months. So they're going to cycle out and leave. So the real bar is better than the average human with 24/7 consistency. And that's what— squint— and that's what you're going to see in these emails, like pretty good with not a lot of errors. That's, that's good enough to crush it.

Yeah, the 24/7 thing I'll get to, and it's part of why it works. So at scale, you'll see in those— in that chart I just showed. So I'll break it out a little bit. Artisan overall, And we're live streaming, by the way, guys. So if you want the slides and all the data, you can also go on our live stream, but you can also keep taking photos. And for Artisan, that one has— it actually went up. It's about a 6% overall response rate now. You know, average for normal SDR is 2 to 4. Qualified has inbound, so that's about the same at 6%. And then Asian Force, we've said this has a higher open rate. I'll tell you why. It's our newest agent. It's also people we ghosted. So it kind of makes sense that they were like, you finally followed up with me. I'm like, sorry, my human wouldn't, but my AI agent would. So also, more importantly, I feel like I see people getting this wrong is not empowering their agents to sell, especially if you have something that is a low ASP, something like a ticket to SaaS for London is a great example. Between our two agents that I empowered to sell, they sold 15% of the ticket revenue for this event, which is kind of crazy because in the past when we've had human SDRs try and sell tickets, I think they sold like literally I could count on one hand. So at that volume and at that scale, I see people getting this wrong a lot, which I'll get into. But why does this work, right? It's because all of these different platforms have different things that they're using to contextualize to add value back to the prospect. So instead of making it too much about SaaS, we really try and tune up our agents and train them so that all the value is about their company, what they're doing, and why they should be at something like a SaaS. I think that's also important as you're training your agents or thinking about rolling this out, is that it should be about adding value, right? I don't know. I feel like somehow that got lost in all the outbound madness, but that should have always been the case. And it's definitely the case with AI. AI will also know your product a lot better than most entry-level, like, SDRs. And so whatever you tell it about the product, it will contextualize for them and add value. So AI does this in a way, you know, obviously 60,000 messages we couldn't do ourselves beforehand. So it's really crazy to see, which we'll dive into.

And maybe one thing to add on that data, I wrote a SaaStr post on this recently, but The agent force, 70% open rate, which on its surface is jaw-droppingly high, right? Let me dig into that for just one second. And the 15% of revenue from London. This is a theme I call having agents do the work humans don't want to do or unwilling to do. Okay. So the 15% of ticket revenue folks coming here, we, for 6 years on and off, we tried to get our human SDRs to help us reach out to return attendees to get them to come to SaaS Trail in London. They just wouldn't do it. It wasn't worth their time. They wanted to hunt a 6-figure sponsorship. So we'd ask them, we'd give them incentives, we'd give them Starbucks cards, we'd beg them. They wouldn't do it. And they said they'd do it. And then finally we would use tools like Momentum and Attentive and we see they didn't actually do it and they lied. Okay. So that 15%, here's the point. It's a lift, a lift without the agent we wouldn't have gotten. So I know a lot of folks, again, going to the service companies that want to turn on an AI SDR and have it magically create revenue. Another way is understand calmly which parts of your funnel, which parts of your go-to-market process is not even happening. Is it even remotely optimized? Where are the leads where it's just not like your team says they want to do them, but they don't want to do the small leads. They want to do the, the low scored leads, you know, and that's what that 15% is. The agent force one, if you didn't read this or catch it, that why is it— this is a Amazing and a tragedy at the same time. These are folks who reached out to us and said they wanted to sponsor Saster for 5 or 6-figure tickets, and we did not respond to them. As our human team did not respond.

Not because we didn't like round robin it, not because people kept crying. It's just because our SDRs just said it wasn't worth their time.

They mentally lead and everyone is like, you're going to say they don't do it. Every salesperson. Is managing their time. They're gonna, they're gonna, they're gonna, they're gonna force rank their leads in their head. They're gonna put all their effort into that one big deal that's gonna close this quarter. A little bit of effort into the small ones and nothing to the bottom. I mean, literally while we're here in London about 5 days ago, I wrote this on Twitter yesterday. I reached out to a vendor for a $10,000 product and I said, I wanna buy it while we're here. I don't have a lot of time. We're here at SaaS Tour London. I have 2 questions. I didn't hear back till yesterday afternoon. I was handed to another rep because it wasn't worth the first rep's time for $10,000. And that rep said, let's get on the phone and talk about your questions. 5-day response. Get on— I mean, you lost the deal, right? So if you can put AI into that scenario or the ones that wasn't worth their time, 70% open rate, even if you grow 15 or 20% faster in 2026 because of AI, it's a gift from heaven. Right. And that's what a lot of this is. Figure out, you know, it's not, it's, it's not all magic, but it is magical if you can get a lift out of the folks that just aren't being touched in your funnel. And I'll keep going. I won't take all your time, but don't assume if you're not deep on your funnel, don't assume every lead is being touched optimally. Don't assume every lead— I have an old Saster post that every lead should be treated like a queen or king. Every lead is precious. I reached out, that $10,000, a lot of money to me. Okay, maybe it's not a lot of money to people. That would be Amelia and I basically spending it out of our own pocket. That's not nothing. Okay. Even in London, like $10,000, you can get by something nice for $10,000. That's like a lot of, what's the steak place?

Flatiron.

Got a lot of flatirons for $10,000. Okay. It wasn't even worth their time to get back to us. It's always worth the AI's time to get back to them. It's always worth it. And then your leads deserve it. So keep going. But that's why there's some of this magical rate is because humans wouldn't do the work.

Yeah, but your AI isn't picky. So just on that, on that point, why is this so great, right? Like, why were human SDRs, when we had them, only 3% doing of what we did now? Part of that is what you're talking about, right? There is downtime, there's holidays. Okay, you know, you have planned time off, sure, but then somehow when it comes to SDRs, there's always a lot of unplanned downtime. For all the humans. And so I think that's something where since the AI works 24/7, I was actually just talking to Philippe who's presenting later. We were saying once you get a few agents, actually your problem becomes not having agents. Your problem is keeping up with your agents. So we'll talk about that a little bit later when you guys come back for his noon session. But because it works 24/7, this is now our biggest issue is actually keeping up with our agents. Like, I get constant Slack notifications. Your AI is doing this, your AI is doing that. And we try and watch it all, right? Like, we wanna make sure AI is saying the right things, treating people properly. So many people yesterday were coming into Zaster saying like, oh, hey Amelia, thanks for your help. I was like, did I help you? Oh, Amelia AI helped me, or Jason AI helped me. I'm like, fantastic. This is real Amelia. And I'm glad my AI agent could be there when I was either sleeping or walking my dog. So it's, it is, it's, you'll enter a new world quickly where you have to learn to actually keep up with your agents once you get the hang of them. But it is also part of why it works so well. I put an example here. I think of which one's this from, it doesn't even matter. But you'll see here, this is an example of like a highly customized email. Again, it's tying back value to the person versus it being too much about Zaster. I think in this case it's using some website. Authorization of what they were looking at and tied it back to them and what they were posting about. And then this one had a meetings link as well. So there's a few things I want to get into. I want to get into this one because a few folks have asked us on our previous webinars to create like a mind map, which I didn't have previously, so I made one for this. A lot of people ask me like, how did you think about like the rollout, right, of AI SDRs? I'm like, well, we had SaaS Re:Annual, we had a few human SDRs, and after SaaS Re:Annual, they quit. And so instead of replacing them with another human, I replaced them with AI. So this is how I started to do that. I split up basically the lowest hanging fruit, right? So I took into our, 'cause it had just been SaaS Re:Annual, so you can take something equivalent for your company. Maybe you've just done a new product release, a new year is coming up, You can do something that adds value and say, okay, I've got all these contacts, so let's say I want to follow up with them in my agent. I'm going to train my agent on these specific contacts. So it should be a subset. You should never just unleash an AI agent on your entire, entire database. Like, do not do that. It will not be customized. It's not going to work like this. Like, you do need to train either sub-agents depending on what platform you use, or do like sub-campaigns, however they call it. But basically you want to make sure that each like thing you're deploying in your AI SDR is specifically trained and tuned for that persona that you're going to go after. So you're going to make little sub-personas, right, of your target buyers. Is it a CRO? Has the CRO done something with us? Is it a CMO? Is it— have they been to our website? Have they churned? Or do I feel like their usage is down and they're about to churn? Like, you can also use this for CS use cases and marketing, like All of this starts to converge, but basically all of these things are things you can put into different sub-agents, train it, make it personalized to those leads. I usually keep it in batches of like 800 to 1,000. I don't try and do more than that. I try and get that hyper-customized on who I'm targeting with each of our sub-agents. And then I'll pick which, in our case, we have more than one outbound AI SDR, and so I'll pick which agent I think is the best one for that. I'll pop it into that agent, train it, fire it up, and then give it the sequence that I want of, okay, I want you to, maybe I want this one to book a meeting. I want this one to sell a ticket. I want this one to follow up with a lead we never followed up with. I want this one to book a meeting. And so I'll also give it different goals for each of those personas. So that's how I kind of start to map it out. Important to know though, I did not start with all of these, right? I actually have more than this, but it's important to stair-step it, right? So start with one. Like a really good one is people you might have ghosted or didn't know that you had ghosted. Another one is if you're getting really good inbound, but you're not able to completely follow up, or maybe you can't follow up after the first meeting, that's a good use case. Like, don't start with something mission critical. Like, you're going to be disappointed if you can't get it to work really quickly. A lot of these have ramp time, which I think is on the next slide. So don't start with something that's too mission critical, because then you might set yourself up for failure.

Ameri, who would, who do you think can implement something like this?

That's a good question. I have that later.

Oh, you do? All right, keep going.

Yeah. All right, these are a few more examples. The other thing I'll say too, and the reason why I put these examples, hopefully you guys can see it. So the first one, this one's an example from Artisan. I think this person came here, but literally the top is, to Jason M. Lemkin's point, is like, It's an okay— like, is it the greatest email on the planet? It's pretty good. Would I have written this? Maybe. I'll give myself a 50/50. This one's inviting them to come to SaaS Re:Annual, and literally this says on the bottom, "Hey, I'm a fan. I'm gonna see if I can come attend." Did your AI SDR write this? Because it got my attention versus trash, which just real Amelia answered. So I was like, it's a mystery for you. It's On the next sample, hold on, let me try and go ahead. Okay, then there's another example where like, just because it's an AI SDR and you are going to have all these hyper-personalized emails, don't expect that every response will be positive. Like most of the time I get like a response like the last one. Sometimes I get a response that's like this, that's like very nice, but I'm not going to be there. And I'm like, cool, I'm going to follow up with this guy later. Like, Not everything is going to be an instant yes. Like, I think that's another fallacy of like the AI SDRs is like, if you're trying to get into book meetings, sell something, it may not do that on the first action. Almost like that was true before the AI SDR with regular SDRs. So all those things, you know, to be true are still true even with the AI. Okay. I want to talk about this one.

No, just on those examples. One, you know, a lot of folks ask us, do they know it's an AI? Do they mind that it's an AI? I think even just in these two examples, you'll see a spectrum here. You'll see, is it an AI? Was the question. So I guess we're not always clear whether it is or isn't. I think the biggest— but a big takeaway is as long as the communication is reasonably good, adds some value, right? And even just a discount is value. Okay. And a little bit of connection. What you can see from the 3 emails you've already seen is people don't mind.

They don't mind.

They don't mind. And they don't, they, what they want is help. And I really think, especially in 2026, we're all going to work in these heterogeneous worlds where we mix AIs and humans. And imagining that I somehow mind talking to a great AI is backwards. A great AI is often going to add more value than a mediocre human. So like, don't, I think my meta-learning, it's a learning. Don't worry about that. Don't worry about people talking to an AI. Worry about if it's good.

Yep.

And adds value. And people are ha— they don't mind. In fact, they don't mind talking back to the AI. Forget about chat, even by email, they're happy to do it, right? You might be happy to do it too. It might be a relief to talk to an AI rather than the human that took 5 days to get back to me this week. Like, it's not a negative.

I don't think it's a negative. To that point, we did hyperchange our inbound flow. Because of AI as well. So that was mostly outbound. For inbound, we use a tool called Qualified on our website. Our old flow 6 months ago used to be you have to fill out a form on our website. I know a lot of people still have that flow today. Then we would get alerted via email and then I would manually route it to whoever I thought was the best AE based on like the company size. We don't really do territory. Company size, category of company, like, okay, do they wrap any competitors? Things like that. So we would route it and then the AE, depending on who it was, would take about 2 or 24 hours to get back to them. And it'd always be like, you know, oh hey, it's so-and-so from Saastr, saw you were interested, let's book a meeting. Like the simplest, maybe worst email response to— no offense to our team— to those inbound leads. And now this is a screenshot of Amelia AI. If you want to try it, it's on saastrlondon.com. You can talk to she/me and you can instantly talk to Amelia. You can book a meeting with her. So she instantly books meetings for our team. Honestly, a lot of the time she books meetings while I'm asleep. So I'll wake up and I'll be like, okay, and the meetings booked are, I'll just go through all the meetings that have been booked overnight instantly. I didn't have to worry about waking up in the morning, routing any leads I might have missed or whoever's in charge of your routing might have missed. And then follow up with them and then hope that the AE then also follows up with them to book the meeting. So now what I do instead, it's already pre-booked. I wake up in the morning, I talk to Dave and our sales team. I'm like, hey, these people booked. Here's the context of the conversation. Here's what we already know about them. Now you can have a better meeting with this person. You can literally skip like steps 1 through 5 and just go to meeting. Like, just pass go.

Yeah, if you, if you, And there's a lot of vendors that can do this pretty well today. If this, if this is already well understood to you and you've already implemented this, great. Okay. But if in 2026 you're still having humans qualify a lot of your prospects and wait, there's no need with AI. You and, and Milly will talk about training. You are going to have to pick a vendor and you're going to have to train it for 30 days with your data. But there's no reason a prospect shouldn't be able to interact fluidly with an AI, not even know they're being qualified. Okay. Not in this yucky objectionable dated way of whether you're worth someone that just graduated from college's time, which was, which has always been kind of gross and uncomfortable. If you've been through it on the other side, the agent should do it elegantly. Great to hear from you. Glad you're interested. Tell me about your team. Tell me about what your goals are. Talk in English through video, audio, typing. Great. You know who would be perfect for you? Let me set something up with David. He's free tomorrow at 11. Does it work for you? Right now at night when you're ready. If you don't have that on your website, implement it. This, like, maybe before the end of the year. This is low-hanging fruit. The products with AI after the, after the fours, after Claude 4 and OpenAI and ChatGPT 4, they've all gotten really good. They can all do this. Calendly on steroids and instantly qualified. It all works today. There's no excuse to not do this. And there's no excuse to have yucky qualification. Like, there may, maybe if you have a 7-figure product and there's a lot of, but even there, The AI can qualify a 7-figure product. What department are you in? What are you trying to get going?

Who's on the buying committee?

There's just no excuse to have yucky qualification. Just eliminate it. And I really think most SDRs will die next year. But the— and I know we throw around these terms differently— the BDR that sort of smarmily qualifies you, that, that's going to die. Like we just— there's no need for that with these tools. So this is low-hanging fruit to implement. And you will— if, if a prospect can have a wonderful time, get their questions answered for real without a used car salesman and book a meeting in real time, you're going to close more, whether it's 1% more or 20% more. It's much better.

Yeah. The other thing I'll say real quick on that point too, is that these folks actually like that they got to book instantly, like to your point of like beforehand, right? We'd had to round robin and do all this other thing and route it and then finally book the meeting, connect the Zoom. Sometimes the Zoom doesn't work. Like it's so instant. It's not like we're getting messages like, okay, AI, like you didn't really answer my questions. And so I had to book a meeting. They're like, it's literally like qualifying them. That's why it's in the name of this one, but it's literally qualifying them while they're on the site. So to your point, it does all that for you. Obviously you have to train it and our qualified, um, maybe it's on this slide.

And that was 130 meetings booked that you had, right?

100, yes. 130 meetings booked in, we've had qualified since August.

And going up, right?

Yeah, and going up. So August was our first month. So you can see we rolled out at the end of the month. Before Dreamforce and then September, October, November. There was a lot of inbound in November, honestly. I was like, I don't think between David and myself, who's our AE, I was like, I don't think we would've booked, we definitely wouldn't have booked this money without the AI 'cause we would've missed them and they were high followed. We were like, okay, we can see what they said to Amelia AI, we can have a better meeting. And like now we have a lot of pipe in December, Jan because of it.

So One other thing on that chart, Jason M. Lemkin.

Yeah.

A lot of folks ask us, and it's all over social media. This is the age of voice. Is voice better? Is chat better? Is video agents better?

People can pick.

Yeah, they can pick. Don't try to answer this question. You're going to find everyone is different and different types of prospects are different. I like to chat, you know, I like to clack. Other folks really like to talk. I can tell you on our, on our first agent Delphi, we've done like 150,000 chats. The ratio is about 80/20, 80%, maybe it's 85/15, 85% people like to chat, 15% like voice. But there may be different buyers. You may have more traditional buyers that want to pick up the phone. Chat obviously is the easiest to implement. Voice is so easy. It takes almost no time.

This was hard.

This is the most work. Okay. Video's a lot of work. And we've just launched it. We actually don't know how well it will perform yet, I don't think, right? But some people—

early metrics. Yeah. Some people like it. They like it. They, they like that it's— part of why we made it video was because so many people used our inbound agent and it does a little bit of outbound that I wanted to start to ramp, to literally ramp our AI like a human to be able to start to sell more. So that's why we added the video because I was like, okay, I need it to add a layer of trust where I feel like on a chat, I wouldn't, I wouldn't buy something on a chat necessarily.

It might add a layer of trust for some people.

Yeah, it might add a layer of trust of like, this is literally me. I went to Qualified for a whole day and filmed this. I think the video's dropping later today so you could see how I did this. It was a whole day.

Yeah.

I had to like look at the camera, like blink slowly and like say weird words, but it was like, it was fun for me. But like, this is why I like this obviously too, but I was like, okay, like It's Amelia AI. Some people know me from SaaS or like the pod and like it just builds trust. Whereas like, okay, if I want her to try to qualify people even more for our meetings, I feel like the video will add an extra layer where they're like, oh, I know Amelia, she spends a lot of time with her AI. I might trust that to like get me maybe even further down the funnel.

But you don't have for, for, for, if you're deploying, you haven't done yet, or you're in the midst of it, you don't have to overanalyze this like on social media. Okay. Just do it. Like, just do it. Like, the whole theme of this is pick an agent at some level. It doesn't matter which one as long as it's a well-trusted vendor for your use case and train it. And then don't overanalyze, is voice better, video or chat, or is this creepy or good or weird? Just do it. And you're probably gonna find chat's the easiest to do first because it's the, like, it just works out of the box. Voice really, I mean, literally for the openers for this event, Amelia did it on 11 labs in about 5 minutes in the green room yesterday. Voice is not a lot of work to train on your voice. It's a little bit work and video is two orders of magnitude more work. So just sequence them in.

Yep.

Don't overanalyze a lot of this stuff. Just a leader, train and go and then sequence. They all, they all work.

Yes. I will say though, it does take a lot of time. I would not leave your agents to their own devices. So I'm going to get into our top learnings here because this ties into a lot of what we have here. So just this last week, I was talking to a go-to-market company in the SaaS community, we all know it if I said it, but they're at $1 billion in revenue, not in valuation, revenue. And I was talking to their head of go-to-market, who's now also their head of AI and their head of sales. And they were like, oh, you know, we're, we're looking at all these different AI SDRs. We want to roll it out. We want to pair it with each human SDR on the sales team. I said, Godspeed. I don't think that will work. I don't think you should just unleash an AI you do not know how to use. They were literally calling me to ask me how we do all this. I was like, okay, so let me get this straight. You don't know how to use it? No, no. You know, we're learning. We're like figuring this out. I was like, valid. But also, if you don't know how to use it, why do you expect that your brand new BDRs and SDRs that you literally just hired in the last 0 to 3 months will know how to use the AI version.

In all fairness, it sounds silly to us, okay? But if we think pre-AI, in the old days, you might buy Outreach or Salesloft, and you should, and you should put processes in place, and you should build templates for your team. But a lot of smaller orgs would just hand it to their SDR teams. Pick your, pick the tool. In fact, a lot of them would even give them budget. Pick the tool you like. Pick Mixmax or Outreach and Salesloft, whichever one you like. Write your own kind of crummy cadences and set them loose. It doesn't work in AI.

It doesn't work.

And so their idea— this was a very strong, well-respected billion-dollar AI B2B company. They wanted to do the same here. They want to just hand the tool to their brand new SDRs, let them train it and figure it out, do all that segmentation and sequencing the media ad and figured it would just magically work.

It's—

if you compare it to like the old days, it's not that silly, actually.

It's not silly at all. And I actually, I talk like when I walked them through why they shouldn't do that, they were like, Oh, you're totally right. Like, they're like, okay, we like— now we're almost a little embarrassed. We thought we wouldn't just unleash it on all of our like BDRs and SDRs that were humans because now it makes sense. So like, we need to feed it. Like, we need to figure out, pick who's going to feed it contacts. Is it sales, marketing, CS? Okay, which contacts? Okay. Once it gets a response, who does that go to? And they were like, oh, we haven't thought about these things yet. I was like, okay, Think about those things first. Maybe do like a little map of how you're going to use the contacts and then roll it out. Like, roll it out with contacts right now that no one's touching. I'm like, whatever agent, SDR thing you're looking at and just buy one like you're looking at for it, just pick one, just pick the best one you think. Yeah. And like, roll out one where you're like, okay, they were like, no, no, no, we literally have this set of— everyone has this set of leads No one is touching. I was like, all right, do that first. And then when that works and you've got it figured out and you've got it tuned in, like, you know, 6 weeks, then you can start to roll it out to like the leads and then, you know, stair-step it, go from there. Also, why this worked is we have a single source of truth, and that's what I told this company as well. I go, you need to have one single source of truth of which agents get which contacts. What follow-up are they getting? What CTA are they getting? What do we do with them when they close? You know, the AI doesn't really ask for a lot of commissions, so you got to figure that part out. Who handles the renewals if the AI closes the entire thing by itself, right? These are all things you should think about as you're deploying it. So that's part of why ours worked so well. I don't really like to say orchestrating, but it is true. We do orchestrate all of our agents through the two people on this stage here. And decide which contacts go to which people. On the amount of time it takes to get these set up, so each of these took about 2 weeks, except for the video thing I just showed you. And that's on purpose. Like, if you get frustrated in a cycle because you're like, oh, all these AI companies are like, you need 2 weeks, you need a month, you need 6 weeks. I'm like, it takes time. Like, literally, people give out like, oh, can I set it up in a day and it'll just work? I'm too busy. I'm like, you gotta make the time. Otherwise, it's not gonna work. Otherwise, it's not gonna work. Each of them require maybe different things for you to do while you're warming up and tuning the different AIs. But again, a lot of this is also on you to see, okay, like in the early days, like we read every single message, every little thing that our AI was saying, we would read. Now I spot check and I have like flags in play. Like basically it'll like throw a flag on the field if it needs me. And a lot of the times, again, if it books a meeting, it's automatic. I don't need to read it. So you can start to build trust with your agents the more you get into it. I mentioned this a little bit earlier, but I think for self-serve models or things that have a lower ASP, I think it's a big oversight that I see a lot of companies, they roll out things like a qualified event or like an artisan and they don't let it sell. They're like, oh, you know, I'm going to, I'm going to book a meeting for every $500 thing because I don't trust the AI. I'm like, Have you tried to trust the AI? Have you tried to give your agent something where it can sell within like certain bounds like ours has? I was like, it works pretty good. I like if you give it certain guardrails and you let it sell, it's actually a pretty good seller.

Just two things to, to, I know you'll get something in the second point. Who do you need to, I, I know you're gonna get to the next one, so let me summarize it quick.

Yep.

And then this last point on lower ASPs. Who do you need to have success? I know you're going to get to it. I think you need two people to make this success. One, you need someone, generally speaking, at the vendor to help you deploy it. A forward deployed engineer, solution architect. They go by different names. Everyone on social media, everyone is for startups is talking about forward deployed engineers and what that term can mean. A lot of things. It probably doesn't mean for you what it meant to Palantir where they invented it. But it means someone that can work with you and your training and get it into production.

Yep.

That's why a lot of these tool— that's why self-serve doesn't work that well. And that's why a lot of the— one reason some of these tools are relatively more expensive.

Yes.

Is because you need a human helping you onboard your product. Okay? You need one. And Amelia will touch on that. If you can't get that help, don't buy the vendor.

Don't buy it.

Don't buy the vendor. No matter how slick the salesperson is, AI or human, if you're not going to get the help to train and deploy, I'd rather have a worse vendor that will give me the help day and night.

And then I trust.

Okay, then the second point is, okay, then, then, and then on your side, this is— well, we're out of time to talk about it today. We'll talk about it more.

Yes.

One way or another, you need a GTM engineer in-house.

Yeah.

Now, is it— are there a million of these folks running around that you can hire on Craigslist? No. Okay. And does that mean different things? I mean, we'll have Clay here later with Amelia. For them, GTM engineering is kind of an onboarding sales role. That's different. In-house, you need a nerd. You need an AI nerd. Okay. They, they could come out of marketing because, you know, obviously a lot of this is an overlap between marketing and sales. So if you have a technical marketer, if you have someone with a little bit of B2C background or a HubSpot nerd or someone like that, they can probably do this. Anyone that's built multiple complex campaigns can do this. Anyone in the market that's senior enough that they just built a lot of campaigns. This is basically a lot of campaigns with AI. Okay. Can someone on RevOps do this? If they're techie. Can your average RevOps person do this? No. Can almost anyone on your sales team that is not in RevOps do this? No. Okay, so slow it down. Try to find the one like GTM nerd on your team, promote them and have them own this. But you need— it needs to be that nerdy person that can implement software. It's not new, but you need that. You need the vendor to help you and one person in-house.

Yes.

And the last one on self-serve. Look, a lot of the vendors that are out here and will be at Annual and that they're rolling out more self-serve product, but it's early because of training. It's early. And I just did, I did a, I did a, with G2 maybe 3 or 4 weeks ago. It's on G2's website. I did a deep dive with the CEO of Zendesk and Zendesk has obviously a lot of AI and they do have a low-end self-serve version. This is interesting. He said with our top customers at Zendesk, we can automate 60 to 80% of support and related interactions. Okay, makes sense. They train them for sometimes for months at Zendesk for their big customers. For our self-serve AI, it's 20%. It's 20. Now, 20% is still meaningful, right? But 20% probably isn't enough for an outbound sales tool. You need to go further. So this is— just bear in mind, this is gonna come, but training, iterating on autopilot with no human, it's not there yet. It's not, it's not quite there. As we record, right? So don't expect, don't expect that folks are going to— they're getting there. I think in 12 months it will be exciting, but we're not— you need humans on both sides.

Yes.

To make the AI work, you need two humans.

You need two humans. My point on that too, actually ties into learning number two, actually, is do it to humans you think will stay at your company. Because literally I was talking— no joke— I was talking to a CMO maybe 6 weeks ago But, uh, they're a $50 million ARR. They're like, he's like, oh, I'm gonna bake off with like 10 different AI SDRs. I was like, first mistake, don't do 10 bake-offs, too many. Like, if you're gonna bake off a few AI SDRs, maybe pick like 3 at most. I was like, that's mistake number one. I was like, this guy's already doomed. But guess what? I went to go look this guy up. He's already gone. Ah, you don't. If you're— was it— if it was true before AI, it's probably still true now. So turnover in GTM was high pre-AI, it's still high now. It actually might be higher if you believe all the LinkedIn hype. So don't stake— if you're a founder here, don't stake your entire like AI go-to-market strategy on one person like your CMO, CRO, if you even sniff hint that they might leave January 1st, which is not far away. Like, you're going to spend the next 4 weeks arguing.

Yeah.

And they're just going to leave on January 1st. So just make sure if you're going to stake your AI or if you're going to build clone or found, you know, AI agents that you deploy around clones of your team, that you're damn sure they're going to stick around and they have a real stake in the company and a real reason to do so. Because otherwise you're going to be left retraining all these agents over again when that person leaves. And then to my last point, because, you know, we got to, we got to go. My last point to yours on picking a vendor, don't enter too many bake-offs, as I just said. When it comes to our budget, we've already touched on this. We have some of these tools are like, you know, $40K to $100K a year-ish, but we replace headcount. But some of the marketing side, which we didn't get time to today, but we have a YouTube session about like our whole marketing go-to-market stack. Some of those are $20. So, like, don't fall into the trap with the vendors who don't help you. So, this is me at Salesforce Tower before, like, before I got on the plane here. They have this giant Christmas tree. That's me. This should be everyone's sales team now. So, it's me, my FDE, which I didn't have, you know, a couple weeks ago, so I now have a forward-deployed engineer, my account manager, and my sales rep. And I'm like, that is the nucleus now. That is the new nucleus of everyone who's on our primary account team. Should be these 3 people, right? You have somebody to escalate to as a CSM. You've got your normal sales rep and you have your FDE. And like, I know, like, I literally text with all these people now. Like, I've gotten to know them. I trust them. I trust them with our data. I mean, in this case, it's Salesforce, so they've always had our data and they use it in Agentforce. But if folks say they don't want to help you or they can't help you, I wouldn't be too eager to give all your data to them. For the AI agents, like just pick folks you trust, talk to them, talk to the humans, ask them for customer references. Like I do references for all of these folks all the time, like all day long. I do like either short references, long references. Like I talk to these people all the time to do references for some of these products that we've mentioned. So like ask them for one. I think too often now, again, in the age of AI, people skip this step, which is maybe now more important than ever. Like just ask for a really good, solid customer reference, a logo that you might know. See if they give you one. They don't give you one, go on to the next vendor. Pick somebody you trust. And then on that, since we didn't get to all the questions.

Yeah.

Yeah.

Oh yeah, we— I think we're out. Well, we're going to take a break here and flip the room. Yes, we will do more on this at 10.

So yeah, so literally we're just going to do a little movie with questions.

Two things. We're going to do a follow-up session on Zoom in like 2 weeks. So bring all your questions and come live. And we'll also vibe code a little thing when we get back where you can type your questions into. We'll answer them either in text.

We'll answer all the questions.

Every— it's a lot of people in this room, but we will answer every single person's question on all of this when we're back and we'll do a follow-up or maybe even two live for folks in this room. Yes.

Hopefully this is helpful.

Thanks everybody.

Thanks guys. Support tab. One that triages tickets, another that catches duplicates, one that spots churn risk. That'd be pretty amazing, right? Happy Fox 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 Happy Fox omnichannel AI-first support stack—chatbot, Copilot, and Autopilot working as one. Check them out at happyfox.com/daster.