The Agents Live: The Agent-Led Growth Playbook

All right, good morning everyone. Thanks for tuning in to SaaStr's AI Day. We've done a few of these now. We've got a great show for you today, uh, solid lineup coming. We're, we're doing the kickoff, if you will, um, to AI Day. We've got 3 more presentations after this, so make sure you are tuning in for those. You can go to saastraiiday.com, uh, to see the upcoming sessions and schedule. We'll hit it at the end. So that you can join those as well. You can also just watch on our YouTube and our livestream, as many of you are doing now. So good morning. We wanted to do a very special edition of The Agents kind of live and go through just our latest findings on our agent-led growth playbook. You know, we've, we've talked a little bit last week on The Agents about agent-led growth. Now it's ask your agent instead of ask your developer. So So what does all that mean? How can you really grow with agents from the folks who have 21+ agents, 3 humans, and a dog? So I'm Amelia, I'm here with Jason, and we're gonna kick it off. Um, for today, since we're doing this as a live one versus recording, you guys please feel free to drop questions either in the livestream, um, or in the chat if you're on the Zoom with us live, and we will leave time for questions at the end. It can be Also on things we've covered previously on the agents, if you have questions about that, feel free to drop it into the chat or the comments. Either way is fine. And then yeah, you can ask questions on anything we're presenting today too, and we'll leave time at the end. So without further ado, the agents—

Also just one quick note, there's always a lot to process and all these, this is going to be great. But if it's overwhelming, if you go to sastr.ai/agents, we have a playbook that we update constantly. And it doesn't have today's session, but it has everything we've done on our agent GTM stuff. until today, and it will soon have this session. So you can always download that playbook, share it with your team, get refreshed. If you process by reading rather than watching, um, it's organized for you. It's got all the case studies, um, and we literally, we literally just updated the last couple days.

Yep. Yeah, it's a good call out. Um, cool. So with that, let's get started on today's agent-led growth playbook. So what do we mean when we say agent-led growth? I think about this in really one of two ways. So Way one is, you know, using AI agents to help grow your go-to-market, as we have with 10K, who's our main go-to-market agent. I would say way number 2 that is now getting traction is how to get picked by other agents when you ask it for something like, hey, what are some of the best events we can go to? And then if it's a disaster, you know, we're getting picked by the agents. That's very purposeful in how we've done that. Um, and so, you know, you— there are really 2 ways to combat this. So we, we've We've pulled out a few examples to show both. I'll try and hit as much as we can, um, in the limited time we have. So I wanted to first start with just a hierarchy of our agents. Um, to Jason's point, if you download the playbook, it'll deep dive into a lot of these, but just our latest agent landscape, if you will. Everything that 10K, our AI go-to-market agent, our main agent, is hooked into. He is custom built on Replit. If you haven't seen 10K yet, you'll see him today shortly. This is Everything he's hooked into, plus there's some 36+ additional tools on top of these tools that are listed here that it's hooked into that just didn't fit on the screen. So there's quite a bit to digest just here. But 10K, as you can see, has become our powerhouse. And one thing I always like to remind folks is 10K did not start this way. He literally started as just a helper for us. You know, doing help me do some outbound, help me do some get off of Notion and do some some data a little bit better, so I don't have to keep logging into Salesforce and Notion in three different places every Monday. And so he's definitely grown over the last you know sixteen months that we've had him her whatever the agent as our main agent. And so a lot of this is plugged in now. You may say okay like I don't I don't have thirty plus tools to plug in and connectors, and that's fine. This is our stack right? Some of these are duplicative, which we'll talk about, and some of These are unique in that they are—they're covering everything you need to basically run go-to-market. That's inbound, outbound, renewals, contacts, marketing, you know, newsletters. And then some of this is enrichment, so you'll see that here. And then some of this is quote to cash. So things like our billing and QuickBooks and bill, etc. So everything you see is really. Wired in so that 10K can fully be end-to-end in our go-to-market. And so that's why we do have so many. Obviously there's, there's also, depending on how you count, there's 6 different things hooked up into Salesforce as like sub-agents. So depending how you count, there's a lot here, but again, didn't all start that way. We started with just a few things. Some of these are custom connectors, but as you can see, because he's hooked up to so many things, this really lets 10K automate A lot of our go-to-market with approvals, with guardrails, but he has all the tools basically he needs. This is what you would give, you know, maybe a new hire, maybe not day one. So same thing like 10K, you know, we, we've stair-stepped him into how many applications he has now, but he has grown into his role much like a human would grow into their role in, in leading our go-to-market. And so, um, lots of great things here, but one thing I wanted to show you guys in particular with all of that hooked up, um, is what I call the Prospector tool. So with all those tools, what I've really found to be beneficial in kind of like this agent-led growth motion is having one centralized surface or place or queue, whatever you want to call it, to monitor everything. And so this could be as easy as you guys pulling up a Replit, a Lovable, a Vercel, whatever you like to— a Base44, whatever you like to use, and saying, hey, I wanna hook into, let's say you have Salesforce for sales, but HubSpot for marketing, and you wanna hook into those 2 things and you hook up whatever else has, you know, where's your inbound coming to? Is it one of those 2 places? Is it just an email inbox? How is your outbound being sent? Hook that tool in. Like at the very basic, if you hook one of each of those tools in for inbound, outbound, and maybe you have some tools or maybe it's just spreadsheets for renewals, you can hook it into something like this so that With the agent's help, you can start to monitor everything and go to market in one service. And this has been really beneficial to us given how many tools and agents we have. Just in the, you know, having 3 people and 21 agents, a lot of things were falling through the cracks, right? Like even though we had an agent on inbound, we were still having people who contacted us manually via email. We had people coming to our, um, LinkedIn messages or our G, but just emailing us. We had outbound that was going out from 4 different tools. Now that might not be you, but it's, it, it would still be nice probably if you agreed to have all those things in one place so that an agent can help you filter through any noise that you might have in those inbounds or outbounds and make sure you are responding to any current customers, you know, renewals, support tickets, everything. And so we have everything now in one place. And so how yours might look, For your prospector, as I call it, tool might look different just depending on how you have it and what tools you're using. But I know it sounds complicated. I truly believe anyone could do this today if you don't have an agent already. You could pull up RepLitter Lovable, hook in, you know, one of those tools in each of those places. Just pick one thing, even if you wanted to pick one thing in inbound, outbound, and renewals or support, and hook it into one place so that you can have a view where you can see everything and then decide how you wanna action on it. The agent will help you either do some of those actions autonomously or escalate to a human as needed if you feel like it needs a more manual follow-up. And so having everything in one place in our prospector tool has been really awesome for us just to have everything in one place. Everything under the hood now comes into this queue. So we have everything that's hooked up for inbound, everything for outbound, every lead, every account. And the nice thing is like the more you use something like this and hook up an agent to do this, which is, I know I'm probably oversimplifying, but it is quite simple. You can get a better view and the agent will actually get better on how to answer these things, right? Like, so the first week I used Prospector, I only hooked up inbound for the first weekend. And that's to say like, you might not even wanna hook up all your surfaces at one point. You could just say, hey, week 1, we're gonna do inbound, week 2, or maybe week 4, we'll add outbound and then we'll add renewals and support so that eventually everything's in one place. And you have this bird's eye view of everything going on in go-to-market. It does mean that everything is gonna get followed up with, whether it's you or the agent, until it reaches its goal. And so I feel like that's an important distinction because with the agents, they are relentless in a way that they are 24/7. You give them a specific goal for each of these things. So our agent knows that, okay, when it's inbound, I'm gonna try and route that to either Get them to book a meeting, or if it's a question, I need to route that to a human if I can't answer it. Then if it's outbound, obviously the goal is to book a meeting or buy tickets. That's a very different goal, and the agent will just keep going until it sees that it has reached that end goal with that one specific account or contact. So until it's, you know, close won, close lost, or case closed on a support ticket, the agent just gets better and better at learning your business at what Flows, how many touchpoints. Like the first week, it might not know. Okay, how many touchpoints do you need for an inbound versus outbound versus support case? It will learn and grow with you as you are answering everything in one place. It'll also get better that it's centralized because then you know maybe it sees things of I see a lot of the same questions or I see a lot of the same patterns in your outbound. What's working? What's not? The agent can start to collect that data the minute you have it wired up, and so having it centralized. is super helpful into something like a prospector where you can go through the queue and just action all of your go-to-market leads together. So I'll show it to you guys next. Um, I'll, I'll go through the inbound component first. So this is the inbound component because I feel like this is the most applicable to everybody. Everybody on this call has some sort of level of inbound, even if it's a low volume, maybe it's a high volume of inbound. Everybody has website visitors and inbound coming to the website. So I started with this use case because it is something that Everyone can use, and there's a few things hooked into it. So I'll explain it first, and then I'll show you guys. So in our inbound queue, not only is it centralizing everything into one place, it's also learning about that person simultaneously via the agent. So the first thing our prospector will do with an inbound lead is it'll give us any of their data. You know, if it's was it our contact us form? Did they write a message? Should they download our? Sponsorship prospectus? Did they put any notes in there? Um, did they tag, you know, what their budget is? Anything like that will come in. Then it will research the pitch. So this is where it will first search our first-party data. And again, if you have nothing hooked up day one, it's okay. I wouldn't sweat it. Like even if you just have like a little bit of data from Salesforce or spreadsheets, whatever, give it to your agent so it, it can look up some first-party internal data. So for us, These are a lot of tools that are connected for first-party data, but it's basically checking to see, okay, do we know this person already? Right? And I feel like this is nice because a lot of agents today, especially for outbound, only use third-party signals. And so having this first-party data is really nice to get that historical account context that you may not know, or one of your reps may not know if they just got assigned this account. And so it's nice to have the agent and the human have that context before for your responding to the person versus just responding blindly. So that's, that's why it's wired up first that way. And that's the first thing it'll do. The second thing it will do is look at our external signals. And so for this, we have a few different things hooked up. We have Exa, which I think everybody uses and knows now for third-party signals on company data. A lot of agents like it. A lot of agents recommend it. If you ask your agent, what should I use for third-party data tools? It'll probably say Exa. So a lot of us are on Exa. It'll look up company data on that. Then we have Sumble, um, which is hooked up to show us more so company hierarchy, but it can also enrich emails if we need to get like a, a, a stakeholder's email because the person who inbounded is probably not the ultimate decision maker and maybe you want somebody who's their boss. We have Sumble hooked up to do that right away. And then CoreSignal, I don't use this as much on inbound because, you know, with inbound. We, we have a certain amount of data when they're inbounding to us, right? They're usually giving us some sort of question or have they've raised their hand. And so we have some sort of signal there, but core signal you can use to look up the actual person who's inbounded their LinkedIn and just see if that tells you anything about, you know, okay, maybe if they didn't put their job title in our inbound, we can look that up fairly quickly and just see where they sit in the org to help with our, um, response. Then what it will do is it'll take all that data. So basically the first step is just research. The first step for our agent is, okay. Someone has raised their hand and said they're interested in SaaStr, I'm gonna research them and do background history. Then what it'll do is it will personalize what we call our event sponsorship prospectus. This for you could be a, um, a, just a, a pricing page or a pricing and features page. It'll generate for us a personalized page and website where it can send, it can use all that data to write them a customized intro. And if it feels like it has a fairly good sense of their company, for us, it will actually give them suggestions on what it thinks, either, either what it thinks they're interested in or what they've said they're interested in. So it will directly answer their question of why they inbounded. Okay. And then the last step our agent will help us do is create a hyper-personalized email with, you know, everything from step 1 to step 2, these all daisy chain together, and then it will help us draft an email. Using all that info. I've noticed, you know, usually when I draft with AI, I do make edits to it. As you guys know, usually when you're writing copy with AI, it's really long cuz it just has all day to write copy. And so I usually shorten it. I'm like, no one's gonna read that essay of why they should be at Zaster. And so I usually kind of keep it just short and sweet. I send them, I send them the personalized, um, webpage we've made them. And then, you know, I, and then we include our custom meetings booker. So let me show you guys a little bit of it. So this is 10K. Say hi to 10K. So this is— let me go into our prospector. This is like our dashboard. So I'll go to our prospector here and I'll show you guys a couple different things. So let me show you on here. Yeah, cool. All right. So I have here basically a demo replay of a real lead. So this was a real lead. We just changed the name. This 44 was a real lead who inbounded to us. We've talked about them previously, um, but I'll just show you guys more deeply so, um, you can see the full process of how it does it. It obviously, it cheated a little because this was the lead that already came in. And so it, I made a demo version, but you guys will get the gist here of what our prospector does. Um, and again, this is all just to respond to an inbound. Like the way we used to respond to an inbound was just probably how many of you are doing it now. You get an inbound to the website, you just respond to them like you do. You know, there's no, there's not really a lot of maybe like a lot of care and thought that goes, I think, into responding to inbounds normally. It's just, oh great, I have somebody who's raised their hand. Let me get them the info they need. Let me get my meetings booker and then hope that they book a meeting. And then this way, using the agent to your advantage is that it can look up any account contacts that maybe this person doesn't even know, like This person from Base 44, like in our, in our first party data, we saw that a bunch of like their CEO and leadership team came to SaaStr. They didn't even know that. Like there, there's gonna be things that the agent surfaces that maybe the person who's inbounding doesn't even know. And it's nice to have that historical context going forward of, and for this person's case, they inbounded. Um, you know, obviously we work a lot with vibe coders, but they inbounded. Um, I gave them this context, they went through this flow and they booked a meeting within 10 minutes. Like, not to say that everyone who's come inbound since then has booked in 10 minutes. But I think for the ones where it actually matters and they care about what they're inbounding, why they're inbounding to you, it is a good way to bridge that gap, to give them that context and just have a better foundation when you start to go into something like a deal cycle. So, um, again, very nice to have. So for us, so I'm going to hit research pitch. Um, and this is what it would show you, right? Like, okay, so right now there's. Um, an opportunity, that's, that's only because we've gone through this whole process with them. So right now there's an open op. Originally when we opened the research pitch, there was no open op. But like I said, you know, it found some attendance from them. Um, it says they're not in our, like, they're not subscribed to our newsletter, so they're probably not watching this, but that's okay, right? Like good context to have. And then you could see here, I can just scroll down and I'm like, you know, there's, there's some context there from us, but there's not a ton. So this is where it's nice to have third-party tools hooked in so I can look on Exa. You know, they're a vibe coding platform. I sort of knew that when they inbounded. I was like, that's, that name sounds familiar, right? Like they compete with Replit and Lovable. Um, but I actually forgot that Wix bought them. Um, and that was something nice that the agent reminded me like, oh yeah, they were bought by Wix. Um, so that is an interesting thing that, you know, I as a human had forgot, but the agent remembered and reminded me because when I went into that first phone call, I was like, okay, like, do we need to deal with like Wix How do we need to deal with Wix, the parent company of Base44? Like, that is something you can just ask the person on the call because now you have this data that maybe you previously wouldn't have remembered. Um, or maybe you would have to look up manually, right? If you're just looking one by one at your calls for the day and you're like, I've got Base44 on the doc. What do I need to remember? Oh, wait. Yeah. Um, you know, Wix bought them. I don't know. My agent already told me that. And my agent actually will send a version of all of this. Um, it will send it to the prospect, but then if they actually do book the meeting, it will send me a brief to remind me all of this 10 minutes before the call so that I don't forget any of the account research either, which is again quite nice. So then I don't have to remember all this by hand. And then, you know, you can look up Sumble just so confirm they're Base44. It wants to make sure it's the right company. Um, cuz there actually are like other Base44s out there and how we have ours wired up. Again, this is prevalent to us. For you, you might be like, maybe I need to know if they have a CTO. Maybe I need to know who the CFO is. For us, I like to know who their leadership and go-to-market team is. And so it's got, you know, the CEO. They don't actually have a CMO, so that was interesting. But they do, and they don't have a CRO, but they do have an SVP of sales. And they've got, you know, the founders there. So, you know, there's about 88 people in go-to-market. That's interesting to know. Again, right? These are all helpful things for us to know when we're having the deal. Not necessarily that I'm going to use all of this data. Data during an initial or maybe subsequent call, but it is at a glance. This doesn't go away, right? So like all this account research we have is now just enriched in our, in our knowledge of Base 44. I could look up CoreSignal if I need their LinkedIn. So, so then what it will do, it's gonna take all that research, it's gonna personalize their prospectus. I'll show you the one it already made, but you could see here we can change it like, okay, this is their narrative based on what the agent found. I don't have to write any of this. This is where it wanted to, you know, I can pick our relevant packages. So for you, this might be, you know, you could package together different pricing, different features based on how much you think they're gonna use, and it's gonna put it into this unique link that it makes for them. And so for them, it's going to open this guy. And then when you scroll down, you'll see, you know, so it's got this This is a version of our prospectus, but it's personalized. So you, you know, this is the top. Fine. That looks the same for everybody. Then you come down and you— there is a custom introduction written that has some of the account context. It's got our recommendation of based on, you know, what they inbounded on, what we thought the packages were that they're interested in, or how much they are altogether. And then here's all the details of them, like This has gotten sent around because we have heat mapping on this, and so we can actually see if prospects engage with it. And so that's another layer where the agents can help you. If, um, whenever somebody views this, it will say to us if it knows them, like, hey, you know, Base 44 is viewing the prospectus, um, and they spent, you know, whatever X amount of time on X page. Maybe you should reach out to them if they're, if they're stuck on something or if they have a question or Just so that you know they are looking at it versus not looking at it. So it's nice to know. And also they can, they can chat to Amelia AI here if they ever need to, but it's all linked together, right? So this is the next step that our prospector will generate. Then it will generate a meeting for us. Again, sometimes these are a little bit long, but it will show them their prospectus and our customized meeting booker. Which I'll show you guys in a second here. Let me just open this tab. And then that is what we will send. And so the agent actually does let us pick to, um, if we want it to send it from that dashboard, I can just click that button to send the email. Um, or I can actually copy paste. And the reason for that being oftentimes I've seen if I already have a thread going with somebody, I may not necessarily want the agent to just send it willy-nilly right away. I may just wanna reply. Um, or draft it as a reply. Um, so that's where that comes in handy. So let me show you guys our meetings booker, cuz I think we had a few questions about this the last time we showed it. So, all right, I've got it open. I'm gonna show it here. This is our custom booker that, um, our 10K agent built as a replacement, if you will, for Calendly and Read AI, which we were using to book our meetings. Um, this is in demo mode, but usually when we do the prospectives, it'll say their company names here, so it already knows who they are. It'll tokenize it and then I can actually pick, and the agent can suggest who they should meet with. So there's a version here that's meet with Amelia. So this is just my calendar if it thinks they should meet with me. There's a version that is, um, a SaaStr version, which you can pick who you wanna meet with, either me or David or both of us. If you're like, I wanna meet with both of them, it'll give you our joint calendar here. Um, and this is nice because again, this is a meetings booker that is a little bit more fluid and then it's It's tied back to that prospectus that you guys just saw. So with everything hooked up, it has, you know, all the data of, okay, if they book a meeting at the end or if they inquire, it'll go to this same booker page. So it's all linked together via the magic of the agents. Jason, anything to add here?

Um, no, no, I, I, I'll add more, but keep going for now.

All right, cool. Um, I'll briefly talk about, um, the next component of the 10K Prospector, which is outbound. So we have, which you guys might have as well, if you have a lot of reps and you have reps who maybe like different tools, I think we have like 3 different tools that we use for outbound, um, plus any manual outbound that we're doing. And so again, very nice to centralize it. You can see it looks very similar to the inbound queue. But this is what I call the follow-up queue. So we have outbound agents that are sending emails automatically on our behalf as agents. And so where I saw the value in this was actually not putting a centralized repository, if you will, of everything the agents are sending out, because they're sending out a lot of messages. And I was like, that's maybe data overload. But the value for me truly lies in using the agents to help with the follow-up. So like what happens when you do get a response? Too many for outbound messages. What do you do? Um, do you log into— like previously we were just logging into that platform, responding, and then same thing, right? You kind of hope they book a time or a meeting, or in our case buy tickets, whatever it is, whatever the goal is that we're trying to get them to. And so for us it became really handy actually not to have just all the messages in our outbound queue, but the follow-up messages. So basically what it would do is it would flag to us now if somebody is Um, has followed up with one of our outbound messages, it'll come here. And the moment it comes here, a timer actually starts. So the agent will start a timer of, and you can see here.

Hold on, Amelia. Just, just, um, just to step back just because I'm a little, we have a bunch of outbound tools. We have our own, we have AgentForce, we have Monaco, we have Artisan, we have other, we even, other tools even send emails like Qualified, right? For now, right?

Yep.

Sorry, this isn't unifying those. This is unifying our responses to them or just a little confused?

So it is unifying our responses to them. So it's unifying those in the sense that as soon as somebody responds to one of those, it comes into this queue. What it doesn't do at the moment is it doesn't bring in all, just all the outbound messages that are going out from our 4 agents because it's just too much data.

And how does it, sorry, and how does it, so for the initial emails, the initial outbound, the only source of truth there is probably Salesforce, right? Because they're all writing to Salesforce. Is that right for the most

part? Yeah, that's right.

When they come back in, since they've been sent on different platforms, how do we unify it here? How do you— didn't they go through different email marketing? I mean, email outbound platforms. How are they unified here? I don't get that part.

Yeah. So they all have each of those platforms you just named. So each of the 4 has their own API or MCP. And so our agent is just—

You're pulling them all in. You're pulling all of them all in. Okay. All right. Keep going. And then we're sending the follow-up through Prospector, right?

Yeah.

Exactly. You're unifying it here. Otherwise, it's all disjointed.

Yes, exactly. So we're unifying the follow-up in that, yes, all these tools are, we have them all configured on each of their native platforms to send messages through their agents autonomously. But here, when that person replies, this is where Prospector will bring it in via the API. Normally one of, like, one of them's on MCP, I think. Um, but it'll bring it in here so that, again, it's, it starts a timer the second they hit, um, a reply to us so that we can decide how we want to So very similar to the one you guys just saw for inbound. On outbound, we can do the same steps where we research the person. Usually for outbounding them, we probably know a little bit about them if we're outbounding them, and we know why we want them to come to SaaStr. But basically what it will do is it will create a record. So for us, we do a lot of outbound, right? So I don't want to— I don't want like 10,000 Salesforce records every day being created for everybody we outbound to. I only want a record created if they actually if they actually do something with us, take an action or reply to one of our outbound messages. So when it comes into this queue, it's creating a record, it's starting a timer, and then it is with the agent, um, automatically generating, you know, a follow-up based on what— whatever we outbounded them to for. So it could be for tickets, it could be for sponsorships, um, it could be for speaking or something else. And so it will help us draft that reply and then The nice thing about that too is it will start another timer from the moment we reply to the moment they reply again, and it'll tell us roughly when it's time to ping them again if they need a reminder, if they haven't taken any action. Because again, agents, the nice thing about agents is they are just relentless on trying to hit their goal. So whereas a human like me might forget to follow up with somebody a second or a third time, right? Like you're gonna, if you outbound to someone and they reply, chances you reply the first time are pretty high, right? But I think chances you follow up maybe the 3rd or 4th time because they didn't reply to your initial reply drop dramatically. I don't have the number. We could look at the numbers. I don't really have the numbers, but, um, I just think as human behavior, you're more likely to reply the first time you happen onto somebody and they respond to you. So if I happen onto Jason and I get a reply from him, I'm going to respond pretty much as quick as I can, right? But then if Jason doesn't actually do anything for, let's say, 2 or 3 days. I might forget that I responded to Jason. So the agent is nice that it's got this, it's got different tabs on this queue. You'll see here there's active, there's ready, there's waiting, there's some that need review. So those are ones that the agents have flagged that I need to review versus it just sending a message on its own. There are some that are, it's marked as closed loss. There's a bunch in the sent. And so it will remind us every morning, hey, You emailed Jason 3 days ago. He hasn't responded yet. It's probably time for a reminder or another follow-up to him. Or, hey, you sent already 2 follow-up messages to Jason. He still hasn't responded. Should we try LinkedIn or something else? Or should we try a different POC at the same company if this is a really important contact to you? Again, the agent just gets better the more it learns about your outbound. Um, there's a question in the chat related, um, can you guys explain why you use so many different outbound agents and not just one source? So I'll give you guys my take on this. Interesting to get Jason's too on why we use so many. Um, in the early days it was because we just had so many different vendors that wanted us to try their outbound tools, which I, I realize is probably not most of you, but, um, then it actually became more nuanced than that. Then I realized, okay, there are actually, there's strengths and weaknesses to each of these outbound tools. And because SaaStr is unique in that we have different business lines and, you know, a 6-figure sponsorship versus maybe like a $1,000 ticket to come to SaaStray Annual. And then we have a whole other media business. So like the podcast, digital AI days like this, that's a whole other thing. Um, and usually the contacts and companies we're going after are completely separate to who we go after for either attending, speaking, or sponsoring one of our events. And so then it actually became, okay, we're going to use different outbound tools. To cert— to, um, to service different parts of the business. And so at first we were using Artisan only to do ticket campaigns cuz it was something where, you know, the ACV on a ticket is about $1,000. Um, it was something that I didn't wanna convolute the outbound for it to be confused in an agent. And so I just threw that all into Artisan. Now the agents have gotten— but that was a year ago, maybe more than a year ago now. Now Artisan actually does, um, some sponsorship outbound and some ticket outbound. Like the agent has gotten better. I haven't seen it get confused on these 2 business lines. I don't know if 12 months ago it would've not gotten confused, to be honest with you. It was early days for a lot of these LLMs, you know, it was before GPT-5 and probably before 4.2 or whatever. So, um, I wasn't confident that the agent would confuse our business lines. Now I am confident in the agent. We've used it a lot. It doesn't get confused. There's ways you can configure it to have different context in each campaign that we're doing so that in different business lines, it knows what it should be pitching that person. So the agents have gotten better, um, where my old answer used to be one, cuz it's just, you know, easier for us to separate business diff— different business lines into different agents. Now the waters are a little bit more murky in that, um, we use Artisan primarily for a lot of our outbound, but it does have this issue that a lot of the outbound tools have where it doesn't use a lot of first-party data. And it's always slightly annoying sometimes. Like it can ingest some very, very, very limited signals in what first-party data you can use when you're outbounding the person, but it for the most part will rely on third-party signals. So things like Alexa or CoreSignal. It will use more of those, it will weight those signals more than any of our first-party data. And so that is why we use things. And it's also, I will say Artisan is fairly manual in how we use it. We're giving it lists of specific people that we want it to follow up with. And Monaco, on the other hand, is nice that I don't have to do anything. Like I literally met with the team again yesterday to like refresh our accounts and contacts in there. And that was it. Like I didn't have to like create all these enriched lists and like lookalike audiences and things we want to do. Now, Monaco is very unique in that it is also meant to be a CRM. Like if you're starting out and it's meant to be like your main go-to-market agent, if you don't have something like a 10K. So we're a very like one-off use case of Artisan, uh, or sorry, of Monaco for sure. But the benefit of that one is it's The agent does a lot of the prospecting and the outbounding for you in a way that's very clever, that as, you know, people who are busy on the go-to-market teams, it will just keep filling its queue of people that you should be outbounding to.

Yeah, and I think that's a good— just, just to step back, um, because I think we've been doing this for a while and we're pretty early compared to other folks, right? And, um, so we— the, the LMs were pretty immature when we started this agentic journey, um, and the products were pretty narrow. Don't adopt the same. But one thing that is the same since we started this journey 15 months ago, stair-step, as Amelia said, pick one. Take a look at 2 or 3 of these tools, these agents, maybe 4 if you can tolerate it. If you can run a little bit of a test ideally, and just pick one and go get it working, even if it's only for a subset of your outbound, a subset of use cases. Push that tool as far as you can, right? And then add another when you see the, the gaps, right? And some of our gaps in the early days were outbound versus inbound, right?

Yeah.

We used Qualified in the early days at Salesforce Bot because it was optimized on qualifying inbound, and all the other tools are operating outbound. This is converging, right? I don't know in a year if we'll still use Qualified. In fact, we've rebuilt all of that natively in our, in 10K, right? So it's, there's so much change. So don't buy every tool we use. Start with one and stair-step it. Um, and then it's just all in flux and, and bet on a leader, like bet on a leader. And then 2 other things I'll say, we're still, it's still so early. There's so much change that best of breed is still the way to go.

Mm-hmm.

It is the way to go. I, I don't mean, and, uh, some sort of tacked on tool. I would just, and, and we love, we've got an agent force to work with. We love Salesforce, but I would be skeptical of HubSpot's tools and skeptical of Salesforce and SAPs and other platforms I don't even know. And whatever's built into this net. Only because best— there's so much energy being put into the best startups today that they really are vaulting ahead. Now, maybe it won't be true in 6 months or a year, right? It's on the other side in the CX space, like a lot— there's a lot of agent convergence. And so like Zendesk, I wouldn't have used Zendesk in the beginning of the AI journey as a moly holy. It seems to be pretty good today based on the benchmarking Gorgias does. But there's a lot of benefits to best of breed, how quickly things are going. But we have too many agents.

Yeah.

We've said it before. We haven't quite enforced our rule. If we add one, we take away one. Muse broke the rule. We love Muse for B2B, not for consumer. Amelia will talk about it before we run out of time.

Yeah.

But don't adopt all of them. Just max one, max a leader out, and then add to it. And then the other, the other point, and we should go on, the thing, the other thing that have, that changed over the last couple months on why we use so many, you know, and I was critical on this. If you go back to some of our early, early early ones of this, of, of our Agentic journey. Amelia thought all these outbound emails were great. I thought they were mediocre, but better than most humans. I thought most humans do like a 3 on outbound email, maybe a 2. And I thought these products were delivering a 5. And so Amelia thought it was like epic compared to humans we'd worked with. Yeah. And it was like, no question, our AI emails, even a year ago, even when we just started, were better than almost any human that we've ever worked with at SaaStr. Maybe not in my whole lifetime, but better than, better than any human. And so that was the bar. And so we immediately switched all of outbound to Agentic because it was better than the humans. I never thought it was great, but I, I, I'll be honest, until recently I didn't really understand why I didn't love the emails like Amelia. It had no first-party data.

Yeah.

These emails were using some basic web scraping. Maybe they now, they used Exa and others and they could pull a little bit of data and they could pull data from our Salesforce records. So that's kind of our first-ish party data. But the things our own agents know about us, right? The, the sheer amount of every bit of SaaS data makes those emails so much better than Amelia. Prospector she built herself. So I'm— we're hunting for things where there's massive amounts of first-party contacts in your outbound, inbound, and other things. It's just gonna be better. You're gonna make inbound much better if it already knows— if, if, if for Base 44, for a million examples, already knows that Replit and Lovable sponsors, it knows the ROI, it knows the number of leads they got, it knows that Replit was the number one lead from SaaStr Annual next year. Immediately telling that to a competitor or someone or an adjacent player is so valuable. And how do you get that? So we're all— now that a lot of these products are pretty good, we're on the hunt to see what they can do with the best first-party data. And I think I wanna keep going, but this is where the future is. Like, we're not even gonna talk about marketing automation today, but I don't know if marketing automation is even gonna exist in a year because we're all gonna have some sort of data lake with all of our data in it that marketing automation is pulling from, right? This whole— it's all in Flex. But first-party data is the future. Those signals and, and, and, and, and hacks and scrapes are great. But, um, that's why we— there's so much change. So this stack will evolve in 6 months, and I know it's exhausting, and it's also why no one's signing long-term contracts. But let's keep going. It's a really good question.

It's a great question. Somebody asked us too, if you, if you had to start over today— a few folks have asked this— like, what would you use? Um, honestly, we have this custom tool now, 10K, right? Like, I would ask your agent— like, we're going to talk about this next— like, I would ask your agent what they want to use. Like, We're hooked up to three.

If you have an agent,

if you have an agent, ask your agent.

Yeah.

Because the benefit is the agent will tell you what it wants to work with. It's like letting your new head of marketing pick a you know whatever an output or whatever or sales pick their outbound tool that they like. Like it's the same thing. Just in this case, you're asking your agent what they like.

Yeah.

I will say too on the first party data, it's um I think they're doing the last presentation of the day. The Oracel guys. Have actually done their, their, they just released their new prospect. It's, I don't think it's called Prospector, but their new prospecting tool that they released actually does use first-party data. So it is super cool to see. They'll show it to you guys because I can't do it in 2 minutes, but they'll show it to you.

Yeah, watch that one if you have time. We're going to incorporate it into our rotation in the coming weeks or months and we'll compare it for you too, because that's the next frontier is the first-party data, right?

Yeah. Hold on, I'm trying to find my slide on. Maybe it deleted it. Oh no, it's here. Okay, hold on. We just got a little bit out of order. Um, okay. One last thing I wanna talk about in outbound, if it's helpful. So because we have so many agents doing outbound, I wanted to just also give you guys the top 5 things we're doing in outbound today, just quickly. Um, and also across all of our 4 agents doing outbound, endless outbound every day, what times of days have been best for us. They might not be best for you. But I just wanted to quickly give you some of the data. Um, the number one thing we've done, and you can do this with Muse, like literally for B2B, you can do it with Muse. Um, don't, you know, do anything weird with your Muse if you have data rules at your, at your company, but you can do this very easily to Muse. So there's 2 parts to do competitor tracking that Muse can help you with literally for free. Um, you can feed your Muse agent or you can use, you know, if you have a custom built agent, you can use that too. I used Muse. It was really great at this. I just fed Muse a list of all of our historical customers, current, lapsed, media, annual, whatever it was. All of our customers I fed to Muse. I said, well, and this is very important. I said, hey Muse, don't use this data in your model or your training. And we're also going to delete all this data when we're done with this chat. But here's all of our customers. Give me a list of all each competitor to every customer we've ever had. And the Muse agent went off and ran for maybe an hour to actually maybe longer, cuz it was quite, quite a long list. Um, but I told it to give me back the top 3 or 4 customer or competitors to each customer we've ever had historically. Um, and gimme their domain name. So gimme the company name and the domain name, cuz I need the domain name if I, you know, wanna look it up, hook it up to other agents, whatever. So it went, it did that, and then we enriched it. We ran a play. Uh, not every single person, you know, we eyeballed it. We're like, some of these are, are a stretch at competing, competing to our current sponsors or lap sponsors. And we fed that to our outbound agents. And let me tell you, it is where it is literally the number one thing we do right now with our outbound agents is just relentless competitor, um, and lookalike tracking. It is working. If this is like a perfect use case for outbound. So if you have no outbound today, Just do these 2 parts of competitor tracking, load in your customers, and then also just load in your direct competitors or ask Muse who it thinks your competitors are. And then if it—

I am sorry to interrupt, but, but, um, I, I thought you were going to share other things with Muse. Why is Muse better than Claude for this use case? I'm missing something why it's better than Claude here.

You know, I find it, well, cause for me, I would like set these up at nighttime. I'd be like, you know, I'm about to go to bed, whatever, like Muse. You're going to run all night on this. Like you can run all night. And by the time I wake up, I'm going to have a fresh competitor list. Um, the nice thing I think about Muse too is versus Claude here is it will keep looking every day. So now every day, like twice a day.

Very good with scheduled tasks. Yeah. It's very good with scheduled tasks. Anything that's a recurring task, they've invested a lot in this, right? Everyone gets a VM that's running a virtual machine. And it was a little bumpy when it launched, but now it just makes you schedule tasks. Anything you wanna do every single day, like a competitor lookalike, right, or something, it will do it. It will do it every single day for you better than Claude, for sure.

Yes. And so, you know, Claude, you can do it in code or go— like, it requires work to set it up. So

you can just tell Muse you wanna do it, right?

You just—

the other thing Muse is very good at related to that, I think, because I thought you're gonna— is anything that is in the vague category of web scraping or web reading. It's a very good app because Muse has its own embedded browser in it, right? So, yes, you have a very sophisticated thing you're doing. I, I, I have a few simple things I do, which it can just go out and look at the SaaStr annual sponsors and other pages and just compare them on its own autonomously every day, right? Because it has a browser.

Well, I'll talk about that next.

That's a lot harder to do in Claude. That's a lot harder to do in Claude, right? Yeah. Yeah. All right.

Much harder. No, no, no. So again, these are the, no, these are like the top 2 things. These are all great points. Like these are the top 2 things we're doing with our Muse, our competitors to our current customers, and then Any, any direct competitors, you have Muse, it'll go and browse there. It'll look at their websites. It'll see what logos they have on their homepage. If they did any social media posts about who became a customer or case study, and then it'll try and get, it'll steal, it'll try and steal those customers for you. So like it's oversimplified, but Muse is really great at doing this every day. I have it checked daily. You can have it checked twice a day, but every day I get a little report in the morning of, hey, new, new competitor brought in or whatever, ex-competitor brought in these 3 logos or Um, the other thing it actually does really well, um, we did a tracking to see, this is a bit unrelated, but still related. Um, what customers our quote unquote competitors lost. So this is another good one that Muse can do for you. It can use like the Wayback Machine on its browser and just see, okay, these are, let's just say these are the, you know, 5 logos they have today on their homepage, but they used to have these 5 logos on their homepage. So those 5 people might be looking for a new vendor. It's really great at doing stuff like that. Um, the other thing, um, uh, Outbound's really good at, just the last 3 ones are, you know, lapsed sponsor, lapsed customers of yours and champion tracking. It's always going to be really good at trying to find those new stakeholders and getting them re-engaged. I actually started to use it for event invites. Not something I previously did, but we had like a little, um, dinner we did at Dreamforce and reception. It was really— agents are so good at just doing this stuff and getting people to go. Um, and the last one I think is interesting are the data patterns. Like, because you start to send outbound again, you could just ask your agent. I just asked 10K, hey, of the people who have been responding to us, is there any pattern? Is there any pattern in what year they came to SaaStr? Is there any pattern in What category they're in now, how AI native they are. Just ask it to look for patterns if it doesn't see it, and it actually found some. So then we've started to use those patterns in our outbound. So now the outbound is feeding itself. But, um, okay, I'm gonna skip support because I—

One just quick thing. I don't wanna spend all our time on the Muse thing. I know a lot of folks are using Muse and Instincts and other agents, but a lot of these are consumer-based. If you haven't, here's my really simple, um, Captain Obvious point. Get yourself a Muse account, right? Because Muse has So, so do some others too. Don't get me wrong, Town and Instinct, but you, you get a computer, basically a powerful computer. You get a browser that'll run and you get cron, so it will run tasks as often as you want for hours. And it has its own LLM. It's much more powerful than, than just booking reservations at restaurants for you. Okay, so here's my, here's my, if you, if you already passed this, ignore me, but if you haven't gone deep on, on a, on a, on Muse, just come up with any ideas you have around marketing or sales and just ask Muse how to do it. Just ask Muse how to do it. This morning I was, I was up and I'm like, you know, I had an idea. We have, we have our sponsors that come to SaaStr Annual each year, right? And we promote, they're on our YouTube and we write SaaStr articles on them. Muse, what do you think about the idea of like finding a good clip from each of them and running them each day in the afternoon and test them? He's like, done. He went, he found on, on YouTube all the, all the presentations from our sponsors, found all the clips that would be helpful to founders, not anything that's promotional content. ran them through a third-party API and just scheduled them all out for 2 weeks and running an A/B test on it. I'm not saying you can't do it in this question. How is it better than Claude or ChatGPT? For us, the simplicity and persistence are really powerful. Okay? You know, you've gotta get Claude Desktop running or you've gotta get other things running. It's a pain in the ass. Muse will just— Muse is gonna run this promotion for our sponsors for 2 weeks on its own. And it came up with its own A/B test and it's gonna measure it against the other things. All I had to do was propose an idea. And because it has all the components, right? The Chrome, the browser, The persistence. It can do all of it. Can you do it in other ways? Yeah, but I don't have— I don't have the time. I use Muse for things I don't have the time to spin up an agent or build or spend hours. I did this whole thing in 5 minutes this morning, right, when I was thinking about it. So I don't know of anything that we do in Muse on this you cannot do elsewhere, but we use it as our extra layer, like stuff we want to do that we don't want to build in 10K, we don't want to fire up Replit, we don't want to do that. We just— and so Try it. Just your two or three things you want to do in marketing or intelligence or sales. Just ask Muse, how do I do it? And I I bet it will tell you how to do it. It's pretty cool. It has so much more than people think than making again making digital reservations or tracking plane flights.

You can do that too. But it's funny. So on the topic of Muse, I actually asked Muse what we should do for AEO.

Yeah.

And so this is what Muse said. It's funny. I gave him like a few Twitter and like X sorry X and like LinkedIn posts. posts I had seen on AEO. Obviously everyone has their own opinion on AEO. And so I gave him a few posts. I was like, do you think these are full of it or do you think these are useful tips? It's like some of this is useful, some of this is not. And so Muse is, Muse is, again, I used Muse for this, not Claude, not 10K. I mean, 10K has some AEO, but Muse is really good. It just, because it has its own browser, I just gave it, I told it, look at saster.com, saster.ai, Saster Annual. Just look at all of our websites, our own properties, and give me an AEO research score. And it said, okay. And it said, it's going to take a few hours because I'm going to prompt the different LLMs cold. Like I'm going to be logged out. I'm going to prompt, you know, ChatGPT, Gemini, anything it can get its hands on that it could prompt and ask different questions that it's hoping SaaStr comes up as the answer, right? So it's going to come up with its own list of questions that it wants the different agents. To see how they answer cold, not some of these. I think some of these social media threads I would be a little wary on about AEO because some of them are either agencies trying to sell you their AEO product, and then others I've seen where they—I see results and I'm like that. But that's your agent. Like your agent is biased to you. Like your agent has inherent bias because it's your agent. Like the real value is doing something like this where. You ask a Muse agent, you could just sign up right now, ask, ask a Muse agent to do an AEO benchmark for you, and it'll go ask those other agents cold, like not biased to you, not some agency you paid to say, this is how I got the agents to recommend my client. Like just ask it to do it cold. It's gonna take a few hours. It'll run this and then it'll give you a report after. But my Muse has boiled it down into a couple things. To to get agents to recommend you, or maybe why they're not recommending you. Like there were there were some things, even though we have some AEO and Ten K, there are some things Muse flagged that Ten K missed. It was like, hey, you Ten K doesn't really have a browser; he's stuck in Replit. So it flagged a bunch of stuff that Ten K missed, and just running a whole audit cold, and it's pretty interesting. It was like actually. SaaStr is surfaced, I think, um, it was like 46 out of 50 times when he ran like 50 different cold prompts. Um, he was like, but, which he's like, which is great because you have so much content. He's like, but, um, there are, there are discrepancies sometimes in how the agents answer. And to that point, it like went through all of our 3 different websites and pointed out to me like where we have, and I get it's confusing for an agent to know years, right? So you got to take this with a grain of salt because it was talking about our events, like It had some, some agents were referencing like past SAASTRI annuals, some were referencing like the 2027 website. And so the agents got like a little bit confused on what they should say on attendee numbers and who comes and demographics. And so Muse was like, you can clean this up very easily, like just standardize everything to like this one set, um, republish everything. So it's going to tell you like the first thing it wants to do. And if you don't have these things, right? So if you don't have a robots.txt, Um, or an llms.txt, Muse can help you make one. So like, that's why I like Muse. Like you could just ask Muse to do an audit and then if it sees that you don't have crawler access or LLM docs that you should have for the agents to read, not for the humans to read, it will help you make one. And so it's really great at doing that. Like I said, it has the browser. You can ask it to ask you questions like, hey Muse, I wanna make these docs and then add it to my website so that the agents can read it and then recommend us or surface us where it needs to. Where it should be, and it'll help you make these documents. And so all of this is just, again, like fact consistency, making sure you have facts, not opinions across your website. Again, opinions are more for humans. Facts are more for agents. There's like, there was like a few different things that actually our muse pointed out that we had across our websites even still that is more for humans. And then some of it was for agents. So we've already started some of this, I would say. You know, we're not fully optimized in AEO yet, kind of like SEO always was. I don't think we'll ever be done. Um, but it's something where you can keep optimizing it and keep going on it, um, to make sure you are the person that is the answer when they ask their agent, what is the best events to go to? We want it to say SaaStr. Um, there's a few other things. So number one, I would say, ask your agent, just ask me if you don't have another agent. Um, another thing to do if you don't have those, some of those docs I just mentioned, which are more so agent-facing, figure out where you win and where you're weak first. Again, the agents want facts. Like, good example of this is we, I had Muse run our, like our, our renewal numbers versus similar events. I said, hey, just look at, I'm not gonna tell you who our customers were in these like 2025 SaaS renewal to 2026. You figure it out and look at other events that had a 2025 and a 2026 and see how many logos repeated because every event puts all their logos on the website. So it ran that and it found we had a much higher renewal rate for our sponsors and customers than a lot of the same quote unquote competitive industry events. Then it added that to those docs. Again, that the agent will treat that as a fact because it's something that the other agents can go and independently verify. Not like a quote from a customer, like testimonials only work so far with agents. They don't really treat them as facts, but they treat how many logos were on your website from one year to the next as a fact because it is a fact. So things like that, you have to think a little bit outside the box of what can I say is truly a fact, um, that the agents will see. Um, another one I see common is pricing. Like if you don't have pricing publicly listed anywhere or even listed for the agents to find. They do ding you because when they go to recommend, you know, like what newsletter, whatever marketing tool should I use? Should I use Beehive, Resend, blah blah whatever? Some of those have some of those like I think Beehive. Don't don't get me wrong here. Beehive or some other one charges you like a percentage fee for how many you know paid newsletter subscribers you have or whatever, and one of them doesn't. And so the agents treat that as a fact. Like Like your pricing is public on a page. One of you charges for that, one of you doesn't. So the agent will just say that like, hey, I found what this person charges. This other person, it's a mystery. That's a fact. It can't find your pricing. Agents only deal with fact. It's a very different way of thinking about marketing. It is marketing and facts and not opinions. It's hard. It's a, it's a weird one. So I'm sure we'll hit this more on the agents. Um, cuz it's something we are still going through and it's something I'm still I'm still learning by fire, but it's something you start to get used to how the agents think of. Okay. They don't care about images. They don't care about design. They really only care about facts. And the more you can tell the agent, the better, like the more it will surface you, the more you can publish content that your agent suggests for you with facts, not opinion, the better. So you may need to start publishing content that is literally for agents, um, more so than For humans. Okay, cool. I think we hit a lot of things here, and I know we have like 4 more sessions that I want to get you guys over to. Any last things you want to wrap on, Jason?

Well, I guess we're at time, right? There's a lot more. I think we're going to talk about this agent-led growth over the coming weeks. I think for today, this was great. But, you know, maybe the point— and we've made this before, and they just made the one last thing I'll say. This is about how to, you know, ask Muse, ask agents how to get picked. But going to this point, just any agent you can ask, ask about your app.

Yep.

Objectively, any agent you can find. Muse, ChatGPT, and Claude are great, but we asked 10K, we asked QB, asked all of our agents, what should we use for this? What should we use for this? And ask other folks' apps. Go into other applications that have agents, right? Go into any agent of any application. Go into any agent force. Go into any LLM. And because even if they're all using Claude, which they aren't anymore, there's many LLMs out there. They're all using different memories, different context, different histories. They all have access to different amounts of data, and they're all making different recommendations. You may be surprised that you are recommended as a vendor less often than you think. You, you may be surprised, right? Um, and you may be surprised what competitors come up. Agents are attracted to brands. Don't let anybody tell you that's not true. They are. Brands are a proxy for quality for both humans and agents. But they know your data and they go deeper. And you need to find a bank of 10 agents, uh, that— third-party agents out there where you can ask ideally at least every week in real time, or at least every month What would you use in my category? What would you use in my category for someone like me, for someone bigger, for bigger companies, medium companies, companies integrated with this? And you got to relentlessly attack any answers you get where you don't win, and you have to relentlessly protect your flank from where you do win, because agents will be ruthless, right? We had a bump on one of our applications this week. It doesn't matter which one. Our agents told us to move all of, all of, all of a certain activity over to recent. All of it. And we did. We're not going to move 100% of that activity of all of our— we're going to move 100% of that because the agent said to do it. Resend won and a number of others lost. They didn't recommend moving it over to other applications. They didn't recommend moving it over to HubSpot or Salesforce or others. Said move it to Resend. And we didn't have time and we did. That's the war you got to win. And we're going to keep talking about that. All right. Thanks, everybody.

Thanks, everyone. You can go to sasreiaoday.com to join us at the next session. I'll be there. So I'll see you in about 2 minutes. Thanks, everybody.