SaaStr's 10+ Top AI Agents: AI SDR, AI BDR, AI RevOps + More: The How, The Who, The Why

Looks fine on mine.

All right, cool. Letting everyone on. Yep. Hello everyone. Oh, I'm gonna go back, go back one slide.

Are we live?

We're live. Welcome to Workshop Wednesday.

Welcome everyone to Workshop Wednesday, or SaaStr AI Live on Wednesday. Yeah, as it's been rebranded. I hope you're rebranding yourself. Um, are Are we, are we good to go, Amelia?

We're good to go. We're streaming. We're here for those who— we're gonna make this one slightly more interactive on when we go through, like, our AI agents. If you have any questions, drop them into the chat as we go. And there's plenty of AI agents and note takers here. Okay.

Okay. So welcome, everybody. What we're gonna do, we'll go through— Amelia will drive the slides. But we're gonna do something that a lot of folks have asked for, which is a bit of a deep dive on what we're doing, how we're using AI at Saastr. We've chatted a lot about it. Um, we've got, as, as Amelia will show you, we've got about 11 core agents we use and about 20 total. But we're going to do a deep dive into like the 6 that we use the most and the 6 that are probably most interesting to you guys. Core AI agents that we use every day, how they work, and show you, show you the actual data. Um, I can give you a little bit of an overview, but then Amelia can, can dive in on, on most of them. Um, and then I'll help Most of them we've bought and we'll show you who we've bought or what commercial apps we use, which is our recommendation. And then we've actually built some ourselves on Replit and we can dive into that. But bring your questions, especially there should be a lot of questions on AASDRs, AIBDRs. The biggest takeaway I can tell you, I don't know if we put it in this deck, the biggest takeaway is there's a lot of vendors there, but on two of the ones we use, Artisan for outbound and Qualified for inbound, of all their customers— this is the most important thing— we're the number one vendor, number one results. Amelia will show you that. And so ask questions about why, because there's a lot of great tools out there. We'll show you what we use, but the most important thing is going really, really deep with AI agents. There's no set and forget, no matter what anybody tells you. In 2025, maybe in 2026, uh, but there's, there's no set and forget. Um, and, uh, just for like us, um, Maybe Amelia, if we skip to slide 3 just for fun.

Sure.

This is our little journey and hopefully you'll be on the same journey, just time shifted. Amelia, tell me if, if we got this wrong. I just made this in Claude in 60 seconds before we started. But we came into the beginning of this year basically using no AI at Saastr other than ChatGPT and Claude. I could be wrong, using nothing. And the first thing we adopted, which some of you guys have seen, is we use an app called Delphi to do an AI agent on Saastr. We've done over 100,000 conversations. Uh, at first we, we just used that, right, going into SaaStr Annual this year, and we used it for everything. We used it for a little bit of sales, a little bit of marketing, a little bit of support. It's really just designed, and we'll, we'll go over it, to be a digital mind, a digital version of me. Um, but what we quickly, instantly saw was just having it was so much better than nothing. It was so much better than nothing. Um, a general well-trained tool is better than nothing. We then we started adding marketing tools. We added Gamma and Higgsfield. We'll talk about and then we added our GTM tools like Artisan and Qualified and Momentum and others and we'll dig into it. And ultimately the journey we're on, we've gone from zero agents at the end of the year to about 20 today and 9 core ones. This should be you. And if you haven't started, it's okay, but it won't be okay soon. But you can do it. And as you see, our adoption has just skyrocketed. And frankly, as we've lost agencies or resources on our tiny internal team, we've just— we've done what everyone on the internet says, which is replace them with AI. And it's not without challenges. It's a higher load on me and Amelia because we have to manage these agents. We don't have— I mean, Amelia's title has changed to SVP and GM to Chief AI Officer. It's not a joke because that's probably what she spends 30% of her time on is managing these agents. It's not free from a soft cost perspective, but it's radically different. So we'll go through we'll go through what we've done. Um, and maybe, let's see, Amelia, should we— we can go through the next ones just briefly and then dig in. Here's just one for fun for guys that we did. I just did the one of our timeline, nothing to $20K in 9 months. This is, I mean, this is a, a, a night and day. I— we, we, we have a SaaS University where I've built a bunch of courses. They're free if you wanna try them, sasuniversity.com. They're still good. I'm redoing them. in the AI age, but I went to ask a question about it, reply in a day. Like, and I went around a, a lot of websites and a lot of them are terrible in terms of AI interaction. And again, on the right is our Delphi agent, the first one. It's not specialized. It is not specialized for sales. We've taken it off for sales and tickets and sponsors that we'll show you. We've, we've put it in the buckets and the parts of SaaStr that are general. But man, waiting a day to get my question versus here, I just asked him 5 minutes ago, when is Saastr Annual? Who are the sponsors? It just tells you everything. There's no excuse to not have extremely high-quality customer interactions instantly through AI in 2025. There's just no excuse. And we'll go through our— we'll go through our stack there. So just for fun, if this is anywhere in your product, just company, just like end it this week. Just, just start. Um, and maybe the next one, and then I'll let Amelia drive. Um, we do have, depending on you count, about 20 agents. We'll go through it. Um, you know, our organization is tiny, but, you know, SaaSer does do 8 figures in revenue with a tiny team. So it's interesting to track us because we have a single-digit amount of headcount doing 8 figures of revenue. That's fairly extreme. It's not unique to us, but it means we really push the model here. But pretty much every single thing you can do with an agent, we are doing. Roughly, you could, you could kind of see it on the slide before. Roughly, we try to try something new about every 2 weeks. Would you say that's right, Amelia?

Yep.

Yeah, we're talking about we try to come up with a new agent idea each week and try one every 2 weeks. And then 1.5 per month is about all our org can tolerate because it's just too much process change even today. Right. But that's— this is kind of the stack we've built up. And then in brief on the next slide, this is what we'll go through and then I'll just let Amelia drive it. We do have a lot of AI apps, but a lot of folks ask what our stack is and this may be different in 90 days. It probably will be because Amelia did a version of this 45 days ago or something. Our AI— In July.

Yeah.

July. Our agent had already done 8,000 customer interactions. The stack's already somewhat different and it may be different later, but We'll go through it. We use an app called Artisan. We actually use 3 instances of it for AI-STR. Amelia will show you. We use— we just rolled out Qualified as BDR on our website to manage inbound, and it's already had a pretty strong impact, right? We'll show you. We've used Gamma almost since inception to automate sales collateral so that it's all dynamic, so it's not the same old friggin' deck that gets out of date and is embarrassing. Um, we love it. Um, we talked about Delphi as a general tool. We've added AI RevOps. I've blogged and talked about it. We've added both Momentum and Attention. They're both great. Amelia will show you why. And then we've built some stuff ourselves and, you know, we'll briefly touch on it. We could do a different session. If you've followed me on Twitter, probably seen a lot about our journey with Replit, but notwithstanding some bumps in the beginning, it's been pretty successful and we've rolled out about 5 apps ourselves. The main learning is only do this if you can't buy it. From 99% of you. Like, it's cool, but we did this for content review and other things, and we'll show you where there is no off-the-shelf app. But don't roll your own. Buy, don't build, right? Use for B2B stuff. There's a lot of cool things you can vibe with Replit and Lovable and others. Try it. We can do a whole nother session on it in the coming weeks. But if you can buy something that's bulletproof, don't build it. Don't believe the hype. But most important, none of these names are as important as putting in the reps and the work And Amelia will show you the actual data in a second. These are all great apps. We stand behind them. We recommend them for sure. But you could pick another two. The most important thing is training them deeply and doing it every day. Every day. So many of the AI failures from, you know, late 2024 into early '25 that you'll see all over social media are just folks that bought something and never trained it. Or they bought something and hired a mediocre agency that did nothing. You've gotta train these. You've got, and you've gotta do it every day. And not only does it take weeks, and Amelia, when you, when you touch on some of these, you can talk about our training. We're pretty good at it. 'Cause we're tiny. Everything is weeks of training, ingestion and training. And then every single day we're in these apps monitoring the outputs, monitoring the interactions, monitoring the discussions and fixing things that are as often as wrong. And if you don't commit to that, you're gonna have extremely mediocre outputs. Just like any, frankly, just like any other tool you buy. If you don't put in the effort, you're not going to get much out of it. So, so many of the failures for AI agents, if you really peel back the layer of the onion, just someone on the team did it and there was no proper training, or there was no data to ingest, a combination of two. So that's far more important than which vendor to use. But we'll show you, we'll show you what we use. And, and maybe with that, Amelia, you could, you could transition into SDR, BDR.

Sure. I am really quickly putting the link to this full deck in the chat so folks have it. Cool. So you can follow along there. There's also more slides in here than we can cover in our time that we have. So you'll see some of the— I've seen some questions come into the chat already that we do answer later in the stack. So if we don't get to it, some of it's there. But with that, I feel you've touched on this one, Jason M. Lemkin. Anything else on getting there by Halloween?

I think we got it. I think we got it. Get it there by Halloween.

Halloween, it's a good deadline. Okay. The first one we'll go through is our AI SDR. So for all his purposes, what it does is we use it as warm outbound. We did a full deep dive on how to do AI specifically and use agents for sales, like Jason said, in July. That one is still very relevant though. So if you haven't watched that full one, that is a full-on deep dive. So we'll touch on a few, like, most recent learnings just because now we've had it fully deployed. For, you know, 100 days or so. So how we've used it in, in our use case is we use it as a quote unquote warm outbound agent. So I have trained it on 3 different use cases, which you'll see here, which I'll walk you guys through to do outbound in different ways. So the way that this is helpful or why you might think about using an ARSR agent with some of our data is because it can do hyper-personalized emails at scale. So in the last 100 days, It's done, you know, almost, it's probably 15,000 as of this morning, right? So it's done about 15,000 messages. I've put a screenshot on the right of one that it's sent out to a prospect here to buy tickets. So this is from our direct ticket sales agent specifically. And you can see it's pretty good, right? It's talking about what they do. And this is part of the training we did with the AI agent is a lot of how we set it up initially. Between ourselves and Artisan was to say we want every email to try and add value to whatever audience it is. So if it's our sponsorship agent, if it's our direct ticket sales agent or a VIP, or a VIP SDR agent, we wanna always train it and tune it and say, you know, we have a lot of really great data at Saastr. Again, our deep dive covers a lot of that, but we've ingested the most relevant data. It also looks at what they're doing on the website. It looks at what they're doing personally. And we've basically set the slider like all the way up on, you know, customize more on them versus on Zaster. So this is part of the hack of like part of why we're number one is yes, Zaster has a really strong brand. Other part of why we're number one is we try and make it as much about them and less about us as possible. When I've talked to other folks who have an AI SEO, they have similar use cases that they're seeing success. The more you can make it about them and adding value to them, the better results you're gonna see.

And Amelia, how do we, I should know this, for both sponsors and attendees, how do we pull in historical data? Do we pull it in from Salesforce and Marketo? And how does that integrate with outbound messaging?

Yeah, great question. So there's a couple ways you can do it. You can pull in from automatic signals. So you can do like a search of Hey, who's talking about SaaS or who's interacting with SaaS or your company? How we've done or how I've done it is I do a direct pull from either Marketo or Salesforce. I'll combine it manually. So on the backend, that does take some time. And then you can do a straight up upload to Artisan. So the most successful campaigns we've had are the ones that have been straight up uploads of our data. And then what it will do from there, it does take some time. So this is an interesting part too. Like, The first time you set up anything like an Artisan, so any AISR, they're all gonna tell you it needs 2 to 3 weeks to warm up. That's common. So keep that in mind. I don't know if we touched on that last time, but they all need time to warm up and basically they all need that time to warm up your mailboxes. So all of these are gonna use, you don't want it to use your primary email. You don't want it to use Amelia@SalesRank.com. You want it to use ameliatryzaster.com. So every AI SDR platform I've talked to needs this 2-week warmup. So that's part of it. Like it's gonna spend time just warming up those domains and mailboxes for you. Then if you do something like a CSV upload, as we've done numerous times, it will take another day or 2 to ingest all those contacts. So again, like if you're starting from scratch, this is already probably like week 3 in the journey. To do that because it's gonna take time to basically find those people and use all the data signals I've uploaded to make sure I— it identifies the right person. So it's gonna find their LinkedIn, it's gonna look at their email, it's gonna check the deliverability of their email, it's gonna look at their company signals. It's basically gonna look at a lot of signals that the AI can do at once to make sure it's the right person. And then it will start to generate the messages like you see here on this slide. So it'll start to generate those messages. It will put them into your pending outbox. You can read them. I've gone through them. Now it's pretty much on autopilot for us. You can read through the messages, make sure it's the right person, make sure it's the right thing you want to say, and then you can automate the send from there. Okay.

And how many— just two questions, one tactical, one on the data. So, so it queues them up before it sends in the beginning. What percent and how many messages did you read manually before you let the AI, uh, do it, uh, mostly automated?

I would say in the beginning I read all the messages. So it was, you know, hundreds and then thousands. Like the more people you start to ingest, like I wouldn't start with a huge list the first time you do this. I think I started with a list of like 600. That's a good mix 'cause it's not gonna find everyone. The match rate is fairly high, but it's not gonna find everyone. So by the time it does that and identifies the right people, yeah, I did about 600. And then in the beginning I would read 'cause you can set the agent to say how many messages do I wanna do? So for us historically, we always knew like if we wanted to email somebody to come to Saastr, you know, an email from me or a human SDR would take about 3 actual human touches. And so we duplicated that in Artisan and said, okay, 3 touches in email and then 1 LinkedIn message to the person. And so your outbox would then have 4 messages per person, right? So 600 times 4, you're looking at 1,200 messages. I would read all of them in the early days. Now I just spot check. I just go into—

but you, in the beginning, you probably read the first 1,000 emails yourself, but now every day you're still spot checking for 20 or 30 minutes per day, probably, right?

Exactly. Yeah. I spot check 20, 30 minutes today of what is going out. And then I also spend probably more time now on the replies of like, how are people replying to our artisan emails? Is this what we want? How do we wanna respond to them? That piece is a bit more manual. Now you can turn it on, so it's like automated in responses, but I still like the, the human responses. I feel like they're still a bit better.

Okay, one last question. I don't want to spend all of our time on Artisan, but this is helpful though. The, um, so for— I mean, we've had probably 50,000 people over a decade come do our SaaStr events in one form or another, and we have hundreds of lab sponsors, right? Or maybe 150. When you're— how do you— and that gets added to the text of the AI SDR. Are you— do you— are you manually creating XLS files and manually uploading to Artisan? And how do you— how does that whole workflow use for lapse or reactivation so that the email comes out great out of that process?

Yeah, it's a great question. So there's a few things it does. I will actually show you guys the backend for one second. I think it'll be helpful. Hold on. Let me get out of full screen here. I'll pull my Artisan guy over. One second. Because it's a great question, right? So let's see, can you guys see this tab now?

I can see it.

Maybe. Okay, great. Okay, great. Okay, so you can see here some of what I'm talking about, right? Like how many messages it's already sent in September, right? So again, I didn't read all 1,000. In the early days, I would have read all, you know, 1,000 whatever. Now I'm just spot checking. Honestly, mostly because sometimes we, because we don't have speakers for London, sometimes it will like pull speakers or speakers from annuals. I'm mostly checking that it didn't like hallucinate any speakers for London. But again, part of that is human error of like, I just haven't trained the AI yet on our current speakers for London because we haven't publicly announced them. So that's, that's a unique SaaS problem, I would say. It's got 207 messages in the queue. And then yeah, to your question here of like the inbox and the outbox, I spend most of my time now in the inbox. So this is where I can see who's engaged with us. I can see what it's sent. You can see if it's LinkedIn or email. You can see I've starred some of these. These are ones I wanna like go back to. And then you can go to the pending messages here. This will be all of our scheduled messages at the moment. So right now there's 200 in this. At any given time. But then to your question of how this works with, like, reaching out to, like, lapsed or churned customers and making sure—

So sorry, you've got 200 folks that have responded to our AI that have a question.

We've got more.

So it's a pretty high— just for folks, that's a pretty high response rate.

Yeah. I mean, I do think we have our response rate here. Total. It's a pretty high, right? So like this one's 7%. That one's high. 5%, you know, Jasper Larsen, who's the CEO, has said like, you know, average is like 2 to 4.

So, but if you do this well for, and the, the 7% is you're trying to get folks to come back to Saastr. That's probably who this batch is, right? Looking right.

Yeah.

That's that batch.

Yep.

But if the email's actually, you know, people saying that email's dead or stuff, if it's actually good, if it's, if it's, if it's adding value, helping them, 7.14% is pretty good for an AI.

Pretty good.

Pretty good, right?

Pretty good.

Yeah.

And again, we talked, we touched on it more last time, but this is still in addition to the marketing emails we do. So it's not that we've stopped the marketing emails that come from, you know, Marketo and are fully designed and are still inviting you back to SaaS as an alumni. We still do those too. It's just that you're getting those maybe 2 or 3 times a month. And then now you're also getting a sequence where you're getting a highly personalized invite from me. Also to London. So it does help reinforce the fact as well.

Yep.

Cool. And then yeah, to your point on like how do we make sure it's the right person, that's where I go into the inbox, right? And you can check the messages of what it's sending, making sure it's the right person. You can change it at any point if you're like, okay, I, I wanna say something specific to this person cuz I actually know them and I've worked with them a little bit more. You can customize it. And then the other thing you can do is you can upload a, I forget where it is, but you can upload a do not send list. So how we've used it is once we do have people who sign on. So if you, if you wanna drive new customers through this, once you get that new customer, so you can upload them to a list and say, stop emailing that entire company. We do the same thing with tickets. We say, okay, this company slash person has now bought tickets. You can stop emailing everybody there in the sequence so that you don't have to manually do that. Like, it'll do that for you. You just tell it who you want it to basically graylist.

But it can't today, and then maybe we can move on. It can't, it doesn't sound like today it auto-ingests our Salesforce or Marketo or other data in the same way that like Delphi auto-ingests our website, right?

Correct. Not today. It takes some training. So I'll go into one of our campaigns here. I'll pick a, just pick this attendee one. So this is our ticket agent, right? And this is the one that sells more directly because again, I've empowered the AI to do more of responses that I have for our sponsorships. This is pretty easy, right? If you say you wanna come back to SaaStr, the AI gives you a code. You don't really need me for that. Like, it's, it's fairly low overhead, low risk, right? The ACV on a ticket to London's like $300. So I've let it go a little bit more hog on this than others. But yeah, you can say, you know, here's the lead. You can see I did a CSV. You can see it's already done 400 out of 800. You know, don't contact people much like our, like a Marketo or a Salesforce. You know, for a sales loft, you can say if they're in other sequences or campaigns, you know, don't contact them. Like, we only wanna do one. The pitch is where you can tune it, right? So this is where I spend a lot of time saying, okay, this is the offering. Here are the pain points. This is where I've put in like the speakers and you can just keep adding to this. So it won't do it automatically today. I do know Jasper from Artisan says that's coming by end of year. But right now you just put it in yourself, you know, pick the, Pick the best pain points and proof points. I continuously update this, you know, use Saastr London and the subject line. People know Saastr, so it's good to put the brand into the beginning, break it up. You know, I've, I've put in different things here as we've kind of gone along. Be excited that this, like, you know, the AI doesn't really have emotions. You have to tell it to get excited. This is an event in London. For us, I always put like dates and locations because for events, you know, if you, again, if you don't specifically tell the AI, sometimes it can make it up generally. So that's where I put that there. So that's where I do a lot of the tuning now. Again, somewhat manually. It does crawl, like when you put in your website, it will generate some of this. I rewrote a lot of it just because it was better. But you can, you can tell it to do that. And then I know, yeah, by end of year, it'll automatically do a lot of what I'm doing now. Cool. I'm sorry, I'm reading the questions real quick before I go on to the next one. There's a question, why did you guys decide to use Artisan versus DIY solution like Clay? Couple reasons, Tracy, you might have some other reasons too. Like, I feel like on build versus buy, right? Like, I do consider— I know a lot of people use Clay. I literally just talked to the Clay team last week, and the co-founder is coming to SaaStr London. So, like, we're fans of Clay. I've used it for different use cases. Like, I haven't used it in this way for specifically an outbound AI SDR. I just find even the cost of setting up an artisan is worth it. Because they'll do a lot of it for you, right? Like they bought all the mailboxes that we have and our aliases. They like helped us set up the list that we did initially. They helped us set up the messaging. You know, they walked us through like the original cadence. Most people that are on Clay, I know, are like technical founders and they're just doing it themselves. They're using it like a Zapier, right? On Acid. They're doing a lot of workflows for themselves and not for their company. Today, just from the folks I've talked to, again, we're super fans. But for me, I needed something just a little bit more out of the box. Like, I just did not have like the patience to say, okay, I can— in theory, you can go set up your own clay agent to do this. But I was like, I just can't. Like, we have other things to do. So that was part of it. I think part of it too was like, you know, they, they, they were at, they were at the last ASRA annual. And so I got to know them a bit more. And part of it was like, I, I— we were actually just talking to Jasper, the founder of Artisan, last week. Part of it too was like, I trusted Jasper and his team with our data. So that's another thing too. Not that I don't trust a solution like a Clay, but part of this was like, okay, we have— you guys have seen it now— like, we have all of our past attendees and Artisan. We have all of our past sponsors. We've got current sponsors. I have VIP speakers in that now. Like, those are very sensitive emails I would like to not get hacked or go to the wind. And I just didn't trust myself to set that up in a clay, which is why I went through an out-of-the-box solution.

Yep. And the other thing, it's a good question. And the, the other, Amelia is the boss of this. She is our chief AI officer. I would say one thing is to just go, go into her. There's a lot of great vendors out there. We're not saying these are the best vendors. We are saying they work for us when they're trained. That's what we're saying. For what you'll find, I mean, there is no, no, we'll, if we'll have a, there are some apps that you could deploy on your own, right, with in a few minutes. And there are some apps if you have engineering resources, you could probably deploying your own. We don't, we don't have a lot of time at Saastr. We have a lot to do with the small team. We don't have any engineers, so we need tools and we have a lot of data.

Okay.

We have more data than almost, we have a decade of data. So there's like some sort of 3x3 matrix where we need tools that can support a lot of data that don't require engineers where the vendor will get us onboarded. And so to be honest, no one we've talked about yet, there are other vendors who wanted to work with Saastr and they fought with us or they wouldn't help us or they wouldn't you know, they said they couldn't ingest the amount of data we have or they didn't want to do with their app what we wanted to do. So there's just— there's a classic—

there's—

it's so vibrant right now in AI, right? There is no HubSpot or Salesforce that— one, you got to do diligence. It's tough. So first, try to figure out who's using what, which is probably why a lot of you here, but then use your spider sense when you talk to them. Do they work the way we want to work? Right? Like, for example, at Clay, there's a massive ecosystem of of agencies that help you deploy Clay, right?

Yeah, exactly.

Now for us, we don't need a $100,000 agency that's to help us. We just don't have the bandwidth to do it. But that might be a huge advantage over Artisan. Like, they're not direct competitors and everything, but like, I don't think Artisan has that network, right? So if you want to hire— this is how HubSpot dominated the world, is agencies, right? So did Shopify. So if you're super enterprise, you might need more than we have, right? But we needed something smart with a lot of data where the vendor would help us and One way or another, most of these complex tools need a forward-deployed engineer, a solution architect, or an agency. Okay.

Yep.

Once in a while, your own team can do this with engineering, but everybody, almost everyone needs help. And so you gotta decide, is this the type of help? Can this vendor provide you the type of help you need? And I would almost pick that vendor, the one that you have the most confidence you can onboard on, not the smooth, smoothest deck or sales process or best demo. Just talk. What's the onboarding process like? How does it match to how you work in your data? And if it doesn't feel right, pull the ripcord and bail and pick somebody else.

Yeah. Yeah. I do know they have self-serve coming right now. Like self-serve to get people started is coming for Artisan in particular. But I think the way they've done it so far is clever. Artisan, Qualified, Momentum, Attention, they've all been the same. Gamma now has all been the same way for us in that they were not unique in that they were like, oh, it's SaaS, or we have to help them get set up. They're like, no, we do this for everyone. Like, we want our AIs to be successful. We want it to work. You know, you want to see the responses and outputs that we've seen and, and enjoyed on the platform so far to be the use case across all companies, right? So again, to Jason M. Lemkin's point, I think talk to them. You know, we use Artisan. There's other vendors like them out there. See who you like, see who you trust. and, and go with that if you're going with an out-of-the-box solution.

The other thing I'll say, I want to stay on track, but good God, if you're in this space, the other thing we've learned is you, you can't have a crappy sales team because we need the help. And I can tell you there's at least 2 leading vendors that we said no to, not because the software is not good enough, but because the sales team fought us. They didn't know the product. They couldn't. Now, this is always true in the old days, but when you're talking about an app with training and deployment, a new category, and the rep fights you, No, like you, I want to know how I'm going to get this deployed. And so just if this is you, goddamn it, listen to the sales calls, get momentum or attention hooked up, be relentless and just get rid of anybody that is not an asset to your prospects because they'll just leave for AI. They'll just leave if they, if they answer, can't help you, can't answer the questions wrong or fight with you. This is not, this is not selling Zoom during a pandemic. It's not gonna work.

So, okay, let's move on. So that's for outbound. Let's talk a bit about inbound. I'll say, I'll preface this too. The reason we moved from support to, or I guess support is your Delphi in a way, to sales. I think that's a natural use case. A lot of folks we chat to say support and sales are their top two use cases for agents. I think the, data's out now on Salesforce Agent Forces, that's their top 2 use cases too across all their agents that they've deployed on Agent Forces for sales or for support. So these are common use cases, right? Like that's not unique to SaaS. I think where, where it became unique to SaaS is because Jason M. Lemkin's StealthFi has now become a bit of a copilot for folks. Like if you haven't tried it yet, the link is in this deck, which is in the chat. You can try it at sastrai.com, use the copilot. People use it for advice. Sometimes people use it for one-off questions, but when we looked at the data, most people come back to it. So they'll actually ask Jason's AI a question. Jason's AI will respond, like, let's say, hey, I'm thinking about hiring this candidate for a VP of sales. Here's like my high-level overview of them. What do you think? Jason's AI will say, you know, hey, I think, you know, you should think about these XYZ things. And then if the founder says to them in the next instance of Delphi, I actually, you know, I'm thinking about hiring that person that we talked about. What should their comp plan be? It'll come up with the comp plan with them. So again, it's, it's becoming more of a copilot, which is kind of cool. And in that use case, we really liked how folks use it for SaaS advice, right? SaaS-style advice. Jason doesn't have time to talk to 1,000 people a day, but his Delphi does. But where I thought it kind of struggled a little bit honestly was in the sales and support side and the inbound side. Like people would ask it going up to SaaS annual question about the annual, questions about sponsoring, and it did okay with the answers. Like it was still better to get an answer in real time than no answer, right? So it was still pretty good, but now it's just gotten a lot better that we use this mix of having a dedicated, right? Even though we have 3 on Artisan, like it just works better to have a dedicated agent working on outbound, and now we have a dedicated agent that works on inbound. So again, it's just that the level of output we've gotten to these is just a lot better than what we originally had on Delphi being Jason M. Lemkin and used for like a little bit of everything. It also kinda led us to these use cases, right? I just saw that people were starting to struggle with things like that. And so that's where we wanted to add those next tools and agents to the mix. So we use QualifiedNow. They've been in the SaaS ecosystem, I would say a long time. I know they've been attending a really long time, so we already knew of them, right? When I was starting to look for things that could do inbound for Saastr as an AI, this was top of mind because I, I also knew a lot of folks who were already using it. So this was also one of those things where I could just backchannel, like, what are people already seeing with Qualify? And they were all positive on it. So that led me to chat with their team. But what it does for us specifically, which actually is a little bit different. So without getting too much of the nuances for what it does for us is you can see it on our website now. So to Jason's point, if you're on saastr.com, you're gonna mostly interact with Delphi. If you're, however, on one of our SaaS for Events pages, so if you're on SaaS for London, SaaS for Annual, or you're on one of the sponsorship pages, media pages, basically anything that has to do with actual SaaS or revenue, this is where our qualified agent lives. This is where you'll see the Amelia AI. Again, we strategically said that Jason Delphi is Jason's AI. We named it after Jason so folks know that they could ask it for advice. We named this one the Amelia AI so that folks knew that there was a difference. It wasn't just like another Jason M. Lemkin AI. This is the one that you specifically go to when you have a question about attending SaaStr or sponsoring SaaStr, and then what you should do about those two things, or media, right? So that's how it's specifically trained. And honestly, it's something where even like, we don't have many sales folks here, but they were a little skeptic on in when I first deployed it because they're like, okay, our, our prior process to Qualified was you would come to SAS Re:Invent or SAS Relaunch and any of our event sites. And if you say were interested in a sponsorship, you would fill out a form. After SAS Re:Invent in May this year, I was like, that feels very antiquated. That feels not very 2025 where our conference is now all AI. We're going all in on AI. It feels very outdated to have a page where you fill out a form And that's the way to get ahold of Sasser. So that was like a kick in the butt for me specifically too, to get away from that kind of old model of how we were doing things. So now our qualified does actually a lot of different things, but part of what it does is it, the form is on the website, but you won't actually ever complete it now. Like it's literally there as like a wrapper. And if you start to say, the moment you start typing in your first name, this Qualified agent, the Amelia AI, will pop up and say, hey, I saw you're interested in, you know, sponsoring SaaSer. Just shoot me your email and then I'll automatically, on the backend, it's round robinning it, but I'll round robin it to a sales rep here and it'll show the meeting booker instantly. Again, I know there's a lot of other things that do that. The magic though of Qualified that we've seen is that it does have all of our data. So unlike Artisan, it does already have full integration with Marketo, full integration with Salesforce. It is so nice that it knows this. So like when they're coming to the site and people will come multiple times usually to our website, it knows, it knows how many times they've been to the website. It knows if they've been to SaaStr Annual, it knows if they sponsored SaaStr Annual. It's no, literally if they opened the SaaStr newsletter from Tuesday, it knows all of this because our Salesforce and Marketo knows all of this. So that is actually— they call it Piper internally at Qualified. That's the magic of using a Piper agent. Ours is called Amelia, is that it knows everything. So in that context for inbound, I actually like that it knows everything because these are folks that are obviously warmer leads, right? They're coming to the website, they're coming multiple times, they're on specific pages to either buy something or get involved in SaaS or For all intents and purposes, they're high-intent visitors. So in this scenario, I do want it to know as much data as humanly possible. And so in that reason, it's very nice that Qualified does that on the backend. So that's what it does.

So it's, and it's, Amelia, it's, it's ingesting our Salesforce and Marketo data like daily.

I set it to weekly cuz I'm like, you know, that's enough. Daily is a lot. Yeah. I'm like, weekly is enough. Okay.

So every, every week it's doing some sort of diff analysis and importing everything that's new from Salesforce or Marketo?

Correct. Yeah.

Okay. And that's automated.

That's automa— yeah. And that's automated. It also will crawl all of our websites once a week too. So like saster.com, it will readjust once a week. So if you write something up about, like, you wrote, like, you did the live coding last week on Workshop Wednesday, it already saw that, it adjusted it, it knows how to reference it. This one will show up, like, next week, right? Saastr Annual, it'll recrawl once a week. So, like, you know, when we add speakers or when we add them to London or make updates to the landing pages, I don't have to worry about pushing it to that. It'll just do that automatically.

Yep.

So yes, that is where we've used it. Part of it was like, okay, I want to, you know, bring our inbound into the whatever century this is, into 2025. And so that was part of like, okay, let's replace the form with something that is a lot more intelligent than just a static form. The other thing we started doing is because it can do the meetings bookings, it's obviously gonna have a better response rate on meetings than we used to have manually, right? So if you filled out the form previously, we would then, I would round robin it manually, then I would send it to rep, the rep would reach out, and then, you know, they'd have to go back and forth on picking a time, booking a meeting. Now it just books the meeting. So literally, David, our team will wake up and he'll be like, yeah, I got like 3 new meetings from Qualified. I'm That's great. That's what, that's what we want. I didn't have to touch it, but I do have all that data, right? I can see the conversations they had that led them to sponsoring. I could see who they were. I can also like, I'll go through with David kind of daily and say, okay, like, hey, who, who did you get from a qualified meeting? Like, let's see what they said on the website, website. And then you can also see like what their company's doing. So you can see if their company is on the website a lot, right? Maybe they're taking it more seriously. You can see what pages they were on. So again, it just gives you like extra like buyer intelligence context for how they came inbound versus again, static form. You're just kind of getting their name and email. You're not getting the context around it. So that piece has been really nice for us. The other thing we've started it on, there's like, and then there's one like somewhat future use case, which is literally in the next couple weeks is, I started to train after we nailed the sponsorship piece, which was the first piece we wanted to solve. I started to train it on direct ticket sales. So much like our RSM, this is a little bit different because this is inbound versus outbound. It can, again, it can see when people are on like the SaaS or London site, it can see if they've been a couple times. Now it will preempt them and say, hey, are you thinking about attending? You are? Okay, great. Here's the discount code to attend. And then if they don't use it, the next time they're on the site, it'll remind them of the discount code. So this is a new thing. We've, we literally adjusted that at the end of August, worked really well. So we're going to keep doing that. And then we, as you can see here, we started a prompt for like, do you wanna inquire about group tickets? My concern on that honestly was that if I turned that on, we were gonna get too many spammers that were like, hey, your chat already gave me a promo code and actually I just didn't like it. So like I'm inquiring about group tickets, but really I've actually gotten the opposite. I've actually gotten highly qualified, qualified, I've gotten highly qualified leads companies that actually need group tickets. So because how you buy tickets on Saastr.com, it's usually good for a group of like 1 to 4. It's, it's kind of tedious actually past 4 people cuz you have to put in all their info. You have to like do different things. So anyways, it's just easier to do it with an actual person. So this has actually led to actual sales of, of group tickets, which has been nice. Cuz I actually thought it was gonna be abused, but it's been used. Properly. So that, that was like the next, we literally just rolled that out in the last 2 weeks. So that was like the next use case we found it. So now we're actually already up to like 2/3 use cases of it. The other thing it does is it does really good support. So if you ask it, like, people have asked the London one a lot for like the hotel block. Again, it's like kind of buried on the website. So that's a common thing it'll answer now. Like, Hey, I forgot where the venue is. Also, is there a hotel block? It'll just answer that for them. You gotta remember too, like London time difference to California where Saastr is, is pretty staggering. So like I can't answer these questions in the middle of the night. And if they come to the website and they can't find it, it's much easier to just interact with the Amelia AI, ask it all the questions. The future use case, which is we're gonna do before London, and this is the goal with me and Qualified is to turn this into even more of a copilot so that folks can ask it like, hey, you know, if you come to the website again, it's gonna know if you're already registered or not. So let's say you're attending, you come to the website in November, you haven't booked any networking meetings yet. Our Amelia AI will be able to tell you like, hey, you haven't booked any meetings yet. You should, here's all the info on the SaaS for Who Do You Want to Meet program for London. You should start networking. Or, hey, I saw you haven't booked any sessions yet. Here are some ones we recommend you should book and go to on Tuesday, December 2nd. So that is its next date. I'm also adding, like, Jason's Delphi already has voice, so you can literally talk to it, which a lot of people do. They call it. So we're adding, like, voice and video to this agent as well. Again, this, this was adopted a lot later than Delphi, so it's doing a little bit of catch-up in that area. But I would say in the use cases we've trained it on has been successful out the gate. And then yeah, I would say too, like you can see here, like some of our much like Artisan, like our perform— I think they're fine with me sharing this. Our performance metrics is like higher than average. So like the black lines on each of these for like engagements and meetings for SaaS has been higher than average, but still really, again, their averages are still fairly high. Ours just happens to be above average. Which again, I think that's part to do with like the brand that Cestra has and how people are coming to our websites versus like a more common product website. I'm going to read some of the questions too.

I think it sounds like, listen, a lot of these behind the scenes, you can read the questions behind the scenes. These apps are a lot of mashups of other APIs, right? They're API, other services that can figure out who's on your website and figure out all these things. But, but the, the, the, what's, I think one of the things, and others can do this, but that works for us for Qualified is we have so much data about so many people over a decade, right? That we have so many returners.

Yeah.

So many of the folks that come back to us and, and, and so that's why we do so well. But you might think if you're a newer startup, oh, I don't have folks that are returning. Of course you do. Everyone's, the newer you are, the more folks are in discovery. How many times do I have to come back to your website? To due diligence, to think about it, to when a sales rep tells me something, when I hear a referral, they're all coming back multiple times. As Milly said, she— we're shocked how many times, now that we track it more carefully, we could have done it before, how many times people come back, right? So you should leverage this, all the data, all the interactions, everything you have to, to get them, right? To personalize the experience and get them.

There's a one thing before we move. We've talked about Delphi a lot, kind of interweaved. There's, there's a question in the chat I think is helpful. So I'll show you guys the back end of Qualified for a second. All right. So somebody asked, if you have a more enterprise ASP, can you still use a tool like Qualified versus having the customer directly connect with the rep? You can do both. I've literally done both on Qualified. Like, one, our ASP for tickets, yes, is a couple hundred bucks for London, slightly more for SaaS or annual. But our ASP for sponsorships is probably $85K. So I don't know if you consider that enterprise, but I would say that's more of an enterprise ASP. The nice thing about it is you can at any time, they call it pounds, you can start talking. I don't know who this is, so I'm not gonna talk to them. At any point, if you see that they're typing, which it'll show you like what they're typing in real time, it'll like, it's literally showing me what page that they're on right now on sastr.com. You can say, hey, it's a bit like you can hit this and start talking to this person. So for us, it's not an either or, like I actually find it very addicting to have this just, this is literally, I have Qualified always on as one of my tabs and I'm continuously seeing what it's doing, where people are going and it'll alert you. It's got like a Slack integration, so it'll alert you, like you can set it for certain signals, but it'll alert you how you want. So if you want to jump in, and have an actual rep start a conversation with them, you can. Like, it's meant to be fully automated. Most of the time people get our fully automated experience, but there have been instances where I see folks are maybe have a more nuanced question related to sponsorships, which is a higher ASP, and I jump in and I start talking to them. And then, you know, I, I get it to show the meeting booker, et cetera. So you can do both. Uh, before we run out of time, um, I guess just anything else you want to hit on Delphi, Jason? We've hit it a bunch, but no, the only thing I would—

listen, if you want to try our general chat, just go to saster.com and just click on the chat on the bottom right if you haven't. A lot of you have, 139,000 conversations. Uh, it's really good. Um, just, just the one reminder from it with— this is the first one we started, and what but there, there are other tools that do this, that create digital clones. But what it does is it mostly automates ingestion. So it automates, you can automate, it'll crawl our website, it'll automate the RSS feed from our blogs. It ingests all my tweets, it ingests every YouTube video we've ever done. Then we got SaaS for that. That's a lot of content. So out of the box, it was really good. The quick thing we learned is then to quickly upload some very specific documents. So when this is the only tool we had and like, well, let's upload the, the sponsor prospectuses for, for our sponsors.

Yeah.

It became magical because all of a sudden it could answer those questions. It would get some stuff wrong, so you have to read it each day, but it would get pretty good. So we'd have to review it daily. The, the meta-learning is, um, listen, to get the results that we've seen from the SDR and the BDR, you have to put in a lot of effort. But even just some general tools, man, this is so much better than get back to you in a day. It's so much better. So if you wanna start, we started first with, or you could use Qualified as a general tool too. Like it's not a perfect support tool, but it could, like these tools all can do more than you think. So pick a platform and just keep investing in it. It's so much better to get back to someone with a 90% perfect answer in real time than any brutal other process. There's just no excuse. So my point is, listen, we, we started with one and now we have 20, right? And 11 core ones. So this general platform was not enough. But man, just having a versatile platform versus nothing is so much better. It's just so much better, right? There's no excuse to not have a super well-trained AI talk to your prospects or customers in real time if that's the way they want to talk without a human. There's just no— so if nothing else, pick a general tool and run with it is the meta-learning.

Yep, for sure. Quickly going through questions. Okay. I answered some in the chat. It's just so we can keep going. All right. Two more if we have time and then Jason M. Lemkin can chat about QuickListable once we've live coded ourselves. Sales Collateral. I'll touch on this one real quick. Like, again, I don't— I actually don't remember how you found Gamma, Jason, but I just knew.

I just tried it. This is one, and it was great. Uh, but then I said to me, like, can we customize this so that every single sponsor prospect gets customized collateral? Because it was driving me nuts at how crappy— like, our decks were pretty good, but they were so uncustomized, I wanted to cry.

Yeah, yeah, it's— yeah, so we use it. That is our primary use case. I actually just talked to the head of product yesterday at Gamma, and they actually said they have actually more B2C customers than SaaS/B2B customers at this moment. But our use case is like the most common use case for the B2B customers, which is automating your pre and post-sales collateral, right? So like to Jason's point, we have right now this prospectus on our website, which you can download, you can talk to our AI and get it. And see it and it's like 100 pages. I don't know. It's crazy long because it has everything. It's got our media, stats, re:annual, London, podcast. There's just a lot of detail in it. So it's fairly long and like hard to digest. So where we started to use a gamma was, okay, if I wanted to talk to somebody who is a prospective customer and I didn't want to send them this 100-page deck, I want to send them something more custom. I now just load it into Gamma. I tell it which slides I wanna keep from our long, long-ass prospectus deck, and I tell it what the company is. I tell it other points and it'll automatically spit out a, a more customized deck for them. I send it to them. It's a lot easier for them to not only use and reference, but it'll get passed around, right? Especially if you host it on Gamma, it'll give you like an alert of anytime somebody reads it or sees it. These will get passed around because we've literally made it for them specifically. Like, so if it's the VP of marketing we're talking to and they need to run it up to the CMO or the CEO, send them a, you know, either pre or post-sale highly customized deck. They can share it around. There's nothing like controversial or secret in it, right? They can share it freely inside and then they don't have to do it. Like a lot of the times we do these decks now and send it to folks They, the number one response I surprisingly get is like, oh, I was gonna have to actually do that myself so I could show the rest of my team, or I could show it to our decision maker or our buyer, or actually legal needs something for like procurement. So a lot of times it will actually, not only does it look a lot better, it feels a lot better, it's more customized to them. It will save your buyer a step. And so obviously if you think about those implications, you want to make your buying process as easy as possible. And so now we like live and breathe by doing these little decks. Again, probably takes like 10 or so minutes to make one, just depending on how customized you wanna get with it. And then it'll spit it out instantly. The other thing I've found is I've used it for like our internal docs or docs I share with like certain VIP groups at Saastr. So like past sponsors, past attendees, if I wanna make something, you know, we don't have a full-time in-house designer, so I can't, you know, He's great, but I can't Slack him and like wait 2 days to get a PDF back that looks SaaS or customized. I just do it now in Gamma. I, you know, I uploaded our brand guidelines. I said, here's our guidelines, here's our colors, here's our logo. Give me something that I can just, you know, use in 60 seconds to pass around because I already have all the content I need for this. So that's where it's been, again, highly well received when we've sent these, because it saves folks a step and also just super easy to use. Like, It is as easy as Canva to use Agama. Last ones we've added, I know Jason, you've had a, a, a post about this that went a little bit viral in the last couple weeks of a true story of how when we've implemented AI RevOps between ourselves and some of the companies Dostr knows/works with, anytime you have an AI RevOps, somebody naturally kind of quits. It's because it knows what you're doing. And so I think for some reps that maybe don't have as much engagement in something like a Salesforce, because this is pushing and pulling directly to Salesforce, it is something where you will like literally catch people out because you're implementing a RevOps tool, right? I bring a moment to my attention because they're good partners of SaaSers, but they help us do automation and we use Salesforce. So specifically in Salesforce, It will analyze all of our sales calls automatically. So now I just get a Slack. There's some setup there, right? It's not just gonna be like, okay, David had a call with XYZ company. Here's the summary of the call. It does do that, but we've also tuned it on— there are certain things I want it to say. So I, I always wanted to do next steps, which you'll see in the screenshot. I always wanted to do, were there any objections? What help does either the person or does David need to get the deal to the next step? It will automatically attach the contacts to an opportunity. So we're really bad at doing this. Your sales team is probably bad at doing this too, right? Like we put a lot of accounts in Salesforce. Sometimes we forget to put the actual individual contacts. It will just create the contacts for you when you have the call. Then you go back into Salesforce, they're already there. You don't have to do anything. For us, we've always, we've also set it so that it will change our stage. So if this is a first step call, which is like a, for us it's like a 0.5 step in Salesforce. It will move them to step 1, like meeting held anytime the meeting has been held and then it's over. So for us it's just nice because I have spent the last, I don't know how many years pulling my teeth trying to get our sales reps to put things into Salesforce. So now instead of pestering them all day to do it, the AI just does it. So they do have to, again, they do have to hook it up to Salesforce, it does give me reports on like how much activity they had in the week, how many calls, action items, red steps. It also has some like, both of them have really good AI where you can automate like the next steps if you wanted to with the AI. But again, there's a lot it does here even though maybe it seems simple to you guys. There's a lot it does behind the scenes similar to like the qualified where it knows these accounts, right? It's hooked up to your Salesforce data. It knows who, how the call went, it'll automatically push all the notes, right? Like we didn't have any call notes in our Salesforce prior to this. Like we just had kind of like loose information. Like if there was a close one deal, it was in Salesforce. Otherwise it was pretty much not in there, which means it did not exist. Part of this too is like the, I guess just other learning on it of doing like an AI RevOps tool is pretty forever starting from when we implemented this. Now we have great Salesforce data, right? We have healthy data on all these accounts. So when it talks to— when Qualified sees that they're on the website or Artisan sees that we're outbounding them, it can reference this, right? Because it's all pushing to Salesforce. Okay. We only have a few minutes left, but if you want to go through— do you want to go through like the vibe coded one?

Yeah, I'll go. We go a little, little, little bit over even if we lose people. It's a new world. But anyhow, there's a lot more that we do. But we just wanted to give you the core GTM flow that we spend the most time in, which is what you just saw, the SDR, BDR, generic AI, and RevOps. But there's a lot more. And it's a bigger topic. Certainly, we've written about it and tweeted about it. But we, we have built a bunch of stuff ourselves in Replit. You could build it in and other tools as well. I just, I mainly picked Replit in the beginning because it was an end-to-end solution. But ultimately, all the leaders in live coding basically use Cloud Code and are more similar than they're different. But where there was nothing to buy, we built. Um, and there's a bunch that are cool. This one is niche, but some niche apps are— could be pretty powerful. So for example, this is specific to us, but it is interesting. We get thousands of speaker submissions a Okay. And if we get a handwritten email from Larry Ellison saying they blew up the quarter on AI, I'd like to speak at Saastr, we'll just— we'll get that one. We'll see that one most likely. But we get thousands and there's gems and it's too many to process. It's way too many to process. And so what we did until through last year was we hired an agency. They worked with us for years. We paid them about $15,000 a month or more. And part of their job was to review all these inbounds. And it was okay for a while, but they got tired of doing it and the volume was really high and they might have heard of HubSpot or Salesforce, but they might not have heard of cool tools like Gamma or Replit or others. And so they'd miss things and it was a lot of volume and then they would hold things and people wouldn't know if they were accepted and they get grouchy cuz they, it would take months to hear back from the agency. And so when the agency didn't wanna do this anymore, um, after a while I just Vibe coded a speaker application and it's pretty damn cool. You could— don't, don't do too many, but you can try it on sasterlondon.com. Do a speaker application yourself, put in it, and it's pretty good. It uses all the criteria that we want and then it instantly grades you on the website. Here's one from 8:57 this morning, 54 out of 100. It tells you what your score is. It tells you why you might or might not be a good fit. It tries not to be triggering and it sends us an email and a Slack notification of who they are and why. And it gives us actually much more brutal feedback on their pitch than it does to the candidate. But man, now everybody knows this is so much better and saves us so much time. And this did take iterations. This was not a 15-minute project. It probably only took me 3 or 4 hours to build this, but it was over multiple days. So, so like it has to be like at least a P2 priority to build it yourself and then maintain it. But pretty damn cool to have an agent reviewing thousands of speaker submissions and content submissions rather than spending hundreds of thousands of dollars a year and getting much lower output from humans that don't want to do this. They just don't want to do this work anymore. So I built an agent. So that one's pretty cool. You can try it and you can— I mean, it's kind of epic. That's an internal, mostly internal use case, but you can try it and do a submission. The next one. What's the next one we got, Amelia? And just a few others you can try that we built ourselves. And again, at the bottom, buy where you can, don't build. We can do a follow-up session on this. I know some of you have followed it. We built this whole speaker review. We do content review with AI agent. We built a startup valuation calculator. You could try it on Saastr.ai. We've had 230,000 valuation calculations where you can value a startup. We then, once we did the, the, the speaker review of the 2,000 submissions. I'm like, let's use a version of that to help founders review their pitch decks to VCs. Uh, I launched that this week. We've already had, I think, 450 bona fide VC decks go through the system. It gives you deep feedback and scores based on 4,000 recent VC rounds and 800 VCs that have come to SaaStr. It's pretty cool. Um, we had our SaaStr AI London website built in Squarespace for years, but it was driving me nuts. I couldn't do what I want to do it this year, so I rebuilt it from from scratch in Replit. That's pretty cool. And then we built a whole new version of our Saastr website, which you can try in Saastr A, which has so much stuff. And all this stuff is cool. Um, it's interesting when I look at the order I made this in, they went— I went from necessary to nice to have. So probably anything below this line, I would not vibe. I would buy it. There's no way you're going to build an Artisan, a Salesforce, a Gamma. You could build the little bits of it. I mean, there's just no way. There's too many man hours. But if you have nothing, the beauty is you can vibe your own agent now. Um, just be patient. We've done a lot on it. So a quick look on other things we can— we've done a bunch on this. Maybe we'll do another in a couple weeks on what we've really learned vibing. Um, you can try them all here, but it— but you really can do it. Just stay small, keep it simple, and, and buy if you can. So I think that's probably it that we have time for, Amelia, unless you want to do anything else today.

Yeah, there was just a question earlier, um, That is probably worth spending 2 minutes on.

Sure.

And it was, how do we think about managing since we now have more agents than people? How do we tackle that? Which I will say a lot of the, let me get there, a lot of the slides in the back half of this deck, which is in the chat for you guys, it'll be on sas.com if you can't find it later if you're rewatching, talks about that. But I don't know if you have just like a 60-second answer you want to share.

I just say two things. One, I know it seems like a basic heuristic. If you're going to implement the number of agents we all can be running and should be running today, um, roughly you're going to spend as much time managing them as you do humans. Maybe it's a little more, maybe it's a little less, but I really— I think you would agree with me, the amount of time you have to spend managing them is about equal to the amount of time you used to spend hiring agencies and people In some ways it's worse because the output's so much higher. You saw like thousands of outputs out of our AIs, right? Rather than a human, which, you know, may have fewer outputs, but there's no arguing. They work 24/7. And so it is, we're more tired with all these AIs because they do more, like they get more crap done. But we're, every time I ask Amelia, would you rather have a person do this? The answer's always no. But you're not just like when you hire a VP of sales, you don't as a founder, you don't get out of sales. When you hire 10 AI agents and you actually make them work, you don't get time back either. You don't get the management time. It's just you got to manage the AIs. So it's just as much time. The other, the second point to make, then we could break. So that's on off, off-the-shelf stuff and buy it. You know, the problem with the stuff that we vibe-coded is great. Okay. You could see it, right? It's cool. Here's the thing though, guys. Like every day I got to fix stuff. None of these vibe-coded apps are fully stable. Now, some really simple stuff is like if you just basically built a static website, it's probably stable, but anything with any complexity breaks or needs upgrade updates. And I compare the low-end stuff to Shopify and Squarespace, but like those, the, the, the, the hardcoded elements of Shopify and Squarespace just always work. Work, right? The webpage always renders, the billing always works. But like all the stuff I vibe coded, every day I'm fixing something that's broken. It breaks. And sometimes the agent broke things. More often it, it wasn't fully spec'd the way we thought. Sometimes, a lot of times API connections are unstable or ingestion ables are unstable. But literally, I don't mind because I am learning, but every day I'm futzing and fixing the ones we built ourselves. If you don't have those capabilities in-house or you don't want to do it, I would not do anything remotely complicated vibe-coded because it isn't just like third-party ones you buy are not set and forget, right? Because you have to monitor it. Anything you vibe, you're the DevOps team, you're the TechOps team, you're the QA team, and they all break every day. They break. Um, so if you're not up for it, um, find a way to buy.

All right.