SaaStr AI Live: The Top 5 Issues Managing Multiple AI Agents In Production

Um, okay, we are live. Hello everybody, welcome back to Workshop Wednesday slash Sastra AI Live. Excited to be here with all of you and your agents here today. Um, we thought we would do a fun one since we have 20, actually almost 30, I think when I counted it this morning, um, agents and Vibe-coded tools and apps in productions. We wanted to pick our, uh, we narrowed it down to 5 plus 1 bonus. We wanted to talk about our top 5 issues. Once you start to manage these multiple agents in production, what does that really mean? What are maybe the nuances folks haven't thought of if you've only deployed like 1 or 2 agents? Kind of like what breaks when you have multiple agents at a high volume and high scale. So, um, we're gonna hit that today. And then side note, today's an exciting day because Jason's gonna be on the Salesforce earnings call later today. So, um, you can watch it at 2 PM Pacific. Um, you can also watch the replay if you miss it. Should be exciting times. So It's a big day at SaaStr in Salesforce. All right, fun times. Let me keep admitting folks and go. Okay, as always, I always like to have this as our first slide because folks go back to it and reference it. So for any of the agents and tools we'll reference today, you can go to saastr.ai/agents. That's where we put our agent directory. Which I have a few more we need to add to it, but this is our depository of all the agents and some community agents that we use across the board. So a lot of the things we'll talk about today are linked there if you're looking for the links. So as always, keep that in mind. Okay, top 5 issues. I'll summarize them here and then I'll go into them in depth. I think there's a lot we've learned in like the now 9/10 months of managing multiple AI agents in production between—

This meeting is being recorded by Update AI.

Between Jason and myself. So I kind of narrowed it down to the top 5 things, and then I have a bonus. And then Jason, obviously you can feel free to add anything I might have missed in these top 5, but I do feel like these are truly the top 5 impacting our day-to-day now that we have so many agents and apps in production out in the world that are AI-powered. So yeah, I'll go just briefly through them and then I'll go through each of the 5 in detail. So Um, I would say issue number 1, and this is maybe first ranked, is context switching becomes extremely difficult when you have 20+ agents. They don't all function the same way. If you guys saw the article that Jason did over the weekend about like how we use Salesforce and AgentForce is kind of like the backbone and the plumbing for a lot of these agents. You know, there's a reason we do that, and that's because a lot of these agents function somewhat independently. Some of them push back to Salesforce, some of them don't, some of them talk to Claude, some of them don't. And so it's a bit of a hodgepodge today. Like, that's the reality of managing all these AI agents and productions. They all speak slightly different languages. They all ingest context very similarly, but still different ways. They all have slightly different personalities, I will say, these agents, and they all kind of want different things from you. And so I think context switching just like in as a human brain keeping up with the agents becomes kind of difficult on a day-to-day basis. That's, so again, I'm just gonna go through these quickly and then we'll dive into one of each of one of them. Number 2, I think the new agent blackout period for, again, for those of you who have maybe added a couple agents, but now you're getting into the high volume of multiple agents. For us at least, even with the help we get from our vendors, which we'll talk about a bit, even with the help we get, like, I feel like every new agent costs us at least 2 weeks of what I call either a blackout period and somewhat chaotic of like, I kind of have to ignore the other agents anytime I bring on a new agent, which, as you might imagine, since they need daily, um, review, it's kind of rough on the entire stack. It's just like this cascading waterfall of effects that happen every time we add, um, a new agent, of which I famously said on LinkedIn I was not going to add anymore, and then I added 2 more in the last 2 weeks. So there goes that. And, and I can share some reasons why if you didn't watch the one from last week. Okay. The other one, I would say number 3, which is a huge issue, actually, maybe it's number 1, but the AI, what we call the AI agent succession planning crisis. I actually asked our agents today what they would do if I had to pass them over. And the answers were actually a little bit surprising. You'll see why when I dig into that one. The fourth one, I would say the agent as the truth teller. One of my agents, more famously than the other, the 10K agent, roasts me all the time, every day, because I talk to it every day. Like, that's the first one I talk to each morning. It will literally roast me all the time, which is funny because when I pushed it on that for this webinar, it said it wasn't roasting me. So that's, that's an interesting learning too, where it said it was being helpful and I thought it was roasting me. And then last but not least, before we get to the bonus one, I think there is a lot of, you know, compliance and security, maybe not issues, but just nuances and maintenance that people don't talk about. That is a real problem, right? I think that's why most high, you know, highly secure— like, I'm not going to vibe go to Salesforce for a lot of reasons, but part of the reason is because Salesforce is secure. Like, that data is there. We've got 10 years of data there. It's the plumbing for a lot of our other agents. But also it's, you know, they've thought about all these things like SOC 2 and ISOs endlessly so that I don't have to. So that alone is like kind of worth it for us to use it and keep using that. But I think there's other issues when you start to vibe code your own apps and agents that most folks probably haven't thought about because you, you're so used to these other tools like Salesforce doing it for you that even myself, I didn't naturally think about doing that myself when I started pushing apps into production. Anything at a high level you want to add, Jason?

No, I think this is, this is enough for today. I think, I think you'll get to the next one. It would be useful just as an overview. When you log in each day, who do you talk to? What does that take? How much time does it take? That, because that feeds into the first point, doesn't it? There's context switching to ask. There's also just this, and part of that is all these different interfaces, all these different exceptions, all the different things when you log in each day to Artisan or Qualified or Agentforce. there's a lot of time in different interfaces and different conflicts. Um, that'd be the main one. And then maybe just a quick reminder to folks, maybe you have it in here, how we segment our base for these different agents, because it's related to the first point, right?

Yeah. Yeah. Um, yeah, I would say I'll come back to this because I feel like this is kind of a good recap. But, um, yeah, to your point on the context switching tax, right? It's, it's It's not even that I think about them as 20 agents anymore. Like, I think the better comparison to make to folks on this call and watching live is you have to almost think about it now, and I think about it now, as managing 20 different people, which like before we had 20 agents, I don't even know what we had maximum headcount here. Let's say 12 at the very, at the very peak. Um, Now we have 20 plus, almost 30 different agents, plus, you know, 3 humans and a dog that we manage. So I think that's a better way to frame it to folks is to think about each agent almost as its own entity, as your own employee that, you know, speaks a different language, has a different personality, needs different things from you. Like to Jason's point, like I'm not immune from logging into each of these these tools every day, seeing what's going on, right? So like, this is a good example of my, my kind of like morning. Even just this morning, like, I actually didn't even have time to go through my whole checklist because we are on this call at 10. So I was like, I haven't actually even made it through all of my agent check-ins for the day, and it's already— now it's going to be like 11 by the time this is over, and I have other calls today. So like, it's basically gonna take me the whole day just to, just to check in with all of our different agents. So I don't really know if this is my morning routine but my daily routine, but So yeah, kind of a quick example of the ones I check the most, right? So I would say I actually log into 10K kind of first because this is our main internal AI VPM. We've done a previous episode about it where we did a deep dive, so you can go back and check that one out. But just as a high level, this is one that is running on internal data. It has some API data. It's literally telling me what I, what I/we should be doing every day for, you know, tickets to Sastre, paid tickets to Sastre Annual, our VIP summits that we have, our sponsorships that we have, our sponsors, uh, media contacts. Like, there's a bunch of high-level, um, it's high-level, you know, goals that I'm trying to hit with $10K. And so I check that one every day just to see where we're at and like, am I falling behind? And that's why this one roasts me all the time. It's like, Amelia, you didn't get, you know, these 5 things that I laid out for you done yesterday. I'm like, I had to go to sleep. Thanks for roasting me. I understand why we're behind. Can you just help me figure out how to get caught up in a way that I can actually deal with the time that I have today? And so then it will kind of, you know, reevaluate its to-do list for me for the day. But 10K literally tells me what to do. So that's why I log into that one first, because actually 10K2 will tell me what to do with our other agents. That's also why I log into that one first and foremost. That one is a mix too of Claude and Replit, so I'm kind of like bouncing between those two entities, which, you know, a lot of Replit runs, I'm pretty sure, on Claude. And so for that one, context switching is not too, too hard because they kind of speak the same language, but they have different nuances in how they interact with you, right? Like 10K as Claude is obviously more like text-based, whereas 10K as the Replit agent has a lot more dashboards, has a lot more numbers in it. So it's kind of a combination of both that I'm talking to at the same time there. Or then I'll go into, you know, what are our outward-facing agents? I'll go into all of our sales agents and ticket agents, and that's Artisan, Qualified, AgentForce, and now Monaco. So that's 4 different agents I have to log into, look at who needs a human response, look at who it's reached out to over the last, you know, day between the last time I checked it yesterday, let's say morning, and see what it's doing, right? Like a lot of times I'll need to refresh what the agents are doing, who they're reaching out to, the copy they're reaching out to. For instance, I have it in the next screenshot here. Like I— you can see where it's roasting me already. Like I had to context switch for this week when I was doing this on Monday because I was like, we have a distinct promotion going out this week for SaaStray annual tickets in particular because the ticket prices go up on March 1st. And I literally had to put that context into 5 different agents. Like literally you can see some of the screenshots here. I was like, I have to go in and tell Artisan that. I have to go in and tell Qualified that. I have to go in and tell Salesforce that. I have to like, 10K already knows that 'cause he's the one that came up with it. But then he's telling me, why aren't you launching LinkedIn ads immediately for this new promotion? Do it right now. I was like, I will do it as soon as I'm done talking to the other agents. And so context switching becomes extremely like, I don't, it's just hard to keep up with, I think sometimes in your brain. Um, and a lot of the times it's hard too for the agents to remember everything you're doing, right? So, um, I think there's kind of like a double-edged sword there.

And so then the, um, at least for us for today, if you're literally logging in and spending anywhere from 10 minutes to an hour with 4 to 5 of our core GTM agents, right? I would say every single day, right? And different user interfaces. You have to remind each other of the changes, the campaigns, because they don't talk to each other. I know we haven't found anything, so it's slightly rhetorical question, but people talk about orchestration agents and master agents. We haven't found anyone that can integrate Agentforce, Artisan, Qualified, Monaco, Momentum, Attendant. Like we don't, there's nothing out there today, even with 10K, which we built ourselves, that at some magical API or other level can integrate everything for you. That product does not exist, right?

No, it doesn't exist. I will say the benefit of Salesforce, um, buying qualified momentum is like eventually by probably Dreamforce, right? Those two things will be integrated into one. So at least I have like two less dashboards. Um, and I'll probably have one centralized one by October. Um, but only because, you know, they've, they've acquired those tools. For most tools, it's, it's completely separate.

Yep.

Yep. So it's a lot.

So you're gonna have a one-on-one with your agent every day. Instead of a weekly one-on-one, instead of a check-in on Slack, if you want to accomplish what we've done or better, you have to do a truly and honestly a one-on-one with your agent every day. Even if you don't want to do the work, you got to do the work to see one-on-one with every agent every day.

Yeah. And I think it's, um, an interesting learning too. I didn't put it here, but like one thing I do with my agents is I ask them every day, like, what should I be mindful of? What should I be looking at? Um, and then sometimes it'll say, hey, like, I can give you a recommendation, but you need to give me some data back, or you need to tell me what you're doing with your other agents, or you need to give me XYZ inputs in order for me to output that. And so that is a very good way, I think, more so to look at it is just think of this as 20 different people that you have to manage every day, right? Sometimes you can get away with doing a one-on-one with a human once a week if you're, you know, lucky. This is, you can't, you can't wait a whole week. Like that's the thing. It's like if you wait a whole week with any of your agents, it'll be so out of date by the time the week's over. Over.

The production, the output is so high.

Yeah, yeah, exactly. You can't— you— it just can't afford to wait a whole week. Or like, you're— or you're probably just wasting— honestly, you're probably wasting your money if you don't spend the time to do it daily, because a lot of these are waiting for you to give it inputs, right? Like, it's such a high volume. Like, Art is a qualified agent for us. Like, I'm constantly giving it more people to reach out to, better messaging, different campaigns, different things. We're like If I don't do that, most of those don't do it themselves. So they're just going to sit there like your agents are just going to idle if you don't check in with them every day, which I also think is just like a waste of time and money because what you want is to have that high volume output. So you're maximizing what the agent can do, right? So it's an interesting nuance. Okay. Number 2, I would say biggest learning is we learned this kind of quickly, but I think it's just more so now getting harder now that we have 20+ almost 30 agents, but you know, in the beginning it used to take us more than 2 weeks to implement some of these agents. I would say now we've pretty much gotten it down to about 2 weeks, um, for a lot of reasons. One reason, like a lot, again, even though context switching is hard, you can share a lot of context between agents. So kind of once you have one, you can do another one. It gets easier. It's not easy, but it gets easier. Um, but I would say we've kind of gotten it down to like a 2-week implementation pipeline for any of our agents. It's true for Monica that we just did. It's true for Happy Fox that we just did. Like each of those just took 2 weeks and that was not spending the entirety of the 2 weeks on it, but it was spending the majority of 2 weeks on it. And so this just gets harder with every new agent that you add because each new agent means you're not spending, you know, that daily time I just mentioned. You're probably not doing that if while you're onboarding a new agent. Like I know I didn't spend the time I needed to with each of our agents. In the way that I should. I usually do when I was onboarding these new agents because I was like, I want to get these off the ground. I want to get them up and running. And so I'm just going to have to do, you know, I'm going to have to do the tradeoff and let the other ones slightly degrade or slash sit idle while I get these new agents up because there's just no way I can do it all. And so I think that's a huge tradeoff when you're looking at bringing in, you know, multiple agents. Is instantly things go stale with your agents, right? While they're idling, they go stale really fast. Not because they can't do things on their own, but because they're again waiting for you to tell them what to do. So if they've run through, like, let's just say one of these outbound agents has run through a whole contact list, but I'm busy onboarding another agent, it's just sitting there. It's literally idling. It's sitting there. It's not doing anything because I haven't had the time to give it new contacts. Like, That is, I think, just a huge nuance folks don't really talk about is anytime, you know, we're all trying to push like, okay, we, we, we also push, like we have a lot of agents and I think there's a lot more agentic things people can be doing. But just realize like anytime you bring on a new one, it's gonna take time. It's gonna cost money for these agents. It's gonna cost some of your, you know, mental overhead and mental bandwidth and mental time. Like these agents do not deploy themselves. And so. Even with all the help you'll get in the world, which you should only pick vendors you get help from. But even with that, like, again, 2 weeks is pretty fast. I think most agents, especially in the early days, took us like a month, maybe 6 weeks to get up and running. Now we've gotten it down to 2 weeks cuz we've gotten pretty good at it. But I'd say in that math, you can really only add 1/1.5 agents per month max. Like otherwise you're running in place cuz you can't really keep up with your current agents to add the new agents. So it doesn't actually— maybe it doesn't make sense to add that new agent if you can't keep up with it or know that you can, that you'll need to do kind of a blackout period to do a new agent. And so I think this is super nuanced, not something I really realized until we had multiple agents in production, that there is a trade-off. Like it's— you're going to get a lift from it for sure, right? If it works the way you think it will. But there's a trade-off. Like it's not free— it's not free scalability instantly. Okay. Um, no. Oh, these are— yeah, these are screenshots I added. So I added some of these screenshots of Monaco. This was like one of our newer agents that we just did. Um, again, we got help from the team to, um, put us into the product. I think it took, you know, probably 2 weeks. Um, conservatively, maybe a little bit less actually, because we, we worked fairly quickly with them. But again, it was one of those things where I blocked out doing other things and actually the Monaco team works over the weekend. So I did a lot of this over the weekend, which was nice. But we got it up and running in basically a week and a half. And you can see here, like in its first week on, you know, it reached out to 64 people. It booked 6 meetings instantly. Some of those are like tier 1 accounts for us, which is pretty dang good. So again, is it worth it? Yes. From that perspective. It was worth it to let our other agents kind of like sit idle, um, slash do a little bit less while we were getting this new agent on board because we believed it would be worth the tradeoffs, right? But just know that you're going to have to probably weigh tradeoffs when you come into something like adding a new agent and maybe just plan for a blackout period. I think that's my kind of meta point is if you plan on it, um, it'll probably work more to your benefit versus just going into it like, oh, I can just add this in a day. That's not really going to work for you or your agents. So, okay, number 3, and this may not apply to all of you, but I think from what I've seen, and this is where I want to get your take too, Jason, I see a lot of folks where it's either a single or like maybe one, like for us it's 2 people, but like it's kind of a single point of failure for our agents. And it's funny because I put some screenshots in the next one of like I asked my agents what, what it would do if I stopped talking to them? And the responses were interesting. But I think this is where folks get themselves into trouble because they think they can, you know, recruit somebody to just manage all the agents. That doesn't really work. They think they can just, you know, roll it out to the entire sales team or the entire CS team. That doesn't really work. There's reasons that doesn't really work. Like, this is, I would say, the biggest double-edged sword of like, I do think having a single point of truth with who is managing the agents, right? Like who, from our perspective, which contacts go into Qualified versus Artisan versus Monaco versus Agent Force. Like that is just all in my brain right now. Um, and I'm sure Claude could write it down if I told it to and give it to other agents. Um, if it was really pushed, but like, but then if like, let's just say even if I got the agents to give me an MD file of everything it knows, with me, I'm not sure the next person can really pick it up in the same way. They might say, okay, you know, Amelia was segmenting a lot of her ticket outbound in Artisan and then a lot of her sponsor outbound in AgentForce and then new cold outbound in Monaco. And then, you know, whatever else we're doing, retargeting and inbound in Qualified. But maybe this new person doesn't agree with that. They may say, hey, agent, like, Oh, you're outbound. Cool. I'm just gonna keep one of you, like one of you, like we only need one of you to stay to do, you know, outbound. If it's truly at the end of the day, those are all doing outbound functions. Somebody else might say, I only need one of those agents to do those things, which might be true. But also like we have found like we have all these tools cuz they're super nuanced. They're like hyper effective in certain verticals and that's why we've chosen to do multiple agents versus just one agent. And so. I think that's an interesting nuance you would lose. Like, again, double-edged sword of like, I think having one person know what's going on with all your agents is super important. But then what happens if, you know, you're in a bigger company, you have maybe multiple people working on agents, or you're at a bigger company and that person leaves? Like, how, like, how do you scale past this n 1? Um, and I don't think there's a good answer for this today. There is no, like, again, crazy good agent that can manage all the other agents on its own today as of February, there's— that doesn't exist, right? Like, I wish there was. I don't know if— I think there will be, but there's not one today. And so I think even with all the help we get, it's— this is just a huge issue that I think the market has not realized yet.

Yeah, it's good. I want to keep going, but yeah, I mean, I mean, obviously we're extreme and tiny. I mean, if you left for some reason, we would, we would have to cease using our agents. There's no, there's no, there's no plan B. So, uh, we can keep going. It is a, you know, we're just all learning that you need some sort of chief agent officer, right? Whether it's, uh, whether it's, uh, whatever you want to call them, someone that can manage all of your agents that I think needs to be a bit of a nerd, a tech nerd, ideally a bit of a marketing nerd, because a lot of these products are If you can manage a complex marketing automation tool, you can probably manage most of the products that we use. You can hack the rest. But, you know, thinking through it, I mean, this is, this is Captain Obvious, but I think if you actually find the person that can deploy these agents for real, that actually can be success, just like a lot of these functions, but immediately they should recruit, you need to recruit somebody else on their team and divide and conquer and make sure you have two.

Yeah.

You gotta have two, right? I mean, one's hard enough. There's a lot of— in the early days of agents, when we talk to folks, there's very few people at a lot of organizations that can do this. It's not that hard, but there's very few people. But you, you, yeah, if you don't have two, you're at existential risk.

Yeah. Like literally, um, when Philippe, who's the CRO at, um, Personio, was in London, he, you know, he famously set up all their agents, but then he was like, I, I started using, um, I think it was Qualified or something like with the sales team and Then he was like, I rolled it out to like a few reps to be able to like, 'cause like Qualified, you can round robin like the inbound, right? You can say, okay, like I'm gonna, I'll be the master. He's like the master controller of it, of like doing the training and the programming. But Qualified's actually an easier one to get the rest of your team on 'cause you can just have them respond to folks when the agent can't. You can set up certain escalations and however you want it, right? So for him, he was like, okay, I rolled it out to like 5 of the top reps on the sales team. And then the very top rep, whose name is also Amelia, um, ended up being the most like affluent at learning how to work with the agent hand in hand as a copilot. And so then he decided from that like kind of pilot test with his humans that Amelia on their team would be the next person he trains on how to use the agent. So like he did kind of, you know, like a little bit of a beta with their team and then said, okay, I know now who wants to actually manage these agents the way that I do. And so now I'm going to train that person. So that's another way to do it too, across your team, is to do it that way. But yeah, this again, double-edged sword though, because now it's like now he has to tell all the agents everything, plus he has to tell Amelia everything.

So like, it is the CRO, arguably.

And he's the CRO, arguably. Yeah, exactly right. Yeah. So it's funny too, I asked our agents what they, what they recommended we do if for some reason I got hit by, I said, if I get hit by a bus, what should you tell other folks about yourself so that they could, like if Jason were to take these agents over 'cause I got hit by a bus, what would you tell him? And the 10K version of Claude said it would, well, these docs are on my computer 'cause I use Claude on machine. It would need to give him these, you know, documents, which I'm like, I don't know, maybe you could guess my laptop password. I don't know, maybe you could get them. But like, first of all, that's already an issue because I'm like, these docs are literally on my computer and probably nowhere else. Um, then it's literally saying the next campaigns he has to run for March and April, which I don't think you would want to do. Um, and then I think it's interesting, it called out what I consider the vibe for Sastre Annual, which I hadn't really thought about. It's just It's a little funny that the agent was like, make sure Jason or whoever takes this over, if you get hit by a bus, knows what the vibe is like. So it's fun. I was like, I do actually call it a moment in time. That's really funny. I use that in like a lot of copy and I use that in like a lot of things. So it's a little funny that the agent picked up on that. It wanted you to know the vibe, Jason, just in case I get hit by a bus. I think the other interesting thing was I asked our SaaS sponsors Um, which is an app we talked a lot about last week, um, how many lines of code it has, which was shocking because I think Jason M. Lemkin, you said it's something like 3,000 is like, you know, average or basic or simple. Um, it said 12,000 and it says— it said it was hefty. But then I said, okay, since this is a hefty vibe-coded app that I made for sponsors, yeah, how would you bring somebody new up to speed? And this is where it gets tough, right? Because I'm like, I think what it says is so complicated that somebody else would give up. It's explaining it's a sponsor portal. Fine, that's easy. But then it says how the authorization works. Every sponsor company is a Clerk org. I'm like, that's already complicated because the agent doesn't have my Clerk login. So you're already kind of fucked here. Two, it goes over the tier system that I created. Then it goes over the sponsor data. And then it says, you know, a lot of the sponsor data is hardcoded and that's 'cause I just give it to the agent and then it has like master tables that it puts it into. It's also nuanced in like how I create that sponsor data is I create it with Claude a lot of the time and then I give it to Replit. So again, already running into a lot of issues here if I get hit by a bus. Key pages, the backend, I don't even know what a fricking PostgreSQL really means.

Well, it's our database, but it's, it's our database. It's too hard for someone to take this over.

Yeah. Like, okay. File uploads go to Replit, object storage, submissions sync to Google via Zapier. Again, you're just running into a lot of issues here. Like the agent doesn't know my Clerk login. It doesn't know my Google Sheets login. It doesn't know my Zapier login. It's a, like, it's, it would be a mess too. Hand it off. And then it was funny because then my agent said, don't get hit by a bus. That was its recommendation. I said, thanks, I'm really trying not to. All right. This is number 4 for me. You've already seen it a little bit in these slides. These agents, as I— they're interesting to work with on a daily basis versus humans because you know, some, sometimes humans have their own ambitions, their own goals. Like, I feel like Claude is famously goal-seeking, but in a very different way than a human being is, right? In a very different way. So I think there are times where I also feel like these agents don't really know the date and time and like, they don't realize, they don't realize on their own that like when you start chatting with them at 6 in the morning and you're still chatting with them at midnight, that you should probably go to sleep at this point. Like you can't keep working with your agent. So like, It's just a little funny that as I, as I considered it, but you'll see my agent's response in the next slide. I consider my 10K is like roasting me all the freaking time. Like you can see it here. It said you're behind on summit outreach. I said, I know I am. Hold on, I lost my slide. You needed to send 200 emails by Feb 15th. You sent 87. You're 56% behind target. Now, maybe you don't think this is roasting me. I thought it was though, because it also told me to block 3 hours and I was like, you know, I don't have 3 hours to do this and catch up. Like, thanks. But also now you're just stressing me out and roasting me versus being helpful. So I thought that was a little bit funny because it was— it, you know, what does the agent care? The agent's just going to tell me the truth. The agent has all the data more than I probably look at it, right? Like, I look at it a lot. But it's got all— like, it knows where I'm behind and knows my calendar, knows all these things. And it still said, you're dropping the ball here. Catch up. It's constantly asking me, what's preventing you from doing this right now? And sometimes I'm like, it's literally 11 PM and I'm tired, so I can't really like focus on launching this XYZ thing that you want me to do right at this very moment because I need to sleep first and then do it. So it's just interesting, right, of like, okay, what then, what do you decide as the human leader of the agents is helpful versus what crosses the line into maybe slightly demoralizing of because your agent knows everything. It's not that it's wrong when it's roasting me, right? That is maybe the uncomfortable truth. Oftentimes it roasts me and I'm like, yeah, well, I deserve that roast. You were right. Like, I didn't do this thing that I'm now behind on. But also like because so much like so like Tenka roasts me the most. Some of the other agents roast me slightly, but like because you're getting it from so many different angles and so many agents because you can't keep up with them, sometimes I feel like it does cross a little bit of a line where I'm like, okay, dude, like today, just like sugarcoat it to me like you would a person. Like today I don't need the hard— like I can't deal with hard truths every day. I don't think anyone can. Like I can to some degree, but I think every day, all the time from your agent. Is a lot. So, um, it shows it. So I have some screenshots here of how it's been roasting me, but it's funny because it said, you know, literally it said to me on Sunday, what's— when it was super late, what's blocking you from the LinkedIn Lads from going live? And I, I literally said it's because it was late. And then I said, hey, you've kind of roasted me a lot lately. Can you, can you bring up some examples for this webinar I'm doing? And it said, looking through the transcript, I haven't really roasted you. It said I've been a tough accountability partner, and then it listed ways— this is only 2 out of 5 ways— then it listed ways it should have roasted me. Um, and I said thanks. This is— I don't know if you're, uh, I don't know if your agents roast you as much, Jason, but I think this is just a, like, a fun, interesting one of, like, you know, just something maybe you don't expect from your agents is that they will always tell you the truth and they have all the data. So I don't know, it's an interesting one for sure. Okay, let's get into the last one and then I'll do the bonus one. And then we've, we've got quite a few questions that I want to try and get to. So compliance and security issues. Any time we run an audit or ask an agent to run a security audit, It takes time to fix these. Like, we ran a huge one the other day on sashr.ai. We haven't even fixed, I would say, even 10% of the issues that came up, um, in the audit because it became super fragile once we started like futzing with the code. And like, then, you know, sometimes when you do a security audit, um, it will go too far. Like, it'll start to lock things down to the point where you're like, I actually can't even use my app anymore because it took this too far. So then you kind of have to peel back one by one. It's something you have to walk the agent through. It can't really do it that great on its own. And so this takes a lot of time. Not only does this take a lot of time, the maintenance becomes a lot harder each time you do a security audit and like implement whatever it recommended with, you know, you together. So then when you start to multiply that across 20 agents, it's almost unfathomable, like how to keep your apps and all your data secure. But Jason, you've been running the security more than I have. Like, you're the one who actually told me that I should run it on SaaS.

There's a lot here. I think, I think, um, I mean, we can spend as much time or a little time here. I, I think it is, it is a reminder that, um, in general, I mean, you, you can see this in the press, agents can do more autonomously. That's why they're agents. What's the definition of an agent? I'm actually not sure, but I think it is a piece of software powered by AI that can complete tasks autonomously. And all of these agents that we're talking about, go-to-market agents and others, have access to your data. And there's an inherent level of security in that Salesforce is our hub, right? As to tie this to the beginning, that is important, I think. I do think it is one of the— if no other reason to use something like Salesforce as your hub is that it is secure, but these agents then take data out and And the honest truth is no agents, no third-party agents are more secure than SellSource. They're all less secure is the reality of it. They're less secure because they're startups, but they're also less secure because the agents can do crazy stuff with them, like publish them to the wild, right? Or do other things. It is just a reality. And then just the final point, um, is once you vibe code yourself, which we have done a lot of, they are inherently less secure than third-party apps that you buy. Today, I don't— that may not last because, um, a lot— all of the, the, the products from Cloud Code to Replit to Level Below that have added more and more security, but they're all custom. They all are less secure than a Salesforce or something. So all of the, the apps like us you're going to run are going to have customer data through them. They're all of them, right? You're not Stripe or PayPal. Like, it's not that dramatic in many cases, but bear in mind you're taking some measured risk. So at least, if nothing else, the ones that you create yourself before you launch and then frequently, at least every month, just go to the agent and say, do a deep security audit on my app and just have it do that. Don't assume the apps do it yourself. That's the best actionable thing I can take. And then just be thoughtful that there are— there are just inherent tradeoffs to this that we're all going to have to learn. Right. And It may make sense as you get going in the agentic revolution to maybe use slightly less sensitive data to start. If you're a conservative enterprise, it just may make sense. Just like it always has been for enterprise software. You always start in a, in a little group or a little function that's safer when you buy a startup. It's just this on steroids because you cannot fully predict what the agent will do with your data. It's just a reality.

Yep.

It's just a reality.

Yep. Yeah, I put a few screenshots here of like—

run another audit.

Yeah, run another one. Thanks, dude. Like, Um, there's a question in the chat too, how are we running these security audits? Um, some of them you can do natively with the agent, right? Like Lovable Replit, you can just ask its agent to run an audit for you. Um, other times, again, you may not need it if it's a full-blown third-party tool, but you should ask that tool. Like, I ask Transfer from Mars all the time, I'm like, hey, what is your— how— like, what's your compliance right now? Like, what do you, 'cause you have so much of my data, like, let's just make sure it's not going anywhere it's not supposed to. So I think that's another thing too, is like when you use third-party tools, like we do, obviously we use Salesforce for a lot and that's ultra, like fairly secure, right? Knock on wood. But for the other ones we use, I just ask them, I'm like, hey, what are you guys up to date on? What are you doing to make sure our data is secure? Okay, last one and then I'll do questions. Um, this is an interesting one. I added this because I got a little frustrated yesterday, but, um, I think this— I added as number 6 because this may not apply to everyone and may not be a bridge that you ever cross, but I think if you get to the point where your agents do work 24/7 and you have almost 30 agents, um, and again, your agents know all the data when they answer a question, they have your APIs, they have your sales, like they have all these things they can look at and are brutally honest slash roast you when you can't keep up with them. Um, I think it, it, it makes it harder sometimes. I know at least for me when I deal with actual humans that sometimes I lose patience with them a lot quicker now because I'm like, what do you mean you don't know the answer? Like, my fucking agents know the answer instantly. Like, what do you mean you don't know the, the answer to this? Very basic question. What do you mean you haven't delivered on XYZ thing that I gave you 3 weeks to do? Or what do you mean you forgot something that I've told you literally 10 times now? It does make you on— sometimes I often start to wonder, I'm like, do I need them or can't I build an agent to do this? Because if I'm losing patience with them, I'll probably lose patience with somebody else in the same boat. And if an agent can just do it and the agent won't lose patience, maybe I should just build an agent to deal with them and they don't have to deal with me. So I think it's just an interesting nuance. I think something else I've seen just to be wary of, this is almost somewhat unrelated but related in that I've, I've really find it now in the age of AI and, you know, the early phases of 2026 when folks, um, tell us that they're gonna help SASTR for, you know, let's just say XYZ thing for SASTR Annual. Um, like I'll give you a concrete example. Uh, a couple months ago I was working with a few parties on just SASTR Annual third-party production, things we normally contract out, things I thought an agent couldn't do. Um, and I went out to a few different agencies and production teams to RFP for it, to get a quote for it, right? Normal. Okay, fine. A few folks told me this would be your deliverable. It was one sentence and this would be the price. And it was something exorbitant like $40K or $50K a month. And I was like, in the age of AI, that's what you're promising me? Something that like is one sentence, one price point and nothing else? Like you couldn't even ask your own internal agent or frickin ask your Claude or ChatGPT? How to make a better proposal. Like, that's terrible in the age of AI. I was like, this is just unacceptable. Like, I just didn't respond to those people. And it's sad because multiple people did it. Like, and I think maybe that's just a symptom of agencies versus software tools. And, um, maybe marketing agencies won't exist in a year because of this. But I literally was like, how in the age of AI can you say you have a one-sentence deliverable and a fixed price point, and you can't even use the basic AI better than our team can. Like, what? I, I was exasperated. I was like, this is so crazy to me that so much of the world is still operating this way. Like, I think, you know, AI is probably going to come for them. But, um, I was like, I don't think anyone in the age of AI should be giving you a fuzzy deliverable and a fuzzy timeline. That's just not acceptable anymore. I don't think that was ever acceptable. I think it's just more unacceptable now. Um, the other thing too is, again, I kind of mentioned earlier, but like, I know sometimes people will roast Claude on social media for like forgetting things of like, your conversation got too long, it has to get back to it, or forgot something in the MD file, or accidentally deleted something. I'm like, yeah, but the, the things that humans forget is honestly sometimes worse. Like the amount of times we've had to resend somebody the same thing, like over and over again, is a lot more than the times you have to remind Claude of something that it maybe forgot because it couldn't, like, it didn't have enough memory or storage. Like, I do think it's a little bit different. Again, I do think person— personal goals and goal-seeking is a lot different than the way that Claude is goal-seeking. Like, Claude and all the different agents and the LLMs are goal-seeking to try and help you hit this goal, right? That's why 10K rows me all the time. It's trying to hit this goal. It— and it doesn't understand why I can't keep up with it fast enough to hit the goal. That's a very different goal than maybe, let's say, a human on the marketing team being like, hey, I actually just want to do— I only want to work on ads. I don't want to work on anything else in marketing, right? That's very different goal-seeking with people than it is with agents and LLMs. I also think this is an interesting nuance, but I do feel like I have now managed more agents ever than humans. And so I'm probably worse at managing people now than I am— like, agents, I get. I can, you know, talk to them first thing in the morning, last thing at night. Like, they're not going to be impatient with me if I don't respond to them right away. They might be idle, but they're not going to be like, you know, the same as the same vibe as somebody in Slack being like, hey, I have a question for you. Can you answer this right now? And sometimes they can't, like. So I do feel like now that I've flipped the switch to managing more agents, I think I've probably become slightly worse at managing the people that are, you know, that work with SaaS, especially leading up to SaaS radial. So it's something that I am hyper-aware of now because I'm like, I— there— I have to remember that there are not agents. I need to be more— I don't know, at some point I'm like, I just don't know what's realistic anymore. I'm like, if the agents work 24/7, um, and feasibly any human can use an agent, why can't they keep up a little bit more? So I don't know, it's like, I, I don't even know what to do with this. I just think it's an interesting— that's why I called it my bonus point. Of once you reach this point, what does that really do, right, in your nucleus of how you operate with the team? The team is agents. Like, they are still part of the team though, right? So it does kind of change not only a lot of your day-to-day and the maintenance that you have to do, but I think in some ways it's kind of made me worse at working with people. All right, anything else you want to add, Jason, before we go to questions?

No, I think it's great. Let's, let's, let's hit the questions. Um, and, uh, I think it's just a reminder— I do want to— but it's a reminder to me that no matter how much people talk about on social media about an orchestration layer, we haven't found one yet. It's, it's on our checklist. Um, but when I look at a lot of the GTM and other vendors, they're just building point solutions, right? No matter what anybody says. So as soon as we master the orchestration layer ourselves, we will share it with you. It is definitely a challenge for 2026, but it's one we haven't even realistically been able to tackle yet. But let's get the questions.

Yeah, it's funny, there are a few questions on why we haven't tried to build our own orchestration layer.

Well, how would— I mean, do all these agents have enough of an API or open system that we can manage? Can they all be managed through an API? We'd have to extract all the data through into cloud or wherever, our own Vibecoded app, and then we'd have to met— and we could get them all in some sort of unified system, but they'd all have to be 100% manageable through their API.

Yeah. Which they're not all today.

Yeah. I don't think it's a priority for any of them. I don't think it matters whether it's MCP or anything there. If they're not exposing their entire— I mean, literally just going in every morning and seeing the exceptions and issues that come up from campaigns the night before. If that can't be fully replicated inside an orchestration layer with some management, it's useless. Like, if it's not as good, it's useless, right? So everything would have to be exposed. Um, it's a good question, and it's not— I'm not saying someone smarter than us couldn't figure it out, but we haven't even begun to figure that out. There's no, there's no obvious off-the-shelf-ish, even Vibe solution today. And, um, uh, platforms are— a lot of these agentic platforms are not as open as they look. Or as you might think, not as a, that's not fair. As, as, as much as you might think there's siloed solutions today.

Yeah.

Right. Yeah. The only thing that is open is Salesforce. It's where they all meet. And if Salesforce could be our, our truly our orchestration layer, that would be actually the simplest path. Right. But it's our management layer, but it's not really a truly an agentic orchestration layer yet.

Right. Yeah. Not yet. Um, let me go back through the older question. Someone asked clarifying questions when you've been talking about your agents and apps. Are you talking about Claude Code, Cowork, Open— we don't use OpenClaw, but OpenClaw, or agents inside of SaaS vendors? I feel like we've touched this every week now. People keep asking why we don't use OpenClaw. I'm like, we don't need it. Right now. So, um, but yeah, you can go to, if you missed out the top of call, you go to saster.ai/agents. You can see all of our agents there. Um, we did a deep dive last week on our build versus buy rule, which is 90/10, um, 90/10 to buy, sort of 90, 90 to buy, 10% to build. Um, and we explained why we build things versus sometimes buy them off the shelf. But most of this is third party slash off the shelf stuff that you can also use/replicate some of the internal stuff you can likely replicate too with your favorite vibe coding app. But yeah, that's, that's our 90/10 rule that we wrote out last time.

Look, just on the OpenClaw thing, I mean, we could laugh about it, right? I mean, I tried it early. The, the reason we don't need OpenClaw, like the challenge we need solved is we don't need our, our agents to be more autonomous. We don't need help running them more often. Like the, at the end of the day, what OpenClaw is clever at is helping you run Claude and other things around the clock when it doesn't naturally do so. It's reasonably clever for that. We don't— we already have our agents who are producing more output than we can keep up with as they're configured to run autonomously by orchestration. We really need them to be unified. We need— we don't need— I'm not even sure we need an AI orchestrating our 20 agents. We need a single interface where the humans meet with the AIs. That's what we really need, right? And maybe that needs some automation. That would be nice, but they're already running on their own. These are not isolated agents that we need OpenClaw to have a heartbeat file to get them going again. They're off to the races. It's just logging into 20 apps every day and dealing with the issues is a headache, right? So we need to simplify the human side and maybe even— let's keep going— maybe even orchestration is the wrong term. We need unification. That's sort of what we need.

Yep. Yep. Um, maybe we'll build one after annual. We'll see. It's only 90 days away. So maybe we'll see. Um, but not before— again, I keep telling myself I'm not going to add another agent or do this thing and then I do it. Um, but likely not before then since SaaStr annual is right around the corner. Um, another question in the chat from Alex. Any other thoughts on scaling past one person to manage all of your agents? Jason M. Lemkin.

Ask the question again.

Yeah, we kind of touched on it earlier, but any additional thoughts on how you scale past 1, 1/2 people managing all the agents?

How do we scale?

In what sense? Past 1 or 2 people managing all the agents.

I don't know. You would know better than me. I don't think it's that complicated. I think that, um, I think that you, I mean, listen, there's obviously conflicts in some ways. A lot of the start of this conversation was about conflicts, potential conflicts between agents and making sure they don't conflict. But if you have two folks on your team that work well together, two humans, they'll just divide and conquer.

Yeah.

I mean, we are, our agents are our go-to, our sales agents are roughly as, as Amelia alluded to, and we've talked about before, are actually fairly clearly segmented. There's, there's raw outbound, right? There's, there's, um, There's inbound, um, there's re-energizing lapse leads, right? There's lapse leads. And, um, I might, I might be missing the fourth one. I'm just have different people manage these different campaigns and have them coordinate every day, just like the agents coordinate every day. That would be fine. If we could find a mini Amelia or a maxi Amelia, um, we would hire him or her tomorrow as a human, like no problem. And they would just divide and conquer. I don't think that is an issue with the agents. No, it actually might let us do more, right? We probably could do a little bit more.

Oh, for sure. Yeah. Yeah. Yeah. I think the issue I've run into is like, I, one, I was like, I'm not going to try and solve this problem before the annual. Um, but two, I, I think the problem I run into is like, I can't find folks that know more about the agents than I do, um, at this very moment in time. Right.

Well, that's, that's, that's, that's a big, I think what you need is someone that not, that for real wants to learn and is a good nerd about tools.

Yeah.

You need someone that has already built— like, listen, everyone should be screening higher is better. You should have someone— someone should come into the interview and have built an agent in Replit or Claude Code or does something. That should be the test.

Yeah.

That should be— in fact, what they do at Replit itself now is they give you a massive amount of credits. I think they give you like $1,000 worth of credits in the interview and they say, build anything you want with $1,000. Come back to us and then we'll do the interview.

Yep.

It's the best test in the world. You should do— if we could— if they don't need to be better than you, but they need to pass that test, right? Here's whatever you need. Just, just show me something you've built that automates GTM or agents that is better than we can. And you know, if you're not too crazy, we'd probably hire you tomorrow. Even if you could be pretty crazy, we'd still hire you. Right. It'd be fine.

Yeah. Yeah. I think that's a good test. Um, someone else about asked about our stack, which we answered. Um, another question is how do you calculate your agent ROI? We've covered this on previous, uh, chats before, but just at a glance, um, since most of these are go-to-market agents, we, we calculate ROI directly in, in meetings booked. Um, but actually meetings booked doesn't even really matter that much. I, I count it more so as, um, closed one revenue. That is literally our, our very top line agent ROI across the board, across all these agents. Is ROI, like true ROI and revenue.

I'd say there's just two things. It is very easy to attribute revenue to each agent. It's very simple.

Yeah, it's super easy.

In fact, there's rarely even a multi-touch issue at the Agentek level. There may be multi-touch at a market— at a marketing level. It's very easy. Which agent interacted with which lead and did that— did that lead close? Like, and then we attribute revenue to them. That's the main way we measure ROI. The reality is Having said that, it's not why we started. We started because we had people that didn't want to do these jobs and we said, can we replace a 6-figure hire with a 5-figure agent? So the truth is, even if these agents performed a little bit worse than they did, they would still be ROI positive on that, on that goal, right? Our goal was basically to achieve the same amount of revenue with half the humans and the agents. That was the way we initially measured ROI. And honestly, you know, you need to deliver them. The one thing I'll say to folks in general, if they're scratching their heads and why is it so hard, I can tell you some of the products that we use, most of the ones we use, they have infinite demand. Like Monaco, we just rolled it out for outbound that we talked about. They're booked for 2 months, including weekends for demos, okay? There is one called Basis, which is an agentic, a product for accounting yesterday. I don't know the team, but they just raised $1 billion yesterday. And they said in their header, one of the challenges is like their demand is simply overwhelming. Like they can't serve the demand. Now, some of that is because you have to train these agents and you have to onboard them and you can't dump it on the customer. So there's a reason for it, but that's how we measure ROI. But if you're a vendor, if you're working in a startup, your ROI needs to be massive in this world. You're gonna be slaughtered by folks that deliver like a 6-figure meeting on day 1 like Monaco did, right? Or take over all of like 2 people did for like each of Artisan and Monaco and Qualified. They replaced like 2, 2 people each, right? That's high ROI. Mediocre ROI is dead. It's just dead. Products that just give you productivity bounces, no one's gonna buy that stuff today. So that's why I challenge folks out there If you're stalled out, if your business is struggling, if people aren't lining up out the door to buy your agent, your agent is not good enough. Anyone that has this massive ROI, they cannot service the demand. They literally cannot service the demand, at least not all the demand. And in fact, what a lot of the vendors that we highlight here, if nothing else, they ruthlessly screen out prospects that are not a good ICP fit because they don't even want to waste their time, anybody's time. They don't want them to churn. They don't want them to bounce. If you don't have enough data, if you're too small, if you're the wrong org, some of the folks we talked about, just, just, they just, it may, it frustrates people. We get emails all the time. Hey, can you get me a meeting with this vendor? I mean, we can, we will. Um, but, but sometimes they're making the right decision. They're so overwhelmed with demand. They've got to focus on their core ICP. So that should be you. I know it's tough to hear. But it should be you. And if it is, if you have this agent where the ROI is so insane, you can generate new business for them the first day or the first week, or you can replace multiple humans with an agent reliably in today's world right now. This may not last in today's world. You will have demand out the door, out the door. And if you're not, don't delude yourself that you just have, that your AI is good enough. It is not.

Yep. Okay. I'll try and hit one or two last questions. Another question from the chat was, what's preventing you guys from letting the agents do more of the work, i.e., emailing on your behalf, updating LinkedIn, letting it do the LinkedIn ads, etc., since it's roasting you on that? Um, I'll give you my quick take, is it can't do it. Like, I would let it do it if it could. Um, these agents can't do that. I have not seen an agent successfully do email marketing at scale. The moment I find one, I'm using it that actually works. I have not found one that actually works. Like, not just writing— all of these just write email copy. They don't actually send it to the right contacts at the right time. They just don't do it. It'll come.

It'll come. Listen, some folks are going to challenge us and say, hey, my startup does it. Listen, we will— we look at everything that's out there. The reality is You know, marketing is behind sales. It's just a reality. Honestly, if you squint at the apps we use, our stack, half of them really are marketing automation, but for sales, right? Go back to the old days, Salesloft, Outreach, companies like this, they were, they are marketing tools just designed for the sales team instead of the marketer, right? That's, these are just those tools on steroids with AI. They're just ahead. It is embarrassing that we do not have a fully automated HubSpot/Marketo today. There's no excuse. It's, it's friggin' 2026. Like, and, and marketing actually is some ways easier problems to solve than sales, right? Instead, we're living, we're living in the past. And even I get, I get folks reach out to me and what they're doing in marketing is very narrow. Like, I can automate a little piece of your LinkedIn ad and maybe we'll add that, right? We haven't historically done that, but I don't need a, I don't need a landing page, right? Or like a little tight optimize my Instagram. I need a, that's why we built our own VP of marketing. I wish we hadn't. So, We are all over this. I hope that in the back half of this year when we do these AI Lives, it's actually all about marketing. That's what I'm optimistic about. But we don't need any, like, we need the kind of ROI in marketing that we see in sales. Otherwise, it's just, it's not worth it. They're not better enough, right? But we would love to, we would love to, this, and there's no reason it can't be automated with AI. There's no reason. There's no reason that you cannot ship better emails than a human can create. There's no reason AI can't take your past campaigns and update them with new visuals, new, new video, new content, and send them out. There's no reason AI can't segment your base better than a human can segment your base. There's just no excuse. There's no excuse to not run programmatic advertising for B2B, not for B2C, without you having to do anything. There's no— like, that all should be doable today. And people are going to say, we do it, we do it. We'll keep looking, friends. But if you can do all of this, then like, do what these leaders have set it all up for us. On their own and show us that it works and we'll, we'll be on your homepage, but it's gotta do all of it. I don't need a landing page or an infographic. It's not enough.

It's not enough. Um, one last question that I'll hit and then we'll wrap is, cause I think it's a good one. Um, question from James in the chat. You guys have talked a lot about your data, either pushing or pulling from Salesforce. Do you have one unified data model that you maintain, um, and that all the agents access in Salesforce, or do you have multiple data models supporting different agents depending on the rules that Jason outlined? So I'll give you my answer and then we can get your take on this, Jason, is we do have one data model that we push back or pull from Salesforce. Most of the time, again, these agents collect too much data. Like once you start to have multiple agents as well, I think what you'll start to find is it's too much data. It's too much data for humans to act on. It's too much data for agents to act on. Like sometimes too much is too much. Like we have purposefully told our Salesforce and our agents to only ingest certain things. Um, so that it's not like, cuz like it literally knows when you're on the website with qualified, like it knows what pages you looked at, knows what pages you looked at on salesforce.com. Like it can pull that into a conversation. That we have with you, either for tickets or sponsorships or whatever it may be. Like, it knows all that. And to some degree, that is useful context. But some other things I don't really— like, it doesn't need to know. Like, it doesn't need to know that Bobby used to work at your company and he came to Saastr 5 years ago, but now Bobby's left. Like, does that really matter? It's a new sale basically at that point. So that's too much context. Like, there is such a thing as, I think, just too much slash irrelevant context that you can get into trouble with. Once you have all these agents producing so much output, right, there's so much they could in theory tell each other. But I think you have to make the decision of what do they actually need to serve your outcome, whether that's closing the deal, buying it, whatever it is, you know, retention as a customer. Like they may not need to know all that. Like some of that might be too much for them as a human to process on the other end of it. Oh, you're muted, Jason.

Yeah, for sure. I mean, certainly, cell sources are source of truth, right, for, for, for these apps. Yeah, you will quickly discover it's true with everything in AI is you need a certain amount of data for it to be great, right? And then, but then any more than that is too much. It's also like memory, like you think you need to your memory to have everything in ChatGPT or Claude. It turns out it doesn't, right? We probably need more memory than both of them have on the consumer apps, but you only need so much for, your campaigns, your support to be great, right? So you'll learn, you'll quickly have too much, too much data. The one thing I would just add, it's just interesting, it's minor, is, you know, our source of truth used to be Marketo. Marketo is a very dated marketing application owned by Adobe. We shouldn't be on it. Even from the old days, we should have been on HubSpot or another solution. But we had someone that briefly worked on us that knew Marketo and brought it in, and we're kind of stuck We won't be stuck on it forever, but we've been stuck on it because it's highly segmented, has massive data. Because that's not really an agentic product and because our agents don't really hook into it. I mean, Amelia, you're the boss. It is dying as our source of truth and it is atrophying to a very low-end basic marketing automation tool that we could probably replace with anything off the shelf, right? So because it used to have all our data, because it sent all the campaigns, it was better than Salesforce. But it is very interesting that the amount of data we have in Salesforce has probably gone up 2 orders of magnitude and our marketing automation solution has died because of it for us.

Yep.

Yeah.

It's literally to the point where like, I think our renewal for Marketo in particular is like right after SaaStr and I was like, I probably don't need it. Like, honestly, like it's to your point, it does nothing. And to our 90/10 rule last week, I'm like, it's, it's terrible for what it is. Like, it's, yeah, I agree with you. I, I stayed on it for so long only because it had so much of our data for the past 10 years. But now I'm like, Actually, any of the good tools can just ingest that in like a day. Like, and then to my point, it doesn't need all that. Like, it only needs a few things that you've probably recently done. It doesn't need all historics for 10 years. Like, for the founders who are part of SaaS, you're like, those people haven't moved on. But a lot of the GTM folks have moved companies probably 5 times by now in the last 8 years. And like, we have all that history in Marketo. Do I need that? No, I just probably, I just need their most recent thing that they've done with us. To give that context back to them when we wanna pitch them for something at SaaS. Or like, I just actually don't even need all that data anymore. Um, so it's become less useful to us.

Like, I don't, uh, I mean, I guess as we would leave, we, we will pro— when we have, if we have the time after SaaS Annual in May, we'll probably find a marketing, an agentic next generation marketing automation tool that integrates just well with Salesforce. Just the data in Salesforce and takes our data from Salesforce agentically and creates campaigns, right? And if we can find someone to do that, forget about Marketo. I don't think we'd want HubSpot either unless HubSpot does it. We don't need these dated marketing automation platforms because Salesforce has been re-energized as our— I mean, you can find your own source of truth. It could be in Snowflake or Databricks or on your C drive now with OpenClaw or whatever you want. But it will become our super hub because it is our source. It is our data lake and source of truth. right? But it could be your others, but we don't need 3. We don't need Marketo. We don't need a Marketo and a HubSpot source of truth. And granted, HubSpot does integrate sales and marketing, but we wouldn't— for years we wanted to move our marketing automation from Marketo to HubSpot because it was more modern for years, but we would never do that today because it's not enough.

Right.

It won't be our source of truth for marketing. Like we don't, we don't need that to be a source of truth. We need something that can agentically send emails and maybe a few texts from our Salesforce data. So that'll be a fun journey to be on the back half of the year. We'll do it all together.

Yeah, we'll do it together. Okay. I think that's— I know there's other questions in the chat. We'll do this. We're trying to do these every other week. So you can send it to me in the meantime and we'll answer it next time. In the meantime, we are, I would say, I don't know, something like 70-ish days away from SaaS Re:Invent. So, We'll hopefully see you guys there. I'd be remiss if I didn't tell you that ticket prices go up on Saturday because Saturday is the last day of the month. Um, so get your tickets now before they jump up because, um, why not do that and take advantage of that? Um, we will talk a lot more about our agents there. We've got some of the best of breed AI native apps coming. Their CEOs are coming to speak. Um, so it'll be a lot we're doing to— And we'll show you—

they'll have everyone there to teach you how to do this stuff. To.

Yes, for real.

I love the CEOs of Replit and Lovable and the CEO of Vercel and all these great people and, you know, CTO of Databricks. It's all great, but we'll have people hands-on showing you how to do these agents, answering your questions. You can pepper them until they fall down exhausted because that— in fact, our whole first day is deploy and it's— I mean, it'll be all the— but the whole point of the first day is to help you deploy these agents yourself if you haven't or to fix your issues. That's the whole point. So It'll be great.

Yep.

See you guys in there.

We'll go, we'll go deeper there and then yeah, we'll see you on here next time. And, uh, thanks everybody for joining.