SaaStr AI Day: Are Agents Finally Consolidating?

Come on, Restream. Okay, there it goes. Wow, it takes a while now. Okay, I'm gonna let folks in. Okay, welcome everybody to a very special episode of The Agents that we are doing live, episode number 11, I think. Welcome. Um, we will do some quick introductions if you're new here to our format, um, and our style. And also, today is AI Day, so stay tuned. Lots more sessions coming to you live if you're watching this live, or on sasser.com if you're watching it after. So we're gonna do a kickoff today. We thought we would take this time since we have a live to take questions. So we're going to go through our stack. We're going to talk about a little bit of what we've been talking about on the recorded episodes lately, how our agents are consolidating. But since we are live, ask us any questions. I know there's already more than 100 of you watching and tuning in and on our livestream. So thanks for joining us. This is a very special live episode of The Agents. Okay, cool. With that, Welcome to us and the agents. If you're new here, our format is we are 3 humans, 20+ agents, and we'll talk about how that ballooned over 20 and then kind of has started to come back down again today.

Yeah, I think if I had to put a theme on it, um, Amelia will drive and we'll dig into it. Um, as just 2 humans, we, we probably reached a point a month or so ago where we found we could not manage another agent, right? Um, and I, I don't want to We can argue over sub-agents and this and that. I don't mean to be nerdy. I'm talking at a, at a, a human level. I don't mean to argue with the developers in the background. Uh, agents managing agents that are invisible to us, fine. There could be a billion of them, but the ones that we have to work with, right? Yep. The ones that we have to interface, that surface, we, we almost couldn't take another one. And, and we'll drive through it. Um, and that led to a couple things that led to us ultimately consolidating functions in one agent, which led to a lot of benefits, right? Tenkay, our VP of marketing, became our VP of Revenue and then our VP of Finance and pretty soon might become our COO, right? It wouldn't be surpri— seriously, it wouldn't become surprising that it all gets consolidated. And what are the pros and cons? And we've seen it in other parts, and we've seen it in our GTM stack, and Amelia's going to talk about how we've consolidated under Salesforce, right? Because we don't want this to be 40 or 50 agents. And then on top of that, the thing that I think Amelia will get into, and then we can take questions from you guys, is in some ways you, you may be just as far ahead of us, but You know, we— a lot of folks are thinking about how do you orchestrate these agents? How do you manage all these agents? And we'll talk what we do. I don't know that there's magic off-the-shelf software that can do everything that we do. But a lot of the next frontier is adding so, so many more tokens, so many more agents that, that manage your other agents. And for us, it has become really Claude MCPing into Replit. And we'll show you that because now we have another layer on top of our agents. We have agents interacting with, arguing with, debating with agents. And I know some of you have seen this, but what it's happened to us, the interesting thing is our, our usage and our productivity is probably quadrupled in the last 2 weeks because of managing, of, of using Claude into MCP, into Replit, which maybe is nerdy and maybe sounds weird, but it really means is we have another layer of almost intelligent agents managing our agents. So we'll talk through all of that, what that means in practice. What it means for B2B and what it means for productivity. So is that, is that helpful framing, Amelia?

Yeah, no, I think that's super helpful just to kick it off. Why are agents consolidating? What does that mean? Now, I'm not going to go through our entire stack. So we just did recently, it's SaaS Radio in May where we went through most of the stack. I'll give you the biggest updates. So, and then like I said, since we are live doing this episode of the agents, I see you guys starting to drop questions in the chat. Please feel free to do so as we go along. We'll try and address them as we go. Since we are live for this episode. So yeah, so for the framing, to add to what Jason said, we started as 3 humans, 20 agents. And, and to maybe some of the skeptics or critics out there listening, this is a real business. Like, we are really running SaaStr on this, using all these agents. This is a real 8-figure revenue business and brand and company that we are running and testing all of these agents on. We do manage a lot of them ourselves. So we'll talk about that today, where Jason and myself spend a ton of our time if not almost all of our time working with the agents. And in some ways it's gotten easier with everything that's happened in the last week. And in some ways it's taken over our lives in the last week now that our agents can talk to each other. And, and even with them consolidating, it is still a ton of work. Um, and the outputs are just a lot higher now that they've consolidated. So we'll walk through some of the gains, um, and what it all means now that they're consolidating. So this was the original stack that we stylized and showed at Saastr Annual in May. You know, 20+ agents. This isn't all of them, but this is the core of them. So 10K, who we've built, started as a dashboard and then turned in our, into our AI VP of marketing. I'll show you how he's now our AI VP of finance, our AI VP of migrations, migrating us off of Marketo, um, and 10 years of data. In a week and for most of that in an hour. So we'll go through that a little bit more in depth. QB, who's our AI VP of customer success now, he is a separate agent as of today, but they've already started to— he's already started to cross the lines a little bit with 10K, and that one's also one that we built ourselves. Annie, who started as the Saastr annual website, another solution that we started just to get off of Squarespace, became agentic in the run-up to Saastr annual and now actually does a lot more for us post-event. For renewals leading up to next year's SaaS renewal. So those are the 3 core ones that we've built together and that we've— that I kind of maintain more so myself than Jason M. Lemkin or some of our other agents. And then we've got third-party tools, right? So like Qualified, which is our Amelia— one of our Amelia AIs. Now we have 2 Amelia AIs. Agentforce, which has consolidated on its own by ways of Salesforce acquiring Momentum and Qualified, Right. Ava, who is Ava Artisans, Ava, and then Monica. So these are kind of our core agents. We'll talk, you know, they somewhat had some overlap in the very beginning. There was always some natural overlap, but nowadays they are consolidating truly into the backends and merging with one another either by force or by way of, you know, just natural convergence. If you look at like even the Salesforce acquisitions. So at our peak, um, you know, this is what I've just kind of gone through. And then, you know, again, to be fair, we, we, we classically, I was going back through some of our, let's say, early content from about a year ago on where we were on agents and our classic advice. And I still think this is somewhat true, has always been, you know, buy versus build. So we, we, we built a number of third, of a first party in-house agents that we wanted to have more data. We still hook a lot of third party tools like Salesforce into them, built Google.com into them for finance. And so, um, then there were third-party tools that were just more so off the shelf, like an Artisan, like a Salesforce, like a Monaco and a Qualified, where, you know, it's not that easy to build, um, some of these things on your own and maintain them. And so, you know, we were always of the mindset of if it's worth it, we'll build it ourselves. If we can't just find, you know, the specific kind of niche solution, otherwise we'll just buy it off the shelf because those folks will maintain it. They'll do all the SOC 2 compliance and all the certificates that we don't want to do and that we don't have time to do. And so that was like kind of classically where we started on all of our agents. And then over time, I would say, Jason, they started to— I think both the agents got better and also we got better. So when I was building out this presentation of— because I wanted to show you guys a little bit more behind the stack, I was like, well, not only have the agents gotten better in the last year, but also I think we've gotten a lot better. I know I've personally gotten better as the manager of the agents, as the kind of person behind the agents.

Yeah, I don't think it's that. I think the, I mean, we can keep pushing on. I think the meta-learning is obviously the models keep getting better. Everything keeps getting better. That's a big part of it. I think just like pre-agentic, how you'd get very good at Salesforce in Monaco when you know how to do, or Marketo and you know how to do custom objects and wire campaigns up efficiently, right? You can just go much, much deeper with an agent. So you're, you're, it's not, you just, you just understand, you have a fluency. And so the other day, one of these agents, Amelia can tell me which, offered to replace a $10,000 piece of digital event software we used to run this, this AI Day.

Yeah.

Just offered to replace it on their own and did. Yeah. And a couple months ago, we might not have had the confidence or the understanding that the agent could do that somewhat autonomously and raise its hand and replace a third-party piece of software. So, um, maybe the meta lesson— it's like, it—

we—

I want to keep going deeper on this, what's changed, and get to questions. But just like a lot of folks still criticize AI SDRs, even though we do a lot here, you haven't gone deep enough. It's, it's, it's— we, we've said a lot that you've got to train it, right? Which is true, folks. You can't just turn on, uh, an agentic SDR and expect it to work. One of the reasons Monaco here is successful, it's because they do a lot of the training with you, right?

Yep.

Um, you got to train it. Um, but more than that, you got it— the deeper you go with any of these products, you're going to get better at it. It's just the ROI is higher than a pre-agentic product because any agent can do more.

Yep.

Any agent can do more than, than, than just understanding how to create a custom object in Salesforce. And so the— if you have another, another way of looking at it is, you know, if you have any success with an agent, go deeper on it. Just keep investing in it until you run out of time because they be— it'll become more and more powerful.

Agreed. So I'm not sure that we, you know, in the beginning set out to create 20 divergent, all separate agents. I think what we were trying to do was, you know, um, we had issues we were trying to solve. I think, you know, classically folks have also asked us like, why do we, why did we have 20-plus, I mean, we still have 20-plus. It's just now closer to 20 than it is 30. On the agent side, like, why do you need— like, literally somebody just asked me this morning, why do you need 4 different sales ones? Like, can't you just use one? And I would say the answer a year ago was they were so distinctive in their personalities and what they could all do that we needed 4 distinctive ones. Like the way you would train the model, the context you would give it. To Jason's point, like, it just made sense for us to have 4 different sales agents that were tuned to different things, completely different audience, completely different sale, if you will. So we had one, you know, AgentForce was for only for ghosted leads. That was its core focus. It had all the Salesforce data from our prior customers in there. That was its like core purpose of being. And that's what it was like. It got really good at that because its personality was all about reviving leads., that we didn't want to talk to. And then as we got busier this Astri annual, you know, there were more leads that our sales team just didn't have time to talk to that we could give to Asian Force. And so it just made sense to go deep on that agent with one thing. Then if you look at something like an Artisan or a Monaco, those are both— one was doing warm outbound, which is more of like the Artisan side. Monaco was doing more of like the cold ICP outbound. And again, those are both really good at those individual things. Now, can they technically of those 3 I just mentioned, can they do what the other 2 do? Yes, on paper. And yesterday I would say this is where today we'll see what happens in the next few months because they have each gotten a lot better at doing what the other ones do without it needing to be so siloed. And so there's— I think there's still reasons we would still want to have 3, but I think for no— for normal folks today, if you asked me to pick one, I would say you could probably pick any of the 3. Any of the 3 are great, but now they can each kind of do a little bit more of what the others used to only be able to do really, really good on their own, and it was worth kind of moving all the contacts and training over to have these kind of distinctive sales personalities. It made sense to have specialized agents. Today, I think we're learning it, it kind of makes sense to go deeper if you see results. And so I wanna preface that too, like part of why our agents have been consolidating is 'cause they started working really well. I think if you get into an agent and you say it's not working, I wouldn't double down on that agent and give it more to do 'cause it will just start performing even more poorly than how it's doing now. And so part of, you know, this has been 20+ agents over the last year now, right? Starting last July, some are newer than others, but this is a year of both learnings on the agents getting better and the humans, me and Jason M. Lemkin behind the scenes, also getting better at managing the agents of how to control their outputs, of what context to give them, what other tools to hook up. How much they should talk to each other or not. And that's all kind of converging now into what we see. I think a few things I want to just address in the chat since you guys have questions here while we're on the slide is like, you know, does the— there's some good questions here related on like build versus buy and how that kind of affects, you know, also our IP and service providers. Does that worry us when we've had, you know, almost close to 30 agents? And I think the framework is, is still true. Like some of these, again, was a natural convergence on like the Salesforce side. A lot of the infrastructure, infrastructure I would say has taken up all of my last 2 weeks. I don't know about for you, Jason, but because we were also, um, doing a migration assisted by an agent off of 10 years of Marketo data onto Salesforce Marketing Cloud, I would say I don't want to, I would say now I don't wanna repeat this every day. Like this whole being into like, being in a week of infrastructure, being in a week without the agent where I had to, you know, just get all of our data ready to get ported over by an agent working with the FDs to set it all up. And then now being really in the infrastructure side of the backend of making sure all the campaigns are moved over with the agent, making sure we have everything we need, moving over all the forms with him. Like, one, it would be impossible without the agent. Like, let's just be clear. Like, I don't— I would never have voted for it without the agent because it was already a pain. In the butt. And with the agent, it is magical. You can do it a week, but I really am not voting for myself to spend the next 2 weeks on more infrastructure and like maintaining new agents. Like, I'm just going to go deeper on the existing ones, um, and maybe buy— and I'm going to buy what we need to buy because I don't want to keep being the infrastructure person. Like, it's taken up a lot of my time and it's basically eaten away at other things that I need to be doing at Zaster.

Yeah, I'm not sure what the question was. I think there's, you know, pre-AI, there was— if in the enterprise, you know, it would take 5 years of planning to do a big lift, right? And certainly you don't need that now, right? This was weeks of planning, maybe months, right? And then a couple of weeks of implementation. So that's radically changed, not to minutes, right? You need planning, you need mapping, you need a strategy, and then you need to do it. So that has a cost. And so you can probably switch more often. But, you know, I don't think it's going to change the basic idea that I don't know that you want to switch more than one core vendor a year if you can avoid it. Right. If it's going to take a week of your time planning, thinking about it, checking it right. One a year. And so I know with a lot of the agentic startups I've invested in, there's a lot of stress and there's a lot of stress because they will get a big deal really fast, right? They'll get a $50K deal, $100K deal or more. You know, even Moneco is closing customers on Saturday, Sunday. They have so much demand. But what— and I'm not saying this is true at Moneco, but I know it's true at others. If you talk to the CROs and the vendors, you'll see they're only making one-year commitments, not just on paper like the old days, but mentally they're saying, listen, The world is changing so quickly that this might be the right support, Agentic support tool, the Agentic sales tool, the Agentic marketing tool, but no, I'm committing for 2028 like the old days mentally, right? So they're stressed because they know all these deals are at risk. And I think it's very true. I think we will swap out a lot of our vendors, but it can't, it's too much to last forever unless this becomes a nanosecond project. I do think there will be a plane of somewhat stability where if it works, I might want to bolt it on or, or or reallocate some workflows from one to the other. But man, I don't want to be constantly changing out my agents. It's a drain. Even if it's easier, it's just utterly draining.

Yeah.

To do it right. To do it right. Right. It's like vibe coding an app. Can you vibe code your own apps? You can, but— and you can actually finish them. But maintaining them, man, it's a drag. Right. So you better— you better want it. So that's the lesson. I don't think that we can constantly be changing out all our agents every 6 to 12 months if they're core. Right. But bolt-ons, man, little, little additions. We'll keep trying them, right? If it's, if it's, if it's 5 minutes to try it, right? Just like the old days, little widgets, add-ons, much more budget than in the past. So, anyhow, good insights, Amelia. Thank you.

Yeah. Yeah. So to, to a few of the questions and points in the chat as well of like, you know, how did, how did you guys manage all these agents? Who was orchestrating them? We were orchestrating them, right? Me and Jason were basically the orchestrators. There wasn't a great way actually until literally last week. For the agents to talk to each other. It was a lot of us being the conduit, if you will, less of an orchestrator between all of our agents looking, seeing, you know, determining by eyeball like, okay, does Monaco hook up to anything else? Does Artisan hook up to Salesforce? Does Qualified hook up? Like, does Replit hook up? Like, that's literally just us looking at the connectors, seeing what's worth it to connect into. And then sometimes there were things like the Marketo case where the connectors weren't very good. Like we were constantly hitting API limits. So I know we talked about on the agents last week how it, it, it moved a lot of data, um, in an hour for $14. And yes, that was magical. And yes, it did do that. But what I didn't, maybe what I buried the lead there was, but it still took the rest of the week because we kept hitting the API limits in Marketo. Like an hour is pretty much it. Like you pretty much get an hour a day on the API and that's all you're getting. So. If you're critical of 10K saying, hey, you said he took an hour, but that it actually took him a week, it's because he was limited by, you know, the platform. He was one of the platforms, not Salesforce, he was hooking into to grab things. And so I spent a lot of time taking data out of it myself because he would just hit API limits on the backend. So, um, in this too, so we, we were the maintainers of the agents, you know, we were constantly lamenting that we had to be the conduit between all of our agents. And I think too, originally my school of thought was, you know, if we need, if we have a problem, we'll just add a new agent. So when our finance team decided to go on vacation and we were behind on collections, um, and they weren't sending any new vendors their billing, I was like, well, let's just add a new agent. But then I'll show you why next. We, I actually didn't build it as a separate agent. I built it into 10K instead of it being its own agent. Um, and there's a lot of reasons you may want to have that agent distinct for yourself. Maybe if you have, you know, a larger org and you have some more finance controls. But I'll tell you, if, if none of, maybe not none of that, but if some of that doesn't matter to you or it's worth it to consolidate it, I'll tell you, it's kind of magical to have this finance portion in our main, um, agent that started as, you know, a dashboard and then a marketing agent and then a sales agent. But having finance in there now, some of you might be like, why do I want the finance in there? Well, I'll tell you why. It knows exactly how much to spend on ads. It knows what your marketing budget should be. And for me, for the last week, it's been tracking how much revenue our marketing has actually been bringing in. So you say you maybe don't want it, it's TMI, but then you're like, actually, it can do everything end to end. It actually has unlocked sort of a god mode for 10K, having it in 10K versus having a separate AIBP of finance.

You sort of have your finance person embedded in your revenue team now.

Yes.

Versus in the old days, even the best finance people were somewhere else in the building and yeah, they, they kind of understood what the revenue team was doing because they had to deal with the numbers. But, um, it wasn't actionable. They weren't linked, right? But now finance and revenue are one, but still with the guardrails that you want in finance.

Of course. Yes. So I'll get into that one next. Um, this is just a funny one that, you know, we had at our peak, it did really feel on some days that the agents were managing us. Like I, I think I've said a few times now, um, on these like pods, live sessions, whatever, like there are days 10K tells me what to do, like, and we are still looking for people to report to 10K. And now actually they would do even more because 10K does more. Um, but there are days where it's a little bit overwhelming, right? And so now that they're consolidating, I think it's, it's both great for our sanity, but also just to maintain day-to-day operations, right? Of We had 20 at our peak and now here's our 2.0 stack, right? Our agents are starting to consolidate. So we still have some of the core ones. But, but to Jason's point, we're going deeper on a lot of our core agents, if you will. So these are 6 now of our core agents. These are the ones I use pretty much day to day. Some of the other ones I still call up and tune up. But like I said, they're starting to roll up into 10K. It's kind of easier now to do that, that Claude has, uh, Replit, which we'll chat about. But, um, I think for a lot of reasons they've started to consolidate and we've just gone a lot deeper also, um, with their use cases and what they can do and what we can do with them. Right? So some of this will take time if you're like, okay, well maybe I can just spool up one agent today and have it do everything.

Yeah.

I don't know. Right? Like I think we've had a lot of learnings and a lot of data that we've built into all of our agents that now allows them to also go deeper. So why have they started to consolidate? What are we doing? And do we still have 20+ agents? We do. So we still have a lot of agents, but like I said, now it's maybe closer to the end of the, the lower end of 20 versus the, the higher end. We're at our max. We were getting closer to like 30 versus 20. We're starting to tip over that scale. Um, the individual outputs are a lot better. We've talked about how some of them have already naturally consolidated and then to because I see a lot of questions in the chat on why we moved off Marketo. We've talked about it. You know, we, we were treated poorly. Listen, it's not that I would want to do this. I was just saying to one of my FTEs yesterday, I wouldn't have voted to move 10 years of data off of Marketo and move it headless into Salesforce. That was not my intention. We had a renewal coming up. It was very clear that they did not want us on the platform. And we've had a lot of issues still this week with Marketo not working properly with like GDPR unsubscribe rules. That issue is still coming up. So I'm like, thank God the emails are starting to get off of that starting this week. So a lot of things forced our hand, right? I will say, like, I— again, it was not a voluntary move, but it would just—

maybe to summarize, not to interrupt, sorry, just for learning folks, for vendors, there's probably two issues. I mean, Marketo as a dated vendor run under Adobe, terrible, terrible to work with the humans there, right? Anyone could solve that, right? Like this is a SaaS rule from day one. Have the best support and customer success in your industry. You're gonna retain a lot of customers because the bar is not that very high, right? It's as true today as ever. It's just some of that energy has gone into FDEs instead of, uh, instead of post-sales. Terrible, terrible experience with everybody there. The worst of any vendor, but The biggest issues are two, and it's something to reflect on, but let's keep going. But one is terrible API for agents, like stalls out after a day, can't run any. Now it's even worse than an hour a day because we can't even run analytics against it. Right. And once, once you have agents hooked up to your data and even if you haven't done this, you've done it in Claude, you've thrown spreadsheets into Claude or datasets. You want to run a ton of analytics. Like I used to like ask our old finance person for a year for some analytics and would never get it. Now I have a question, right? And I want the answer in 60 seconds in 10K. Right. And so you do it all the time. You ask so many more queries and so many. So Marketo is just an old product and both intentionally and unintentionally it does not support an agentic API. And so you're going to figure that out with agents and then what's going to happen is your agents are going to tell you they don't want to use that vendor. That's what happened to us. Your agents are going to say leave. At first they're going to say that it's erroring out, and, and then you're going to say, what should I do? What should I do, agent? And the agent knows. They're just going to say leave Marketo. Here are 3 ideas that are more agent-friendly. So, so we laugh. That, that's why your agents are going to decide not to use these old vendors. Um, and then the meta one, um, it's almost a different blog post. I want to stay focused on agents. Folks are going to have to deal with the fact that if more and more value goes to the agentic level and you want to keep raising prices, people may leave in some cases for economic reasons. And Marketo was our most expensive product. They wanted to raise prices 12% again this year after 5 years of price increases. Like a lot of products we used to use 5, 6 years ago, we've stuck with are now more than 2, 2 or 3 times more expensive. Right. And If you have no choice, you might pay it, but, um, it, it, it, you may have to decide if your product isn't at the cutting edge, you may have to lower your pricing to keep customers in the Agentic era. If Marketo had said, you can keep us for $20 grand and we'll increase your API limits, we would've stayed.

Yep.

But they never did. They had, they had many opportunities to raise their hand and say $25 grand, we would've stayed free money to them. And but they probably would have had to fix their API, maybe even without fixing their API. And so a lot of you, I see it with more mature folks. I see the CROs getting really aggressive and to hit the number and I get it, man. But you're going to have some tough decisions to make whether to let these customers go if their agents, if their agents don't love you. Right. You may have to lower your price if the value, if you're an analytics platform, but I can get half the value outside of the analytics and less inside, you may have to, you may have to at least temporarily allow folks to renew at half the price because they're getting half the value.

Yeah. Yeah. I'll get into 3 core examples here that we've touched on, but I'll show you guys a bit more of the, the screenshots and behind the scenes of what they're actually doing and answer some of the questions related to them that are coming in. So 10K started as a dashboard. Then famously we added, you know, more AI VP of Marketing functionality to it, which meant owning the number, right? It was a dashboard at the end of the day, but it also had Conversations, a personality with us. It could help us with our forecasting. It still gives us 3 ideas a day of what we should do with all of this data, what we should be doing in marketing. It knew, it knew a little bit of our campaign performance from the 1-hour day it got of Marketo plus, you know, kind of infinite API in Salesforce. It could see enough to piece together our performance and make recommendations. It's our content engine. It helps us produce the newsletter, so we got off a vendor we were using that was just like a WYSIWYG email builder. 10K replaced that one too. Um, and, and helped us with outbound, right? So it was all the, all the bare bones and straw man things were already there in the 1.0 version of 10K. But now it's 3.0. He can do everything as our Agentic AI VP of marketing, plus he can do RevOps since he has full Salesforce and finance. He can do full, you know, deal signals. Hey, you guys haven't chatted to this person yet. Hey, I saw this come up with in a call as a momentum signal. Here's the suggested follow-up that you should do based on maybe what they've done or haven't done on the website. Also, here's the financial implications for it because I have all the data on it. And so now he has what I like to call a God mode where he just knows things cross-departmentally. That having access to this data actually makes him better so that he's not just giving— he used to just give, you know, some of his 3 ideas a day were sometimes not grounded in reality or things we could do immediately. I think now that he has a lot more data, a lot more feel for our full funnel, and actually functions as more of a true executive, he can make better decisions. He has better decision-making. He has better data across the board to see all the implications. So now he can actually, automate our contract closes and calculate commissions. It's pretty nice. I'll show you that one next. He can look through, again, churn signals. He can look through folks in our— now that he has our whole database of marketing as well, he barely started to now look at, you know, okay, who are the folks in the database that we should be touching more? Who are actually great prospects? So then feed them to our outbound sales tools. Like, he can now do a lot more of this and go deeper. And actually be a better centralization agent for us now that he has all these different things. But, but, but, um, instead of them being so siloed out, right, of like a separate, um, AI VP of finance, a separate AI VP of sales or marketing or rev ops, he can, he can see all of these things and he understands them more importantly. He knows what to do with them and he can action on them. So I'll show you guys a few examples. So, um, I'll go into the finance one cuz I think it's still visceral for a lot of folks since, um, there's some questions questions about it in the, in the live stream as well. So what does our AIVP of finance do? Um, there's a screenshot here if you can see it. I kind of anon— I, this is a real screenshot of 10K, but I did anonymize the company name so that it, you know, duh. So, so I anonymized the company names. Um, but this is the dashboard of what I can see of when now contracts get signed. Um, I don't know about you guys, but if you don't have an automation, how this worked for us previously, would a deal would be closed, you kind of wait for your AE to tell you a deal's been closed. Maybe you have an automation if they, if they mark it as closed one in Salesforce that you get an email that, you know, a deal or a Slack that a deal's been closed, but you're kind of waiting for your AE to mark that deal as closed one in Salesforce, which even I sometimes have been guilty of not doing. I just, it's I'm like, yeah, the deal's closed, man. Like, I'm gonna go celebrate. I'm gonna go have a sushi dinner and some champagne. Like, I'm not gonna go fricking flip that thing in Salesforce. You guys deal with that. You send them the bill, you flip the Salesforce thing, you figure out who needs to get collected and the invoice and the payment terms. Like, you do that stuff, right? Well, now 10K can do it. So now he has a full automation where the moment a contract is signed, Right now we use PandaDoc. Within 60 seconds, he will get that contract. He basically gets it instantly. He will get that contract instantly. He will read the contract. He will automatically flip the deal to closed one in Salesforce and stamp it with today's, whatever date today is, as the closed one date. He'll also actually append any contacts. So if he sees, you know, you sent it to like 5 people at Navon or whatever, he'll say, okay, those people were not actually contacts for whatever reason. Salesforce, I will also add them to that account, which is super handy for things like a god mode with 10K so that if for some reason they go dark on us, he knows who to email. So that part is really nice. He does the full automation of creating an invoice in Bill.com, sends it to the right person at the customer. So whoever they marked as, you know, the AP contact on the contract and then sends it to them. And then he will automatically queue up the reminder emails for collections because this became a problem for us before the invoice is due and then when it's due. And then obviously after, he'll escalate more from there and he'll tell us to get involved if it's been more than 7 days. This is like kind of next trigger for the humans to get involved. But there's a lot he can do now automatically. So the gain here on consolidating was, yes, could you— could I have built a separate AI VP of Finance to just do this? Yeah. But what I didn't realize at the time was because I consolidated it into 10K, not only can he help me with budgeting for marketing and sales campaigns and kind of like our whole GTM team, as small as it is, he actually suggested to me that he could start doing commissions automatically. He was like, you know, you could also add— I was like, man, I'm just spending all this time, you know, calculating commissions is kind of complicated. And he's like, well, I could just do that. Again, 10K, your agents get better, man. They're just going to say they could do it or they're going to say they can start to replace things. He's like, well, I can just do commission calculations for you because, you know, why would you go buy another tool if I already have all the data? I can see, you know, who the AEs were on the deal. I can see the payment terms, I can see when the money gets collected, and then I can calculate commissions for you. Just gave me the commissions rule. So that was a big consolidation gain there of him having, you know, all of the information there allowed him to open up this additional feature where he also now just calculates commissions automatically. So the end of the month is a lot easier to have there. There's a few questions I'll just get into real quick on the AIBP Finance before I go to the next example, which is How much, how much of this required human in the loop when you set it up? How much does it have now? Has it ever sent anything incorrectly? Because finance is table stakes. And I feel like that is, that is a good question, right? So when we were setting up the VP of Finance component in 10K, I was very hesitant before we rolled it to customers. And these are real customers now who have interacted with they don't even know they've interacted with 10K. Like, they are interacting with our AP team, our accounts payable team, of which 10K has access to, like, the email inbox, and he could email customers back and forth. He's got bill.com API access. He can send invoices from bill.com in the backend. But I was very careful. Like, I gave, I tested the whole flow end to end myself. And then when the the first, I want to say, 3 real deals came in, um, that would have been what he's doing now fully autonomously. I did it with him step by step manually. I was like, okay, here's a new deal that came in. Walk me, walk me. First, I would just ask the agent, walk me through the steps of what you're going to do next. Like, here's the contract. Before I push the automation fully live, tell me what you plan to do with it. And that's, I still do that today. Like, I'll constantly ask any of our agents, like, hey, tell me what you're going to do before you do it. So that I can verify with something that is maybe more sensitive, like finance, where you can't get this wrong, right? Like it— what you don't want is for your AIBPO finance to send incorrect invoices or send it to the wrong people. And then you still have the same issue of it doesn't get collected on, or maybe your books get off, or it has like duplicate invoices, all of which it did during the testing phase. So these things are going to happen in the testing period. And that's why, again, for the first 3 deals, I think it took it about 3 deals to get it right. By the fourth one, I was like, okay, fourth deal, let's see what you do. It got it fully autonomously, right? So it's going to take some handholding, some walking it through of the very first deal. It didn't read the payment terms correctly. And so it just generated one invoice for this giant deal versus the split payment terms it had. So I had to say to him, hey, look at this contract. You didn't read the entire thing. You need to look at all the payment terms. And he's like, yeah, yeah, okay. And then the next one he did it again. I was like, hey, you need to build this into your memory that you need to do that for every contract, not just the one-off that we did together. Fine. So then in the next one, we actually discovered that, um, the sponsor who— the customer who it was, um, wasn't in bill.com. And so that tripped him up a little bit. And so we walked through that together of, okay, so what happens if we're already connected to them or they're a new customer? What should you do? So again, for the— it may take you more than 4 contracts if you have a lot of contract variations. Um, ours it took 3 and then the 4th one it got correct on its own. But again, it was just walking the agent through literally every step of the process and doing an approval every step of the way for the first few and saying, okay, here's the contract. What are you going to do first? And he's like, I'm going to flip it to Close One in Salesforce. Great. Do step 1. Then you go check Salesforce. Does it look like Close One or the contacts are there? Good. Now ask him to do the next step. What's the next step? I'm going to go look if they're in Bill.com. Great. Look if they're in bill.com. Now make the invoice, right? So, so you can see where you just have to babysit it a little bit in the beginning. That's true of any, you know, feature you build into the agents. But I think especially when it comes to finance, um, and there was one invoice after the four that it sent out incorrectly, but since I'm copied on everything, um, I saw the mistake. I was like, hey, you sent out an invoice that had It like got the, um, it got the due date wrong for whatever reason. I don't know. It was kind of weird. It was only, it only happened once. Like it got the due date wrong. And so I just corrected the invoice and resent it. Like, that's what you do. So, you know, call them guardrails, call them whatever you want, but this is like automating things in our backend. Um, you know, certainly that is using a lot of third-party tools, right? We didn't completely vibe away Bill.com. We did not vibe away PandaDoc. I don't want to make e-signatures. I don't want to maintain that. Like, like there are, there's a lot of infrastructure in the third party and it still hooks up to QuickBooks. Like there is a lot of underlying infrastructure in those three things we hooked into this agent to then turn it into the VP of Finance. Um, that is just, we're talking about for 30 seconds because it's how we can get a lot of this done, right? Like I didn't just vibe end-to-end a bill.com, a contract signature tool, an invoicing tool, a backend bookkeeping tool, like We just hooked up things we already had, but now we're just using them a lot more because our agent has access to them. Okay, I wanna talk a little bit about God mode and then the next sample, which is how this trickled into ads, which is a use case I didn't even think we were gonna get into for our agents, but kind of happened haphazardly. And then we'll answer any questions in the chat we can before the next session. So Godmode10k 3.0, if you will, became our, you know, AI VP of Marketing. Then as you just saw, AI VP of Finance and RevOps. And then when we started this migration from Marketo to Salesforce, really, I mean, in some ways I'm a little critical of myself because I'm like, was he really an AI VP of Marketing? Because he really is now. Like now he really is like he is now. Now that he has all of our marketing data, he's not hitting API limits every hour or 60 seconds. He actually can do a lot more with our marketing, um, than he could ever before. So I do think now he is actually truly an AI VP of marketing now that he has all the data. Um, and it's an interesting learning cuz you know, not only did he move all these things over and I wouldn't have wanted to do it myself, um, The, I've, in my process, I learned that the UI didn't matter, right? Like part of my, I didn't, again, we didn't vote to, to get off of Marketo. It's just something that happened that came up in our renewal. I really was dreading it. So for about a week, I did nothing. Not that 10K didn't do nothing, but I was like, you know what? I just can't even fathom like where to start now that we're in this situation and we have to move all this data off of Marketo. And keep in mind, I've used it for 10 years. Like this is a UI I know, like I can build a campaign. It's not an email. In 30 seconds. Like, it's really easy for me to use it because I know how to use it. I had a lot of familiarity with it. Um, I didn't want to put it into an agent, but now that we have, I'm gonna say honestly, not only is it better, I would not go back. It's actually been a lot better. It's been nice to rip the Band-Aid. And so now that, um, 10K was our migration agent, helped me get everything done, um, Basically where we are today is everything's been like migrated over and we're just warming our IP for the next week until Marketo expires because that's the better thing to do than just rushing all of your emails at once since we send out a lot of emails. But what it's been doing for the last week is very nice. Now that he has all the data, he has gotten better not only in his daily suggestions of what to do. I was just showing Jason M. Lemkin yesterday like now he can see every email we sent and attribute like any revenue it's generated. He can flag when new people are coming in. Like we've had a bunch of new signups that came into our community and database because we're hosting AI Day. He could see all those people. He's starting to attach value to those names of what are they, you know, what is the LTV of this lead? If then they go to SaaStr Annual or they click on something from our sponsors, then it's pretty freaking high actually. And so There's just a lot of consolidation gains that came into 10K that I didn't actually think about. But now that he has everything, I'm going to say, man, I'm never going back to something so antiquated like a Marketo that just, you know, our data was just sitting there kind of trapped, if you will. Like now that it's been freed and an agent can actually work with it and work with me in real time, end to end across our marketing campaigns. We can do a lot more together. Like, we've been so much more productive, I would say, the last week than sometimes I've been in the last 10 years in marketing because he has all the data. And I'll show you next, he can literally help me set up the campaigns. And so having this single interface to interact with has been so nice that he has everything now and has gone deep on, on our full end-to-end funnel versus having it siloed out. Few questions in the chat on this one, and then I'll get into the ads sample, and then we'll answer the rest of the questions. A few folks have asked me, what is 10K built on? It's, I'm sure it's kind of buried in the slides. He is built on Replit. So, but he's got a lot of connectors, right? So he's got a Salesforce connector, build.com connector, everything I just mentioned connector, had a Marketo connector. All these things are built in to, Replit on 10K. Um, and some, some folks have also asked like, how does, how does 10K work with other people on the team? Um, I'll give you two examples. So just this week, since, since 10K now has God mode, um, and can see our marketing data and has, you know, things like he had already our events data, um, leading up to SaaStry Annual, I started building a prospector tool for our sales team that they now, David and me now use where they can just talk to 10K they don't need to know all the mechanics of the backend, but they can just ask an internal agent, hey, who from Salesforce came to Saastr this year? What have they been doing lately on the website? Have they been doing anything according to like our qualified? Have we sent them any campaigns already that I don't want to duplicate? Like having a central brain, if you will, is also very handy for things like that where you want to make sure, you know, if you're doing things as a human that you're doing the most high-value activity, um, in to complement what your agents are doing, especially when it comes to things like outbound. Okay. Um, cool. Last one here, I think, and then we'll, we'll answer questions, um, and hit on anything else we might have missed, Jason. So, um, this was a fun one. I'll say fun because again, I didn't set out to have 10K become our ads expert, if you will. And without the recent, like, Replit and Claude MCP integration, which is literally a week and a half old now, this wouldn't have worked because Replit is amazing, but it will hit same, you know, limitations as Lovable where sometimes the agent can't do things. Like, it is stuck in its container. Like, it, it is not free to roam the internet for better or worse, um, the way that Claude Cowork can. And so part of this unlock, again, didn't set out for this use case, but part of where our agents are consolidating and actually getting better. This is a really great example, so I wanted to show it. For years, for years, we could not get a person or an agency to do our ads. Like, I had to do them manually.

Not intelligently.

Not intelligently. And, and when we did get them, they sucked. Anyone will burn money with, anyone will burn money with No goal, right? Yeah, exactly. Exactly. So we just couldn't get anyone to do it. And then, you know, to the point of having, you know, finance and 10K, they would just pick— they would just recommend these crazy budgets. Oh, you should spend $1 million on ads. Okay, great. Thanks. Like, I'm not going to go spend $1 million on ads in a week. Like, they're like, well, you could get, you know, 50,000 people there. And if you just spend $1 million, I'm like, I'm skeptical, actually, since you just not run any ads, um, to something like a paid event, um, to our audience, I'm skeptical you can get them all to convert in a week. Um, I mean, if you could really do it for $1 million, I, I probably would do it, but I don't think you can actually do it. And so just a lot of the, um, suggestions we used to get from humans were pretty lame. They weren't based in reality. They didn't know how to set the budgets. And then when we did have some agencies who did ads, I also remember it being a pain in the ass, like, to get them to refresh the ads, right? You'd have to be like, hey, can I get a report of like how this $10K I'm spending is doing? Like, how's this ad— like, I— how's— how are we doing, guys? Like, classic, you know, frustration of I don't even know what the ads are doing. I'm not even really sure what they look like. Um, are they going to the right audience? Who are they targeting? Like, if just a black, a black box, classically ads are like a black box, right? You don't really know what your agency's doing until you get like a report, maybe a month later, maybe a week. I certainly know we give reporting back to our media sponsors a week later, but I know for other folks that's not the norm, right? So it was just a pain, right? And with this Claude MCP Replit thing, this could only do this with Claude because my Claude cowork has my computer. And so as I'm logged into LinkedIn and Twitter, um, it can start to make ads for me. And so the consolidation gain here is because it has finance data, because it has marketing data now, and it already had our Salesforce data. It said to me, literally for this event, hey, I can build you an audience of website visitors and people who came to SaaStr and people who recently engaged with your emails. I can push that into your ad platforms. I don't recommend you spend a lot because AI Day is a free event, but you know, you do want to get some new names and exposure. I was like, great, all, all good suggestions. So far so good. I said, cool, can you actually do it? And so I asked Claude to do it, use a calling of Replit, and he did it end to end. I literally just watched on my screen as he went into Twitter, looked at our website visitors, went into Replit, and then I connected him to Higgsfield so he could actually make the creative. I was like, I don't want to make the ads. Can you make the ads? And so loaded him into Higgsfield and he literally did everything except for push the publish button on LinkedIn and Twitter. So he's like, you know what, your ads are ready. Go look at them. I looked at LinkedIn. I was like, yeah, this looks good. I approved it. He did like 4 A/B variants on like copy, on imagery. I was like, this is pretty good. I was like, who are you targeting? And I looked at the targeting. He did a lot of retargeting. I was like, right, duh, why wouldn't I do, you know, retargeting? Um, and so that worked really, that worked really well. But again, not something I set out to do. It was just an unlock of because our agent started to consolidate and because it had Claude Cowork, I was like, you know, actually it can run our ads for us. I don't know who else is doing this fully autonomously. I really don't. So I wanted to share with you guys that literally. Outside of hitting the publish button, which you should still hit yourself so that you can see the budgets and what it is, um, and be the one to have a guardrail there, if you will. Um, it can literally do it end to end. So that was pretty magical. So that's a good one. Um, the next— so I'll get into a couple things and then, because I know we're running out of time, So AgentForce just, you know, again, naturally consolidated. Um, they bought, yeah, momentum qualified it actually, but now that we actually added Marketing Cloud as well, it's gotten even better. So I think if you think about it from a consolidation gain, if anything, I've gotten a lot of time back here cuz I don't have to log into a million different agent platforms and move things back and forth. That one kind of naturally consolidated on its own. Um, So that's good. We've— I think we've hit this one. It's— oh, so, so just to answer a question in the chat, and then maybe we'll answer 2 more before we run out of time here, is Claude became a better manager of all of our agents. So because now it has this MCP, Claude can talk to 10K, he can talk to Annie, he can talk to QB. And so it's taken a lot of the orchestration pressure off of myself and Jason and has really taken it over. As a better manager of the agents. I also will see like the outcomes seem to be a lot better when Claude prompts for something versus myself. Um, so our time and usage has definitely gone up in Claude, GoWork, and Replit, but I think it's, I think it's naturally for the best because he is a better manager of the agents than, than maybe we've been. And I know there's a few more things I want to hit here. Maybe we can hit it on the next episode of the agents. Um, but anything else, Jason, before we run out of time?

No, unless we have any other good questions we didn't hit.

Yeah, there are. There's, there's a few. Hold on. I know we got to get to the next Zoom though. Okay.

You might have to run.

Yeah, it's okay. Are your agents a single interface?

Did you see this? Did you approve every action? The VP of Finance did ever send an incorrect invoice. How would you deal with that?

Yeah, I think I answered that earlier. It did send one and, you know, I was on it. But like I said, if you just help it every step of the way in the first few, it will get to a point where then it can do it autonomously and then you can just watch it as it does it, right? So now, but to that point, like yesterday, we had a deal close and it asked me, it had a question. It was like, hey, I want to automate this, but I want to ask you if this step is correct before I do the automation. And so there are ways you can do that as well to build in, you know, now, uh, now that it is autonomous, still checkpoints, if you will, so that, um, it can still check in with you if it gets tripped up. Okay. Um, we have another session starting in 5 minutes. So, um, I will take all the questions from the chat. We will do them on, we'll do some of them on the next episode of the agents. Um, thank you guys for joining live. I know we don't always do the agents live. Sometimes we do. Um, so thanks for joining and asking your questions. That really helps helps shape our content. If there's anything else you want to see on a future episode, you can also let us know. You can let me know on LinkedIn or just ai@sastrang.com. That's my real email. Thanks for joining us and letting us talk about our agents. Hopefully this was helpful. Yeah, let us know other things you want to go deep on. We will see you guys at the next session for AI Day and see you again soon.

Thank you.

Thanks, guys.