Welcome everybody back to your favorite show, The Agents with Amelia Larue, our chief AI officer, and me. This week, our AI VP of marketing, 10K, shipped three campaigns on its own. Are we ready to go fully autonomous with no humans in the loop? Why we pay Salesforce more than we did before all our agents, even with less humans. I signed up for a new product this week, AgentMail, because Claude told me to. Is this the future? And 54% of CIOs told Redpoint they're switching their vendors out. They're consolidating, what that means, and what we've learned on it. We'll get into that and we'll get into who's getting cut first. And before we get there, if you have questions on your AI agents, I know so many of you do, drop your questions in the comments, especially on YouTube, and we will try to answer every single question on YouTube. So, use that as a way to reach out and talk to us. We'll answer every single one. So, Amelia, thanks for being back and doing this again this week. It's fun to do this every week.
Fun to do this week. Okay. First couple things I want to talk about in our shop, in our little family of 20 AI agents and three humans. First one at a micro level. Our AI VP of marketing, 10K. And to be clear, we don't call it a CMO, do we?
No.
Okay. Someone kind of hazed me on LinkedIn this week. I didn't even know we'd talk about this. Saying, 'Oh, Amelia is going to go demo their AI CMO at SaaStr and it's not going to have the strategic counsel and advice a true CMO has.' And we're like, 'Agreed.'
And we actually didn't know what to call 10K, but I think AI VP of marketing is good enough. Don't you think it's not a CMO?
It's not a CMO. It's true to what it does every day for us. So, if you guys haven't seen our AI VP of marketing yet, we call it 10K. I originally built it as a dashboard. It started as just a super simple agent, you know, pull in some things from Salesforce, pull in from some systems of records we had just internally, basically pull it all together, help me keep on track as we get busy in the run-up to SaaStr Annual, of where we are year-over-year, and then we layered in what should we be doing. And that is the piece that's now gotten a lot better the more we've been iterating with it and spending more time with it. And so it's become now a true AI VP of marketing, not a CMO, in its role as a VP of marketing. It has been telling me what to do, but now we've shipped this, which we'll talk about, where it's giving us campaigns and now it's starting to run them autonomously.
Okay, good. So, and so what we added this week, and it's in the age of AI, it's not amazing, but I do think it changed, is it started generating real ideas for campaigns to run. We have not yet let it run them autonomously through Resend or otherwise, even Salesforce, but now every day it's come up with campaigns to run, how to target more VCs, how to target folks that should be sponsors that aren't. How are we thinking about these campaign ideas compared to humans? And when do you think we'd let it go autonomously? Like literally run campaigns to hundreds of thousands of folks on its own.
I would let it do it if all the APIs truly worked. Now, some of our APIs we use for SaaStr have some limitations. Marketo, which we'll talk about more, has a lot of limitations. It hits its API limits pretty easily. So it really can't do much with actually sending the emails. Resend, I don't know if I love for our actual emails, right? It feels a little generic-y. But what it did do, so this weekend it gave us, it sent us both an email of its top three ideas. I think the first one was you should get all the single ticket buyers to see if they want to bring a buddy, you upgrade to a team pack basically, which is three.
So, let's step back for a minute. Over the weekend, Saturday and Sunday, it works weekends 99.97%, right? 10K. And he sends us his three best ideas of the day based on looking at the very latest data across the organization. Three best ideas of the day, right?
Yep. So I think Saturday morning it said, 'Okay, first thing you should do is you've got a lot of your paid tickets year-over-year is higher, right?' Which is awesome for SaaStr Annual, 'but you have a lot more single ticket buyers than you did historically. So you should try and basically do an upsell campaign to them.' This was 10K's first idea on a Saturday. And what it did though, I think the next part which was pretty awesome is it laid all the groundwork for us because it can't actually execute this autonomously yet, but it said, 'Okay, here's the list of people that I've already pulled for you from your ticketing data of who's a single ticket buyer who only has maybe one ticket or two tickets and they're bringing their co-founder, but they look like they're at a company either size or scale where they could feasibly bring more people.' So, it already segmented the list for us, which usually that's me doing, right? That's a human going out segmenting your database, finding these people in the list to then reach out to. Then it drafted the email copy. It said here's, so it identified who to send the marketing email to. Then next it created the segment for me. It said here are the exact people to send the emails to. Plus also I've already suppressed anybody who I think you know will not do this maybe from you know different signals you guys have in the data but like I've already, there are current ticket buyers but I've made sure these are the right people that have the highest propensity to buy. So it already segmented the list for me. I didn't have to do anything there. It gave me the list and then it wrote me the copy. So it basically did everything short of hitting the send button. And so that's the piece I literally just did this morning and it took me a couple days lag. But that's pretty magical where our AI VP of marketing, much like true things I used to do as our VP of marketing, would segment the list, write the copy, identify the right folks, but also he came up with this campaign. This is not something I came up with or Jason came up with. This is something he came up with based on data, based on where we're tracking today with how many days to SaaStr Annual, and said this is what you should be running.
So it's building the dashboards. It's reviewing the data daily. It's coming up with its three best ideas seven days a week. And then it's giving you all the components, doing the segmentation, analyzing everything from hundreds of thousands of folks and data points to come up with all the campaigns and ideas. It's not doing it autonomously. So for folks that, it's pretty incredible. For folks that say, 'Hey, that's not a VP of marketing.' Okay, maybe not. Some of that stuff is director of marketing. Some of it's AI VP of marketing. Maybe some of it's marketing manager if you could find this person. But it's doing all of these things every single day. And even if it only does 50% of what a great VP of marketing would do, 50% of a great VP of marketing is pretty good for 99% of people. They don't have 50% of a great VP of marketing, right?
Yeah. And I think also if you have inconsistencies, this is where I've started to see the AI will start to beat even you and I because it's consistent, right? We saw it on the customer success kind of layer for us this week with QB, who's our AI VP of customer success, but I'm starting to see it with 10K where unlike a true human VP of marketing, it's not going to forget these, right? So, it's now, it remembers all the campaigns it's coming up with. It's remembering all the email copy and today it asked me when I logged into Replit if I sent these yet. It was like, 'Hey, I came up with all these ideas for you.'
Have you done the work yet?
Hit the button. Yeah. Have you done the work yet? Which...
I actually like that. I like that it's again acting as our VP of marketing. So I don't have to think about these demand gen campaigns, right, to increase our revenue. It's literally doing that for us. It's just stopping short right now, which is not a failure actually on its part, but more of a failure of the APIs of some of the tools we use that it stopped short of hitting send, but yep, in like a month or two post-Annual, it probably could hit send on a lot of these things. And then we can move on to the next point. But so we built our AI VP of marketing, our AI sales team, our third-party tools we use, Artisan, Qualified, Monaco, and Agentforce, right? Those are our four. For the most part, those do run autonomously. If we set them up, review the text, review the guardrails they have, they've sent over 100,000 emails on their own, right? So what would it take for a tool we've built ourselves to that comfortable we can unleash it in the same autonomous way like third parties with lots of engineers working on them have done?
Yeah, I think we would just have to duplicate the steps, right? So, a lot of our third-party tools have additional security layers that our AI 10K VP of marketing doesn't have yet. It also, I would say the biggest piece of it sending emails will be the warm-up period. So, a lot of what our Monaco and Agentforce do well is they do the warm-up for you for the most part, right? Like they're going to spool up custom email domains. They're going to warm it up with data. They're going to send a few emails that are not your true emails yet from your AI SDR. They do all that warm-up for like four to six weeks. So, you know, could we do that with hooking up another third-party tool to our 10K VP of marketing? I'm sure we could, but it would take time, right? It would take that time. Plus, it would take all that labor to say, 'Okay, we've got to warm up all these domains.'
Enable to get the deliverability. Because that's the biggest piece right now.
API that claim they give you pre-warmed domains.
I know. So, and we will try it post-Annual. Like I'm curious to see how...
We need a warmed email campaign though to replace what we do on Marketo for folks who have already received years of emails from us. Do we really need warmed up emails or couldn't we just, is email really the big issue or is it having confidence that these marketing emails are properly sent and run? Is the warmed up email really the big issue?
For me, that's part of it, right? I feel like having a warm domain so that the deliverability feasibly goes through. That's part of why all of our AI SDRs are pretty good because they go through once they're warmed up from, you know, a Salesforce or once you put in your first campaign, that thing goes through like that. Like our delivery is like 99.9%. So, for me, that ties all together of what makes a really good email and one that folks would want to actually receive and click on, right? And so, I feel like that would be kind of the first piece for me. Now, yes, part of what we've done already on QB and now what I'm starting to do on 10K is to use Resend as like a backfill because these folks are already in Marketo because they're folks that are already in our Salesforce to say, okay, can we send kind of one-off emails? And it would be okay for deliverability to get them to do a specific call to action, but I'm not going to move all my emails there today, right? I'm not going to move all of our newsletters.
Yeah.
It's a little bit out of order, but we've talked about building an AI VP of finance, right, on the last show. We even played around with some AI SDR tool we built ourselves. So, I went to Claude and I asked what we should use if we wanted to build this outbound AI campaigns and it said to use a YC company called AgentMail that's pre-warmed and set up. So, it's interesting. First of all, I did what Claude said. I hooked it up. You haven't even played with it yet. It did work in about one minute. And the emails, it says they're warmed. I got the emails instantly to the top of my email. And it was interesting both because I used what Claude told me to do. It was hyper-agentic because I did it in one minute, right? And it did seem to, you know, with six emails sent, it did seem to have high deliverability. I don't know. But it was sort of an interesting agentic discovery. Here's the visceral part for us too, right? Like I post-Annual and we have a bit more time. Like I feel like it would take us a day to fully build out, you know, and maybe we'll do it over this weekend like before SaaStr Annual. So it's for the session, but I feel like he's already at a point where he has started to send light emails for us in the same way that QB has sent emails to our customers.
Our AI VP of marketing. Correct. Yeah. So he's started to send some of these segmented. Again, right now he mostly stops short of hitting the button for a few things, you know, hyper-segmented list, we're talking like a hundred or so people. We have our Resend hooked up, he can start to hit the button, but I think beyond that, part of what makes me nervous is what happened with our Marketo. So if you guys have been following along, last week we talked about how our Marketo API was really dated and we hadn't even hooked it up. So this weekend I hooked it up to try and solve the issue we've been having, which is still ongoing.
Let's step back for, I want to hear the story, but we've been using Marketo for marketing. It's a very old platform, but we've been using it since you started, so since 2018 or so, mainly because a prior person just switched us without telling anybody. So we were prisoners. Our data was hostage here, right? And among, and we actually, the API is so old. Replit said some of it hadn't been touched since 2014. At first you thought actually it was impossible to use it with the agent. It was so old.
First I thought, okay, obviously there's no native integration like there is for Salesforce or Claude or other things. So I was like, you already know it's going to be a handicap, which is what we said last week. It's not, it's going to suck because it's not native. Yeah. But we've had this ongoing issue where we have had folks unsubscribe from either one of our newsletters or all of our emails. It's after we sent a lot of emails and it wasn't accepting them. Like they would unsubscribe. I would see them as an unsubscribe in Marketo. And for some reason they would still get one of our newsletters even after they were, you know, checked the box, unsubscribed. And for some folks, even after, you know, they asked us, hey, I unsubscribed. I still got an email. You need to delete me under CAN-SPAM rules. And so we would delete them immediately. And they would still somehow get these newsletters, which baffled me because if you're deleted, you're not in my database. So how are you still getting an email? That's not something I can fix. That's not a setting on a page or a form got out of sync. That's not something I can fix as a user. That's something on the platform. Right? So, we reached out to Marketo and how this all comes full circle is this is why I'm just a little bit nervous to move some of this autonomously to our AI VP of marketing because right now even our legacy platform cannot handle unsubscribes the way it should. But here's kind of the double-edged sword. So, because it wasn't able to handle unsubscribes the way you normally would expect, I went into Replit. I finally hooked up the Marketo API after our past episode and I said, 'I know this is going to be dated and it's going to be terrible because it's not native and there's no integration, but what can we do?' Like, can you help me see the problem, etc. I explained to Replit what I just explained to you guys. It said it couldn't find the issue because again, these contacts were deleted. So, like, how are they still getting emails? It said reach out to Marketo, which we had already done. But beyond that, it said, 'Here's what I'll help you do. I'll help you build an unsubscribe page that actually works because now that I have API access, I can unsubscribe these people for you automatically. No forms needed, no nothing. It'll be cleaner because right now it was a form, you know, you hit the page, you fill out the form, you put in your email, you fill out the form, you hit unsubscribe, then on the back end, it does all the steps to unsubscribe you.' Because Replit now had some API access even though it was dated. It said it could do that for us without all the middle steps, right? Which seemed a little cleaner. So what we did is we built this custom unsubscribe flow using Replit, a live coding tool, to solve a problem we're still having ongoing with a very legacy platform. And so I feel like it's put us into a precarious position where I'm like, okay, I'm obviously not going to stay on Marketo, but the fact that a vibe coding tool could fix this problem for us in the interim.
Yeah. How, like, we can just take that further. Like, to your point, we can probably just start sending all of our emails if we hook up, you know, the right APIs, get all warm, etc. And if in theory we can find the right tool to actually handle unsubscribe, actually handle, you know, some of the data processes that it's doing now, why wouldn't we just do that?
We might. It's also an interesting learning as the dated leaders, the old folks of B2B and SaaS get older and older in the agentic era, as we're able to, like, you know, a few months ago when you had a massive feature issue like the unsubscribe button doesn't work in an email marketing platform, pretty bad. You were a prisoner. You had no choice. The only choice you had was to migrate. Can we leave to HubSpot right now if you can vibe a hack, vibe a little agent or an agentic fix so that the fix is plugged in? That's extremely powerful, right? You can solve your own problems. It also just makes the vendors look worse.
It's just terrible that Amelia can vibe a solution to a critical issue, a CAN-SPAM issue in Marketo. That she can vibe an answer in an hour of work, but their whole engineering team forces her to have an all-hands call and weeks later has no proposed solution. They're just going to make you look worse and worse and worse. As vendors, if you can't work at this pace, if your customers have to vibe their own fixes to your bugs, now they can. They don't have to be technical. Like an API is exposed to everybody. So, it just makes you look worse if we can do it in an hour and you can't do it in a month.
Yeah, we're on day 10 now of no resolution from the Marketo side, which...
Yeah, I see. This is the page you vibed to then go into the Marketo API and just delete the record because they're unable to do it.
Correct. So, now this will do it for us. And actually, if you're signed in, which I'm not, it'll...
If you click this from the email, it'll pre-populate your email address, which is pretty cool.
It's already, that's better. And then these boxes you can just check of like, yeah, what do you want? Do you want to just unsubscribe from everything? And now it'll directly do it from the API. And Replit did this in about 10 minutes. It maybe took us 20 minutes to hook up the API because there were some, again, it's old. There were some issues. But as soon as we hooked it up, it took him maybe 10 minutes to make this page. I just swapped out our default unsubscribe link in Marketo to this one and there you go. I haven't had, knock on wood, the newsletter comes out every other day. We sent one yesterday. I did not get any angry unsubscribe emails. So I actually think the Replit band-aid solution worked. And yet on the flip side with these legacy SaaS tools, we're on day 10 of no solution from the Marketo side. And you know, it literally took them until day seven, which was Monday, to say that they were investigating this and they wanted us to get on a call, which I did. And basically what they told me on the call, the actual humans there, said that they hadn't really investigated yet and replicated the issue and they didn't really know what was going on. I said, 'Well, then why did I get on a call?' So, it's just interesting the vast differences of support there, right? Like fixing it yourself in 20 minutes slash waiting now 10 days for a non-solution.
Yeah, it's, I mean, it is pretty bad if you have to hack your own solution here. It is interesting. I want to move on to the next topic, but you could imagine where you show up to a mediocre customer success or other call and there's not a problem. They just tell you to go vibe your own solution. Hey, it's 2026. I can't help you, but here's $50 of credits on Vercel or Replit. You go solve the problem. I'm not going to do it. You go solve the problem. This should accelerate our turning off these dated platforms because they look terrible, but I wonder if it'll make mediocre CS folks even lazier. You go solve it, right? And yeah, someone had a comment on LinkedIn saying, 'I'm surprised Marketo didn't hook you up to some consulting firm that was going to charge you $40,000 to fix the problem.' Maybe they will, right? But maybe the AI version of it is you go solve it. You figure it out. We're just a 20-year-old software platform that charges you $60,000 a year. We don't provide support. Okay. So, I want to go to the opposite topic next. One thing because I think it'll be helpful to folks. One thing that over the last week or so has been in the news. I mean, Cursor and other things have been in the news more, but one was Marc and team at Salesforce did a big announcement of this release on headless Salesforce. I don't want to talk about the product itself, but what I thought would be interesting was to talk about how our Salesforce is already headless inside of 10K and others. I have never used our Salesforce instance ever. I don't even have a password.
I don't even have a password to my knowledge or I don't even have an account. Yeah. Yeah. You're having...
I don't. Jason at SaaStr...
does not exist in Salesforce.
does not exist.
Okay. But here's what's interesting. Now that we run Salesforce headless among other places inside of 10K, our AI VP of marketing, I check Salesforce every day. I never used it for a decade. Now I use Salesforce because it's headless every day because that data is exposed inside of our AI VP of marketing that has all of our 360 data and that I can use. So that's interesting. The other thing that's interesting that I want you to dig on, which I should have known, we're down to a couple seats on Salesforce per se, right? Because our team is 20 agents or humans, but we pay more for Salesforce than we used to.
This is the thing. So there's a lot of folks out there that are saying the death of the seat is the death of traditional software. And to some extent it may be, but with our agents using so much data and so much access, we pay more and it's okay. We wish we didn't, but it's okay. So you want to talk about the cost and then how headless works a little bit.
Yeah. So on the cost piece, so our Salesforce bill as it stands, and this is before I do anything on the marketing side to fix the issues we're having, but as it stands today versus a year ago, so a year ago, April 2025, we had more humans. So our Salesforce bill was something like, I want to say 12 to 15k ballpark off the top of my head for, you know, price per se, whatever. It's like a couple grandish per seat and there's like an API seat, whatever, it's 12 to 16k. That was our bill last year. This year right now as it stands in April of 2026 is I believe something like 22 and change and users came off.
On a percentage basis, that's a lot more.
It is a lot more. However, I don't mind because the price obviously per seat went down because we don't have as many users. You're not even a user, you were never a user. So basically, I'm paying for me and David and I think one more like admin seat for some other API thing. So, let's just say three users effectively on account and the rest of it is just data usage. It's, you know, our Agentforce usage. And they did base it off of like how much we used and blew through all of our like free Agentforce credits and all that jazz. And they're like, this is how many you're going to need for the next year. So, at least it's for, you know, the next year. And they also, you know, they did it, that also included an increase of us to the Slack Enterprise plan so that we could use Slackbot. We were also on a free trial, but then we used, we blew through all those in like a day, I think, and Slack was like, upgrade. So we upgraded. That was like only, you know, maybe a thousand of those of the bigger increase. So that one wasn't, that one is honestly, if you're just upgrading anything, just upgrade that one because it's pretty cheap for what you get out of it. But so that's how it's changed. So now instead of paying for seats, we're paying for pretty much data calls and actions, right? And you could, I know some people are critical of Salesforce on like how much they charge per action. Do I wish it was a little bit less? Yes. But I think they'll get there. I think part of how Slackbot works now of, you know, being able to call up your agents kind of for you is a good way to do it where it's a little bit unclear how they charge on all that right now and how it all gets worked out. Maybe Marc and team don't care as long as you're using it and getting your agents deployed. So I do think our Salesforce, even though we're calling a lot of data and we may call more data next year, I still think it may go down a little bit year-over-year because it'll decrease the cost of all the data actions. And since we're starting to store more and do more with not only Salesforce itself, but mainly the APIs for this headless kind of agentic Salesforce run, I don't know, maybe it's the wrong intuition, but I think ours will go down even though we're calling more data because the data is getting more efficient though because now it's becoming one place, right? So that's helping some of its efficiencies too.
So we, for folks that are thinking through this, we have cut our Salesforce over the last two years. Let's go back two years. We've cut our Salesforce seats back 60% or 70%, right? We're basically down to two plus an API, right? So, we've kind of... So, if Salesforce hadn't raised prices and nothing else had changed, our bill would be less than half. We'd be paying less than half of what we were.
Now, we're paying 80% more because of our agents and data. Instead of paying half, like if that... Now, if I were a CIO at a Fortune 500 company, I saw my Salesforce bill went up 80%, of course, we'd have to have a conversation. A million. But I think the meta thing for folks to take away is it's worth it to us to pay more, right? And that the seat story even at Salesforce, even at the classic, the folks that basically invented the classic seat story, is not that simple. Now that we have 20 agents running and really four or five core ones that run through Salesforce that are pushing data and pulling data out, mostly the sales folks one, we may do more marketing. Net-net, we're paying 80% more and we don't, we, you know, we're going to get better at it so we will drive some cost down. Our agents will get better, we'll use less data, maybe Salesforce pricing will come down a bit, but in the wash, it's a good deal for both sides.
Right, people are missing the story here. And so as more folks go headless, this, I think this is for everything, for every core app. I think it's because we have some, it's such a positive to be headless, right? And now I don't, maybe in six months, 10K, our VP of marketing, maybe we'll have a different agent that's like our support, but right now our VP of marketing runs everything for us, right? It is becoming our hub and it just makes Salesforce so much more powerful. To see Salesforce not this isolated app with leads, contacts, and opportunities, but it's integrated into every single metric, every single attendee, every single dollar. It's tied to how many newsletter subscribers we have and how many social media followers we have. It's all integrated inside of 10K and it actually means we can do anything and run any analytic and any query we want against it. This headless thing. So, anything else? I think the economics now we use are in. Anything else that is interesting running headless that folks might not get?
Only because I think we're a little... they just announced this, right? So like a week ago. So I think this is from what I've heard from the folks at Salesforce, right? This is still a newer concept for most of their customer community that they're, you know, I think people were still getting their heads around just Agentforce and now headless Agentforce, they're like, what does that even mean? We were just getting on the Agentforce train and now you're like, here's this shiny bullet train to get on, which is headless. So I think that piece is interesting, but just because we've already been living it for six months, the headless piece, you know, to some degree, I don't, because the API is good, I think more apps and tools will become headless, right? Like as long as you have a central... I think why headless for Salesforce in particular works is one, we already had a bunch of data in it. So to Jason's point, did he need to log into Salesforce? No, he just needed a dashboard that hooked up into the API so he could see what all the rest of the team and the business is doing. Plus, he can talk to it. Like I think the biggest thing about headless with the API is you can talk to it in a way a CEO can't normally really talk to their data in Salesforce. Like you can talk to Slackbot if you have enough data and it kind of replicates that vibe now, which I think is pretty cool. So I think they're catching up in that regard. But outside of that, I think being able to just talk to your apps with your API and say, 'Tell me what's in the data. Give me ideas the way our AI VP of marketing has. Tell me where we're behind. Give me a list of folks that have turned that maybe I should reach out to that the sales team hasn't reached out to. You can just talk to it.' And I feel like that is really the piece of it that is helpful, I think, for us. That sometimes Salesforce, I think they'll even admit, sometimes it's a little bit clunky to get things done and it's hard sometimes to find things or know what the right filter is to get the right report or see the right dashboard. Or sometimes that stuff is hard and as a CEO you may not have time or the want to just go through all the filters and figure out how to do it properly. But now if you just talk to your agent or talk to Slackbot, hey, run me a report of all these things and it does it for you. Maybe it doesn't matter what's on the back end. As long as the API works and it can get you that outcome, I feel like more apps will start to become this way.
Yeah, it was interesting. I'm on the board of a startup called Owner that you know, that's sort of a next-generation agentic Toast or Square for restaurants. They're at about nine figures in revenue, growing triple digits. And the CTO just made a Loom and he shared it with me. You saw it too. They run headless Salesforce and he built this before they called it headless, but he did well. He was inspired by us and he wrote his own on a hackathon. And as a CTO, what he did was a couple things that were pretty amazing. They built their own AI VP of customer service like we did, but it does even more. It's pretty cool, right? They didn't really have this before. So that one human can manage thousands and thousands of restaurant issues now with their AI and it runs all underneath on Salesforce data. More interestingly, as a CTO, he took Momentum, which we used to, that Salesforce bought, and he used all the calls with all the prospects and one lost and one won. Okay, one, he looked at both and used AI and Salesforce to analyze every single call to figure out what the top feature gaps are that they needed to win these deals and the real reasons they chose Owner over other solutions. And that was pretty cool. And his point was great. He's like, I used to not enjoy working with Salesforce. Like it's not as a CTO, right? I mean, I love Dean, but at the end he's like, I didn't used to enjoy working with Salesforce. Now, we have this incredibly powerful data set that is figuring out why we win customers, why we lose them, and can in real time analyze every single one of our 10,000 existing restaurants and customers headless. That is so powerful, right? And I suspect they'll pay more for data, right? They probably have 200 seats, 200 Salesforce. So this headless thing, it's a big deal. Okay. Next topic I want to talk about is very micro. So that was very macro, headless Salesforce. I want to talk about another thing you built this week. It's a very micro N equals 1 app. As I think I'm going to say that D. E. Shaw coined this term, but maybe he didn't, maybe I copied him, who knows, but an N equals 1 app, which is something that's super powerful but only you need. So in a way our AI VP of marketing is that and that the custom version we worked on, use only this, but everyone needs an AI VP of marketing. But you built an AI parking pass app. I really...
Literally no one needs this but us. Well, maybe others like us, maybe Coachella and Stagecoach need it.
But why did we build an AI VP of parking pass and it's for us, it's actually disruptive, right?
For the shuttles, hit us up. All right.
So just for some context and background, I've talked about it, I think on episode one, but historically, because this relates to everything of like costs and efficiencies and people. So, just a quick backstory, we typically use an agency or two this time of year to help backstop, backfill some of our humans. Even now still with AI, we still use a couple agencies to help backfill our humans slash, you know, we have some folks that are really good at things that the agents can't do yet. So, we always backfill every year. So, I said a couple weeks ago how much our bill has gone down with both of these agencies because we have such great efficiencies with our AI. And actually they like it because then they don't have to do these tedious tasks that they used to have to do. They can focus on the fun stuff. So we have more fun stuff at SaaStr and they can help me more with like the human in the loop support and like speaker support again that they previously didn't have time to because all their time was spent on like mundane, you know, step and repeat tasks. So that's a little bit of the context. So every year typically we'll see our bill go up the month before Annual, which is now, because we have to distribute parking passes manually. Super niche problem, right? Like...
4,000 parking passes manually.
They're all individual PDFs. They're not just a code. They're 4,000 individual PDFs in a massive file, right?
So, the first thing I did is I gave the folder of 4,000 PDFs to Claude Code. I said, separate the PDFs because last year we had a person who did this and it took them about a week to go into Acrobat. So tedious. And literally separate the pages one by one and then save them as their own file.
That's 4,000. Cut in Acrobat to individual pages. Save it as a file. Get the file name right. Not blow your brains out because you did the same one three times.
Yeah, that's a full-time week.
It's a full-time week. Super tedious.
So how did Claude rip through it in, I want to say, two minutes? He said, 'Here you go. All your files individually. I put it in a folder for you on your Mac.' I said, 'What the...' I said, actually, if you're done in two minutes, these parking passes we get from the venue don't have the address on them. It'll say like you're assigned to the east lot or the west lot. I'm like, 'Most people don't know what that means. They're not going to go look that up.' I said, 'Can you append the actual address, the street address where they need to go so when they're driving to the venue, they know exactly where to go?' He said, 'Sure thing.' Maybe five minutes later, I got all 4,000 back in a folder, not duplicates. He just updated the existing files, which I thought was pretty cool. He updated the existing files with the address in a nice banner at the bottom. So, he highlighted it and stylized it already and said, 'This is the actual parking lot address you need.' And I said, 'Fantastic.' So, then I took the folder into our Replit that powers a 26 that you see here. Building a form, super easy. I was like, 'Hey, here's all the info I need to assign these.' Okay, great. Then I said, 'Okay, here's the kicker because this has to work now. Can you take this folder of 4,000 parking passes and wire up logic so that if I say I'm an attendee and I need all three days, that you will give me the right parking pass?' And it said, 'Yes, here's what I need you to do.' And he said, 'Put the parking passes in a folder by day and by type,' which Claude just rearranged the folders in five minutes. So he did that. And then give me the logic of who gets what lot. Like, okay, attendees go into like the east lot, you know, speakers going to the south lot, sponsors going to the west. So he was like, 'Give me all the scenarios. If I'm filling this out, what do I need?' And then I gave him all that. I gave him the folders again and he said, 'Okay, but can you put that into Google Drive so that I could actually access all the PDFs?' So, I did that. Then he said, 'Okay, now give me the logic.' He did some test runs, right? Okay, give me the logic. If I say I'm a sponsor and instead of picking all three days, I say I need, you know, just Tuesday and Wednesday. I think he said, I think my theory is I think I should give them a parking pass for Tuesday and a parking pass for Wednesday instead of blowing one of your all three days parking passes which is good for all three days. I think I should give them individual passes for Tuesday and Wednesday. Is that correct?' I said yes, that logic is correct. Typically when we've done this manually, if somebody doesn't request all three days, we give them the individual days that they need. He said, okay, great. He said, 'Let me rattle this up for you.' He's probably working for like 15 or so minutes in the background. And then he said, 'Okay, it's ready. Do a test.' And I said, 'Okay, let's freaking try this.' So, I filled out my email. I'll fill it out right now and I'll show you guys what it gives me. So, I'll do my email plus like 12 or something because I tested this so many times.
All right. So, I'm going to put in my name.
I'm going to put in my company. Does it really need to ask for this? But what it's doing is it's checking our API integration just lightly to see if this person seems like they have a ticket to SaaStr versus giving it to a spam bot. So, I was like, I don't know if I need all these fields, but it felt better about giving it to actual people and not, you know, getting hit on the form constantly by putting that in. So, I'll submit that request.
Okay. Check your inbox. So, your parking pass and lot assignment have been emailed to you. Okay. So, now let's spool up my Gmail here. Okay. So, here it is. It's the first email in my inbox. It is my two parking passes because I picked two single days. And magically, here are my two individual PDFs. So, here's the address that Claude appended. That's cool. So, I picked Tuesday. There's Tuesday.
I picked Wednesday. There's Wednesday.
And then the last thing I did with Replit before we launched this is I said make sure you burn these. So, like whenever you use a parking pass, they are sent...
I said, I was like, whatever you do, don't send it to someone else. I was like, hardcode that in your memory. Don't send it to anybody else. This is one and done. Never repeat the same PDF or parking pass twice. And it said, 'Okay, you got it.' And then it said, 'Also, just as a backup,' he said, 'Give me a Zapier webhook and I'll put it into a Google Sheet of which parking passes I assigned to which person. I'll also store it in the database.' So, if you're ever worried that I assigned the same pass to the people, you can spot check it and make sure I didn't. So, super niche use case, but this, not only did the setup of it used to take us a week...
Literally distributing parking passes was a full-time thing for all of us. Like, we would all be on the inbox and we'd be like, uh, someone else has requested a parking pass. We would have a Google Drive. We'd have to make sure between the humans, we didn't use the same PDFs for a...
A lot of human error in that process with multiple people who did not want to do this job for weeks.
Well, I want to send people...
Yeah, you have more important things to...
do. But it took so much time that it was one of those things I had to help the team with, right? I was like, it's just one of those things that's such a big time. Like every year I have to give people parking passes. So I said to myself this year, instead of that, I'm going to take this — maybe it took me 90 minutes total with, you know, testing and hooking up all the PDFs and getting it all together — like I'm going to take 90 minutes one shot on Sunday and get this all done and routed versus hundreds of hours leading up to disaster, which is a huge disruption.
Okay, this is a good transition to our next topic of where you find your Amelia. So you were able to have the experience and the foresight to identify, okay, what issues are coming up for a large event, and everyone's got their own issues, whether it's a release, an event — what are the big blockers that we could create an app, vibe an app that could solve massive blockers, right? And so you quickly identified one that takes hundreds of hours of people's time, is incredibly tedious, prone to error, prone to friction, and creates hundreds or maybe even a thousand hours of it. You'd have to have a massive team to support this, or frankly, what most folks do is they don't even try. They don't even try to — like you just show up to a large event and it's terrible. Like we showed up to Coachella this year in our team trip with our agents and our three people, and there's like a three-hour wait for the bus at our hotel, right? They just decide not — these problems are so complicated, without an agent you can't solve them with humans, right? So you solved it. You had the foresight to see it, to care, to scope out which top issues that were worth solving, right? Because we can't solve everything agentically, and then you executed in about 90 minutes with a lot of experience. So this isn't now — this isn't the first app you've built, right, or applet that you've built. So I want to give you a quote and then ask you a question. Where do we find it, Amelia? Aaron Levy was on Harry Stebbings' 20VC this week and said there's going to be a whole new category of $500K to a million dollar new AE jobs for operators over the coming years. They're somewhat technical like you. You're not an engineer. You can't really code, but at this point you're getting somewhat technical. They're deep in AI, MCPs, and CLIs and skills. We need these folks. And I know we would both agree absolutely everyone needs one of these folks, right? You don't even need a hundred, but where the hell do you find you? Everyone needs like what I just described. Every organization needs this, right? You can imagine a hundred blockers you have in your company, right? Um, someone could even just vibe code the fixes to Marketo if Marketo had a couple of folks like this, right? So, um, any thoughts on this? Because I think Aaron's right, but I think he's understating how hard it is to find these folks and what it takes to build them internally.
