How We Vibe Coded Our AI VP of Customer Success "Qbee" with Jason + Amelia

Wait, no, it's being recorded.

Anyone know anyone on the product team at Zoom?

Just not sure if they care.

Doesn't feel, doesn't feel like it.

Please note the call is being recorded. All right, all right, that one's getting booted for doing it multiple times. And all right, now we're down to the people. I think all the bots are in. Okay, let's just check our stream real quick and then we'll get going. Appreciate your guys's patience. Just sometimes I see it. You see it? Okay, awesome. It's good.

Welcome everybody.

Welcome. All right, sweet. Let me share my screen.

Maybe for a little context before we start, um, Amelia will drive most of this. Just, I know a lot of you have followed our Agentic journey. Um, we started off, uh, rolling out a bunch of AI SDRs. We built a bunch of AI VC tools that did a million startup valuations. And then we started building our own agents. We talked earlier on AI Workshop Wednesday, and we'll do another one on our AI VP of marketing. We've written a lot about it. It really runs our team meetings and every day logs into Slack and tells us what to do and stresses us out and nags us. It's pretty good. And then Amelia built that. She built, I built the early ones, then Amelia took it over, built our AI VP of marketing for what she needed and then built And she'll, she'll drive, talk you through this. What now we see is our AI VP of customer success. It started off as a project management tool for our sponsors at SaaStr Annual, SaaStr Annual in May. Please everyone come, we'll have 10,000 folks, the best of AI and B2B, May 12th through 14th. And it quickly grew into this AI VP of customer success tool. And I just want to highlight two things and then I'll let Amelia show it to you so you can see it. Um, uh, one is just at a meta level, this is the best thing we've built so far. This AIVP, it works better. It has fewer bugs. It adds more value versus unit of time increased. So I think the obvious reminder is if you're going to build your own apps, vibe your own apps in Replit and Lovable or v0 or others, um, just understand you'll get better and better as the journey goes on, like everything. And, um, I started this journey. Amelia took it over, took the baton. The VP of marketing is pretty good, but this one's even better, and the next one will be even better. So it may take you 6 months to build the AI agent of your dreams. You may need to build multiple agents and apps to get there, but ours just keep getting better and better. Okay. And the second point, this one is so important. We're going to show you everything our AI VP of customer success can do. And I know a lot of you, and what I want you to do is see what we built because you can build it yourself. If we built it, you can build it. We built this with no engineers and it adds dramatic value to our our little tiny team doing 8 figures of revenue. What I think isn't helpful that I see a lot of folks do on social media, and it's fine that they do it, but I think they're missing the point. Like, oh well, that only works for SaaS, or oh, that only works for a media or community or events business. No. First, anything we do that is so niche to us we don't think will help anybody, we don't share that stuff with you. We don't, we don't want to share like the hyper niche stuff that only helps if you want a popcorn machine in your booth. Okay? We're not sharing, sharing that stuff for the most part. We're sharing stuff that if you think about it is universal to software and B2B companies. What is universal is going back to a larger install base than you can realistically touch with humans, going back to them when they have hundreds and hundreds of deliverables and things to do that is difficult for a human to follow up. So what I want you to do today, and then I'll let Amelia take it, is not look at some of what we built and say, oh, that's, that's an n=1 app specific to SaaS. Or it is. That's true. It is an n=1 app we built our own AVA. But think, wow, what if I built my own version of this customized to my business? Not an off-the-shelf piece of software. Could I tweak the workflows that Sastra has built, the follow-ups, the, the, the dashboards, the, the agent, and do my own version? Because you probably can. That's the better learning from this. You probably can. So I'll shut up, but, but think more about this of how can you make a prime version, an iteration of this for your own business rather than take shots at how any individual marketing campaign or workflow we do isn't what you need in your, in your vertical, SaaS for dentists. That's not the point. So thanks. I'll let Amelia drive and then I'll kibitz and comment.

Yep. Love it. Thanks for the, thanks for the tee up. I think it's super helpful for folks to have some context. I added some context here as well, just more specifically on our AI VP of customer success that we'll get into in detail. And then I go through the steps of how we actually built this. So first and foremost, if you want to see Qiby in action just to see like what we've built, you can go to sasresponsors.com. We've shown it before. It's changed a lot since the last time we've shown it. You won't be able to do everything because if you're not a sponsor, you can't sign in. So if you're not one of our customers, you may not be able to do everything in it without signing in. But it was one of those things where we did decide when we were building out this platform and then it turned into an agent that not everything needed to be gated. Some of this is still, you know, folks have multiple people on their teams that are part of, you know, our customers, our sponsor teams. They have agencies, they have folks who are CEOs, they're not going to want to log into something like this. And so we purposefully only gated things that required, you know, data entry, hotel, you know, things like the registration codes you can't do, you can't submit anything until you log in. Keeps it a little bit more secure, but you can go to sasresponders.com and see what it looks like for yourself. The other thing you can do, to Jason's point, if you do make something like this for yourself, is you can actually give this URL to your, your vibe coding agent and say, hey, this is one I saw that I liked, or I like this format. If you're do— if you're gonna do something similar, that's like a customer portal or customer management tool or whatever, just at the very start, um, you can actually give it URLs like this cuz it can pretty much copy it. Um, so you can use it as kind of like a format too. So feel free to use it as a format. So, um, let me go full screen for a bit now and then we'll jump around between, uh, the live front end of Kiwi, which is sasresponders.com. There's some other things too I'll show you, um, that are not just contained to the website. And then the backend, which in this case lives on Replit. But let me go slideshow. All right, great. All right. So yeah, something Jason touched on, which is important, is Cuby did not start this way. That was not his name. He didn't start this way. He was just meant to be a custom, basically project management tool that I wanted to use. With VibeCoding in this day and age to replace an existing tool we literally paid for for the last 2 years that had no AI functionality at all. It had no AI functionality, nothing agentic. It was very much just a portal. It was just like a customer portal. People could log in and submit us things, which they can still do now, but you'll see the key differences in a moment of, you know, why it was worth putting all this extra effort into VibeCoding and putting, making it so much more agentic and hooking it up to different things. And so you'll see that in a second, but that's, you know, that's really what it started as was we took something that wasn't working that great, didn't see any vast improvements on, was kind of hard for both us and our customers to use. And so, you know, having vibed good at a few different other apps already, we thought, you know, we could probably make something better. But it did evolve from there, right? So I say that as a, if you see everything that Kiwi does now, that he didn't start that way. And we built him in January, so he's had almost 3 full months now of like, I add things to him all the time, every week. Um, sometimes it was daily for a while in the beginning, just as folks were starting to use it in production. And so he didn't start this way. He didn't start fully agentic. That was actually not even my end goal. I didn't even think about that. I just wanted something better to replace something we were using that wasn't all that great and was kind of causing some friction. And I thought we could save time by doing this. I didn't realize till after the fact how much time it would save us, how involved QB could get with our customers, how much of a better experience it would be, not only for us, but for them. And so yeah, that was kind of some downstream effects I really did not engineer in the beginning on purpose. That was more a, because it was agentic and built as a native AI platform, we could do different things with it. And so it just, evolved from there. So yeah, it really began as just a simple, you know, let me just vibe code a project management tool. I want to be able to, you know, my initial spec was let me just be able to assign people's tasks. You know, we're going to need some single sign-on, uh, we're going to need to make sure it has some light automations so that people can, you know, get reminders, know what their next task is, basically so that they have a guide for their success at Zaster as a customer. Sponsor of ours. So when we deployed QB into production back in Jan, that's all he did, right? So I'll preface that, like, even if you're playing around with him now, if you're looking at sasresponsors.com while I'm chatting, like, you're not going to see the full magic of it because you're not a customer, you're not logged in. And I'll show you some of those things in these slides. But that's all he did for a couple of weeks. And that was fine, right? That was absolutely fine. I was like, you know what, this is already better than what we were using previously, which was like a SaaS tool we were paying for. This is already better than what it was. And so I was happy with that. But then I realized he could do a lot more. So a couple of weeks ago, I basically had a moment where, you know, for years— and I'm sure you guys do this too, or many of you probably still do this now— I started doing— I started to write my weekly sponsor email, right? I was like, okay, it's Sunday. I usually send this on Sunday night or Monday so people have it before they get too busy of You know, as a customer of ours, what is basically their tasks or deliverables? What are high-level things they know? Like we launched the agenda or like attendance is 140% of last year. Like I include some like context at the beginning, but then it's pretty much like a laundry list of the things they should be doing, need to be doing, doing upcoming, and then just various information, right, about being a customer of Saastr and what would make them successful as a sponsor of Saastr Annual in particular this year. Like you know, how to register their team, how they could do customer passes, how they can market their presence at the show, booth locations, all that kind of like periphery stuff. So that was all built into the initial spec, which I'll show you again in one of the slides. But once we actually— and this is, I think, key for folks to remember before we get into how we do it and like reverse engineer it— is again, it didn't start this way and it definitely stair-stepped it once we deployed to production. I think a lot of of what we see is people buy good apps and they never get to production, or they get to production with like 10 people or just their internal teams, right? Like our AI VP of Marketing is just for our internal teams. And so it's a lot, it's built out a lot different because it never needed to be forward-facing. I think one of the things we unlocked here was like once we went to production with our actual users, actual customers, and they started adding more people on their team, we, that's when I was like, oh, we can do a heck of a lot more. With QB since he actually is, he has agent capabilities we just haven't built into him yet because that wasn't his original spec. But we could do it right. Like now that I see people are actually using it, now that we see how they're using it, like we started to see, okay, who's logging in, who's not logging in, what are they doing at certain times of day, what are they not doing? We just didn't have that level of granularity before, right? Like we would just have whatever they submitted to us. That was literally it. Like, our old portal didn't even tell us if they logged in or not. Like, it was so hard to use. We had like no insights on anything. Um, and it was also one of those things where if one person on the team of a customer of ours submitted something, like if I submitted something, Jason couldn't see it. Like, it would— it wasn't persistent and had like a lot of flaws in it. And so we just never had a lot of visibility into what our actual customers were doing until we had Quby because we built it in a way, um, you know, we use Clerk in this instance. You can use other things, but between how we did single sign-on and how we built it into our Live-coded app, we got a whole lot more granular visibility on what people were actually doing and then again what they weren't doing. And so that was—

One way maybe, Amelia, just to help folks on the screen, this chart on the side, this task completion. Amelia will go more into this, but to simplify this, this is basically a sophisticated onboard automation. Okay, there are, and, and this may be more complicated than some of your onboarding or less depending on, on what you're doing, but there's a lot to do, but there's essentially 13 core tasks with subtasks that each customer sponsor has to do. And you can see rather than hoping that a human does it or an agency or a third party, Qubi just does it, manages it, reaches out to them, tells them all the deficits, kindly reminds them, identifies the gaps, uses AI to anal— analyze all the uploads to make sure they're the assets and data that we need, manages it each day, and then pushes a Slack update and an email update every single day with the gaps. So it follows up, it manages it, does it all for us. Um, and this, um, I, if you have the world's perfect group of humans and they do every task they're supposed to do every single day and they update your task management tool and it's all there, it, I guess, I guess that would be as good as this, but we could never find a core of humans that wanted to do these hundreds and hundreds of onboarding management every day, especially because for us, our customers, our sponsors, a lot of times they don't want to do the work, right? And that's why you see it. And this is why it's so clever. Um, you know, a lot of the onboarding work customers don't want to do. They get busy. They, they, they don't know why they have to do it versus you do it. And QBee just manages it all, uh, without judging or drama every single day to completion so that these progress meters all get there. And It's fabulous. And this is— Amelia went through it, but if you think about, hey, do I have any gaps like this either on my onboarding or post-launch to manage the customers after? And one of the reasons QB really means two things, but one of it is like a QBR, but a QBR every day instead of maybe someone does it once a quarter. QB is checking in as often as the customers can take it every single day to make sure things happen. So think about, could you use an Agentic product to make your onboarding better so everything happens and everyone's touched? And can you make post-onboarding more successful? That's what QBee really does for us, for context. And that's what you see summarized in QBee itself here. And Amelia will show you more of it.

Yep. Yeah. So it was kind of, you know, it was a big unlock for us to see like something like this one we previously didn't have this level of insight into, right? Of like, okay, at a glance, this is my admin view, right? And this is in the email that our whole team gets. So like, Jason sees this, David on our team sees this, our production team can see this, like, okay, these are where folks are at. And like, you know, for some of these people, I'm like, okay, I just, you know, some of our diamond sponsors, I get it. They have, they have agencies they're working with. They're getting the agencies, you know, logged into the platform. At the time I took the screenshot and so they were like a little bit behind. But again, I still think it's good to keep them honest, right? Like usually our top tier sponsors always also have more deliverables than others. So it was important to see that, but this was something, again, we previously couldn't get this level of insight, but then the big unlock was automating this insight, right? So like to Jason's point, making sure that these folks would get the emails as often as they kind of wanted or we thought they needed at this point to get their things done and also be successful, right? Like the whole point is to help them deploy it faster. In a timely way that helps them hopefully do it with like less stress. So for us, you know, this is, this is what it's become now for QB. And then to Jason's point, I went back and looked for this time last year because we did SaaS renewal in May of last year, basically around the same dates and weeks. I looked at the amount of human hours and inputs and the amount of emails we sent last year for, you know, the first quarter basically. So January through March 31st yesterday and the same first, you know, Q1 of 2025. And it was dramatic the amount of hours we've reduced not only internally but also with our agencies and production teams. So that's, you know, kind of a 3x multiplier there of how much human hours was needed. It was dramatic, right? Like Basically across the board, I saw something like a 70% decrease of literally billable hours. Like these are, again, these are agencies and production teams we work with. You may have something equivalent with like, you know, CSMs or internal teams. You may do that internally just to make it more, you know, comparable for some of the folks on this call. But again, that's a huge difference. Like it's huge. 70, it's a lot. I'm like, that's, literally thousands upon thousands of dollars each month for the last 3 months. And it compounded, right? Because January, maybe folks aren't as caring about their SaaS or annual sponsorship that's in May. You've got to think just naturally in February, March, more people cared about what was going on with their SaaS or annual sponsorship in May because it was coming up. And you might have a similar where, you know, maybe getting them deployed is a little bit slow at the start, but they've got a big launch or something, or you've got a big release coming up where you want to push them to sort of get to a finish line, right? And so that was super important for us to see that so dramatically, just such a huge dip. I'm like, okay, if you, if you even weighed out the cost of us building QBee versus the amount we save, it's not even comparable. It's like apples to oranges. Like, you know, I know people sometimes drag on all these live coding apps take money and all that, I'm like, yeah, but it's a couple bucks, you know, it's probably— we've spent, I don't know how much now on Kiwi, maybe a couple thousand, but we've saved tens of thousands of dollars on human input hours. And so you do come out ahead in the long run. But not only that, I don't even look at it at a pure cost basis, but just from our team's perspective of how much we've needed to input with folks has decreased alongside the amount of like physical hours it took. And so for us, that's like a big jump in how we were operating. Communicating with our customers in the past.

Hey, just one note, just because I should have written more about this and I want to keep going, but on cost, because I got this in the chat, I get it a lot. Well, how much does this cost, right? Beyond what we pay for Replit or other platforms. I put, I don't know what the AI costs are on Qubi, but I put a total $200 a month cap on all token usage across these apps. We can afford more, but I just wanted to not worry about it, right? That's the allowance. This is for Saastr.ai, the pitch deck uploader, valuation calculator. It can, it can, all the apps we've built for all times, 10K QB, and we haven't hit the $200 a month cap on AI use yet. So just as we could talk about it more, there are certainly applications that will burn lots of tokens, you know, massive coding sessions. And, and, and, but within these apps, within all the apps we've built, we have not hit $200 a month of tokens. So just realize for many use cases like this, the soft costs are material. You got to build it, you got to maintain it. Maybe we'll touch on this or do a session. We're going to do another one on what it takes to maintain these apps. But your hard costs for AI are probably going to be pretty low for 95% of these apps.

Yep. And there's some good questions in the chat I'm going to get to next. But I just wanted to preface of, again, I think I see some good comments and chatter in the chat. Of, you know, thinking about this as an unlock for, again, it's gonna cost you more at the start to build it with anything like a Replit level Vercel V0, of course, cuz you're spending more time with the agent. But I'll actually show you a couple things where you could probably get it down. Um, things like giving it a website like ours, actually super helpful for your agent. Like if you just gave it a spec and know like, hey, I actually like the design style of this website, or I like some of this functionality. And like the more time you build— I'll get into it next— but like the more time you build out that spec to be as granular as possible, actually the better your agent will be and it will keep your cost down because then you don't have to keep iterating with it. But, um, that's just a little bit aside. Okay, so I'll show you a couple more things and then I'll show you how we— how you can reverse engineer it. So, um, this is also another good example, right, of how QB actually went from customer/project management tool, a little bit of both, to fully agentic. So Back in the day, I put a little then and now on the screen here, but back in the day, right, again, static, no single sign-on, no way we could really track anything in any sort of, not even a dashboard, not even analytics, nothing, right? Just kind of hope for the best. So minimal reporting required a lot of manual human hours, as you saw, that we had to supplement either with ourselves or with our, you know, outside third parties, which cost, you know, significant amount of monies. And then when we email, this is the important thing too, When we emailed folks of what to do, for the last 8 years I've been at Saastr, it was always newsletter style, even for customers, right? Even for Diamonds sometimes, which now I would never do. But the, like, the set of information, again, not individualized, not customized, due to either lack of time because things got really busy at Saastr or lack of want. And I think this is important too, because Jason put this in the article, like, I tried to get more of our agencies in years past to do something that QBee has done now, which is why don't you take every single customer of ours that's a sponsor, email them, you know, their registration codes. Every, every sponsor has a set of at least 4 registration codes, so they already didn't want to do that. Um, why don't you also, you know, give them a little update about where they're at in either, you know, if they still need to pick their booth or not, where are they at in the queue if they haven't, um, logged in yet? Can we see that? You know, we were trying to basically finagle the data, which nobody wanted to do because it wasn't all in one spot. So like, I literally just couldn't find— I couldn't find people to do this. I literally tried. I could not find people who were like, okay, I can sit down. You have 100 sponsors. I got 4 links for sponsors. That's 400 links I got to go and find that. I got to see if I can figure out if they've been invited to the portal or not, sort of manually. I can't really see if they've logged in unless they submitted something. That's really the only way to tell. And they didn't want to do it. So like in years past, I've had kind of hit or miss on folks actually doing this with our customers. And when it did happen, I would say it only happened maybe once and it took that person a week. QB in the screenshot on the right, you'll see sent all my sponsor emails for the week, individualized, highly personalized, highly customized in minutes. Literally, you can see these are all at 11:11, 11:12 AM. You know, he's got a go to resend and, and send these. Um, but that aside, Kubi was done with this in 10 minutes. 100 people got a highly customized, personalized email saying exactly what they needed to do, exactly where they are, unique sponsor and like updates. Like we have people who don't do booth attachments, or they do things like lounges, they have custom deliverables, or they have extra signage because they're on like the lanyards. They got all that. I didn't— they didn't have to go look for it. Like I didn't have to have a human go look for it or forget that they were the lanyard sponsor. Like QBee just remembers all this and he remembers everything about every customer of ours, every like little niche thing that they have going on. He just remembers all this and he does it instantly. So again, just a huge unlock here that it's now, you know, so personalized to the, to the degree that our engage— I put it in a previous slide here, but like our engagement to, to our users, to our customers has gone up more than tenfold. Like, we couldn't get people to really log into the one last time, and to some degree I was like, you know what, it's fine, like, we'll just send them the registration codes, they can just upload, they can just email us their booth graphics. Like, that's how people were doing it last year. Um, now I have, outside of the screenshot you saw, I have basically everybody but 3 people who've logged in. That's a dramatic uplift in how people are also like interacting with us, um, and how they don't mind that it's an AI. I don't even know if they've realized until this webinar that most of it was an AI and not us. I think a lot of people thought it was me and David. They were like, because we're copied on QBee on the things that QBee does, they're like, oh, hey, or they'd be like, hey, SaaS or team. I'm like, yeah, you're just talking to QBee right now, but that's okay.

I think you hit something that I didn't fully realize. So you said it, this unlock, which is think about on the other side when you're, when you're, when you're buying a product, right? And then some some vendor sends you some another product you've got to log into and train and set up. Like, you just don't want to do it, you know? And the fact that Qubi, uh, has epically higher compliance means even if it wasn't better than the off-the-shelf software per se, it would be worth it alone because it removes— we've been able to build it, you've been able to build it such a way it removes all that friction, right? It's not another, you know, you— when someone like— I was asked the other day by one of our partners to log into Asana and set a bunch of stuff up for them. I'm not going to do that. I pay you 5 figures a year and you want me to create another Asana account and upload stuff for you and then use some weird, you know, Okta account to log into your Asana? Like, for— like, I'm paying you, right? And so, so many of these post-sales products, post-sales motions put too much of a burden on the customer. And so no one just does it like we No, forget about one, the humans wouldn't do the work prior to, to QB, right? But the software we use, people didn't wanna use either, right? They didn't wanna use it. So that's in line. Another reason to, to do it and to vibe it yourself is that people just won't use the cr— the, the, the now mediocre software, uh, because you're, you're putting, you're asking too much of the customer. You're not making their lives easier, right? You're making their lives harder after they give you money.

Yep. Okay, so this is a, this is just a quick example of one of those emails I was just prefacing. But you know, the screenshot before is basically QBee doing this all in a matter of minutes. And then here you can see the before and after, right? So here is what we used to send people again. And this is literally, I took this screenshot from March 3rd, 2025. This is just a year ago. We are sending it newsletter style. You can see it here. Here's a generic link to our portal, which most people didn't bother to log into. Um, here's a link to— we do also a weekly webinar just for sponsors that I link to folks to, you know, booth selection is kind of dated here. It says it started, we'll email you once the window opens. I'm like, now everyone's picked their booth. Activations. This is a huge pain in our butt that we've been able to automate with QBee separately. Um, things like travel schedule, all that stuff. So again, very generic too. If you go to the one on the right and I picked one of— I picked TikTok, who's a newer sponsor of ours. Um, they've, they've, I should say they sponsored in the past or they're, they're newer signing on to Saastr Annual specifically. Um, but they got everything done in like a day. And again, this is something where TikTok specifically last time they did London was very slow to submit stuff. Big company, right? Like slow to submit stuff. We couldn't get them to really turn anything on time. I remember going to London and being like, TikTok stuff is not done. Like, well, I don't have logos. I got nothing. And so even just a dramatic increase of, TikTok went from being kind of a pain in the butt sponsor, cuz I couldn't get them to do anything to they literally uploaded everything, almost everything. You could see they did, they skipped 2. They still have to do 2 more things. But they did almost everything in a day. Again, just made, made it so much easier for them. This is, you know, screenshots of their email. They can see exactly what they've already completed. The last 2 things they have to do. Their ticket codes are in here specifically. That's not in the newsletter style. They click this, it goes to their their TikTok-specific registration codes. It goes to their TikTok-specific discount code. Again, just things we couldn't do. All right, so I'm gonna get into how we deployed it because I see you all asking about that in the chat. So if you want to reverse engineer Qubi and make your own and deploy your own similar agent, here's what you should do. So let me back out of full screen for a second. The— and if you're wondering, the slides I did put kind of in the chat. Closer to the beginning, but they are there. So I see you guys here, but, um, yeah, once you access it, first write a spec. So you can see here, I actually put two things. I put my original spec, which is the one on the right. Um, so you can see it with Claude. Um, oh, sorry. Claude opened on my cowork. Oh, let me share my whole screen. Of course he did. He was like, you have cowork. Um, All right, so this you can see is my original spec. It's actually not that good. I looked at this the other day and I was like, it's terrible. So, but it's still a good example. So if you've never written a spec for a Vibe-coded app, this is still a good basic level example, I would say, of how to write a spec for your Vibe-coded app. Basically tell it everything you want, right? So when I was going to build QBee as a portal, I thought, I was like, okay, he needs some user flows, he needs a dashboard for me, he needs a checklist for all of our customers, he needs to be a hub for everybody, right? All the information they need. So for us, that's logistics details, any speaker sessions, etc. He needs like an asset library, right? He needs some sort of training. He needs to be able to have uploads, single sign-on, everything in their contract, you know, QB should know so that he's never like, okay, I'm going to send— like, he— I— what I didn't want is like, okay, we have some sponsors who have speaking sessions and some that don't. I didn't want him to send, you know, somebody who didn't, an email saying, hey, time to submit your speakers. That would suck because then that person would be like, oh great, you're giving me a free speaking session. No, um, as much as I would love to, we don't have that much time at Zapster. So there was some very important nuances there, and I'll show you how I did it. I did not load those contracts directly into the portal. Say that right now, do not load your customer data directly into the portal. I'll show you how I agent hop later, okay, for a security layer, but just a preference there. Things like badges that you just saw, registration, FAQs, support information. They can actually do add-on orders. Anyways, you can see everything I wrote here into it, the technical requirements, blah, blah, blah, blah, blah. You know, it needs a database, it needs single sign-on, we use Clerk. MVP and future phases, and then all this. Okay, so that was V1 of my spec. Then I asked QB on Monday to write me a new spec. I said, hey.

Sorry, just to be clear, if that spreadsheet, there's a link, you can see it. If that It's okay. Just start in Claude. Don't start it.

Yeah.

And just say, I need help writing a spec for a sponsor portal and it will help you do it together. You can do it in the Vibe Coding apps, but we both have found you get a double benefit if you write it in Claude and then get a spec out of it and then put that in the agent and then kind of tweak it again. Right. That's how we like to do it. But you don't have to know how to do all the bullets and everything that Amelia said. Just, just iterate it. Right. And you don't need And to be clear, if folks haven't done this themselves or they haven't done it in a while, in 2026, you don't need to be a prompt engineer. Like, you don't need to know— like, that is dated when you need to be a prompt engineer. Okay? You just need to talk with Claude or I guess ChatGPT. We prefer Claude and just say, I want to build the spec for an application. Here's what I want it to do and write up all the things you want to do and just iterate it. You don't need a prompt magician. The agents are so good. AI is so good that that's dated. Iterate it, make sure you understand everything it is, then give it to your vibe coding app and let it run with it.

Yeah.

Amelia, go through it. But yeah, it is not— this is— you do not need to be an engineer or some prompt wizard in 2026. Those days are long behind us.

And again, that was my simple one. That was the one I started with. If you're intimidated by it, then just use that one, right? If you're like, okay, I'm just gonna try to recreate Amelia's first original thing, that's fine. And that's fine. And the thing is too, you can give it other people's prompts. You could say, hey, this is the one that SaaS did. You could say, hey, this is the one that SaaS did. Here's the prompt that they used. Here's the website that it ultimately ended up as. Give it sas4sponsors.com. You will save yourself a ton of time from the work I already did. So do that too. But second, here's my— I asked KB this week, right, to give me what it would consider the spec now. And you can see it's a heck of a lot longer. It's got a lot more to the tech stack because we added some agent hopping stuff. Um, it's got some more theming and design. It's got some more nuance basically, right? Because like I said, as we started using it, we saw that we could use it as more of an agent and less of a straight up portal. So you can see it, there was literally nothing about emails in that original spec. Now it has like a whole schema for emails and email blasts and how it's going to route things to people. Um, public faces versus, you know, not public faces. Um, like the task lists are more built out. So. If cost is one of your concerns, I would say you want to get your spec closer to this one than my original one. Because again, benefits of doing this in Claude versus doing an agent, at the end of the day, it's still going to be a little cheaper to write this with Claude or ChatGPT, um, more fully baked out for you than it is to put it into any of like, you know, the major vibe coding apps directly into the agent. So you guys can purchase this, it's linked in the, um, it's linked in the slides. So you guys, you can use it. Um, again, if you want to use it to your benefit. So, all right. So first things first, write a spec. Again, the more granular you can get, the better, but this is not how my spec started. So if you don't get it all perfect, it's fine. The bigger thing here is to get it into production. That is really the goal. It's getting into production. Specs aside, but specs do help. Um, two, right? Okay. So for us, we took our fave, we took our spec, we loaded it into our fave vibe coding platform. Again, you can use Replit, Lovable, Vercel v0, any of them, they're all also gonna be at SaaStr. Um, so you can talk to them there too, and they will show you how to do it. If this intimidates you, they're gonna all be there in May. It's only 5 weeks away. So you can also just do a follow-up there with them. Just go to their booth. Um, but what you're gonna do next is you're gonna give the agent and your 5 coding platform, um, the spec. You know, you can give it some high-level design. Again, I gave it a website I already liked that we already had. When I first gave it to the Replit agent in this case. And I said, here's my spec, here's my kind of like websites that we already have going on. 'Cause I like the design, I like the look and feel. Here's some other, I saw other people's like customer knowledge bases that were really good. I was like, these are the ones that I wanna copy. And so give me, hit me with your best shot. And Kiwi does not look the way he does today, but he looks pretty good. And so the first thing I did, and this took the longest amount of time, is I had to test literally every single thing. Function, everything, sign in, upload your logo. You can see here, uh, submit your address. I tested all of these one by one to make sure that before I sent it to our first customer, it wasn't gonna break. 'Cause my biggest fear when I was rolling this out was this all seems really fine and dandy, but what if it breaks? Like if it doesn't actually get into production and people can't actually submit to us the information they need or get information back from us. Agent or not, then what is the point, right? It's got to actually freaking work in production. And so I tested everything one by one, basic or not, using AI or not. I tested every single input and output in QB. Some folks are asking in the chat, can you see QB in action? Just go to sashresponders.com. You could see it. Now you won't be able to log in, but you could see what he's doing. And then we had to hook up email capabilities, right? And at the time I was like, I'm just going to hook up email so I get an an email, but then obviously we realized he could email everybody personalized. You know, create a database. Most 5 coding apps will do that by default. Sometimes you just need to make sure you make one. You know, port over any content, any knowledge you want. Again, this is a big thing. Sometimes you can crawl your website if you need to move some, you know, if you've got FAQs lying around, that's just like a Google Sheet file, it's fine. Just give it all, give it all to the agent. Except You know, sensitive data. Don't give that to the agent. Give everything else to the agent that's not going to be sensitive. But yeah, I literally tested everything. And then this agent hop thing I've said a couple of times is for sensitive data, never give it to your agent directly. Never, ever, ever. Like I did not. Again, this is a— we call him QB now because he does real-time QBRs. But at the very beginning, when I wanted him to know all the contracts, all the customers, who was going to need access to the portal, I just did a Salesforce integration. I was like, all that— one, all that data is already in Salesforce. We already marked these things as closed one with the contacts when they close. So one, I'm not gonna have to reinvent the wheel if I just do a custom— I forget what it's called, it's in here somewhere. Um, I just did a custom Salesforce integration. Two, then QB directly does not have that data. He has to call up Salesforce. So he doesn't have, you know, 100 contracts just sitting in his knowledge base, which would be scarier. So I definitely agent-hopped his, his original, his original customer database and ongoing database is mostly in Salesforce. And then in terms of like specific users and people who folks add, again, we use Clerk for single sign-on. So all that lives in Clerk. He doesn't have that directly to the point where sometimes He forgets who's in Clerk, or actually he doesn't forget who's in Clerk. He just does this annoying thing, which is super nuanced, which he sometimes will not send it to people that are pending in, in single sign-on. He'll be like, I'll send it to Jason and Amelia because they logged in as part of the Salesforce team, but I won't send it to Justin because he hasn't logged in yet. And I'm like, no, no, no, you need to send it to everybody. But anyways, that's how I agent hop. There's other things like, you know, all the registration links. That's from Ms. Bo. So he doesn't have that information directly either. So that's just to add on to that.

I think that's a good best practice. I think the agents, we try to, we use Salesforce as our hub for all our agents. We've talked about it. And so there is inherently an additional layer of security by having that be our system of record. It's not perfect. The agents are still going to touch the data, right? It's beyond the scope of this. I think that if you need to store sensitive information in your app, it's okay. Just realize the more you do, the higher the burden it's going to be on you to become a security expert, right? Which you don't want to do. You're going to— instead of testing things once or twice, you're going to need to do constant security audits, constant reviews, try to break things. And we can do a different session on what we've learned there. I just don't think most folks want to become security experts. So to the extent you can, if you have your most sensitive data, like for us, we just don't want contracts leaked, right? We just don't. Would it be the complete end of the world? No, but it'd be, it'd be bad, right? So we store them in Salesforce, right? Instead of the agent. I think that's a good practice. But if your use case, that doesn't work, I wouldn't say don't do it. Just realize you're going to have to spend a lot more time than just using the default way the app is built. You should budget hours and ongoing maintenance to making sure your app is secure as you can make it. But I wouldn't say you have to do this agent hopping, but I think it's a great best practice.

Yeah, it's, it's— I know it's a, it's a new term. I don't think I coined it, but I don't know. I just call it agent hopping. I'm just— it's— and it's not— again, it's not that it's anything magical or that I'm— in this case, I'm not even calling up Agent Force, which I think some people thought from the name I was calling up Agent Force. No, I'm just hopping data with all my different AI tools, agents, whatever you want to call them. Again, so it's more secure in that nothing is ever in one spot. And then we do, you know, we have like Zapier keys that are secretly stored, other API keys that are stored that, again, we just load different sets of data and things from other tools that are, that are more compliant and more secure than the built-in. Agents. So, okay, so that's just one, two, number three, and only go through this period. Just pick how you want to deploy logins, right? Um, if it's something like this where you're going to have customers logging in, listen, it was a pain in the butt to do this. And candidly, um, I think it's gotten easier now. I know for a fact it's easier and lovable in v0 because they have it built in. Um, but again, just decide how you want your single sign-on system of record and decide how you want that data to get there, right? Maybe don't load it all in one spot. Again, I segment our data a lot so that it's always hopping between different things to call, you know, it's got to call all these different APIs to figure out the full picture. So hopefully that's more secure, knock on freaking wood, than most things. All right, again, I've touched on this, but maybe more importantly here, What the biggest thing on a, on a, on a vibe-coded app like QB is to deploy him. So I just deployed him to a few folks at first. I basically picked one customer at each of our levels and I said, okay, I'm going to deploy to them. I'm going to see what breaks, what doesn't, and definitely stuff broke. Definitely some stuff didn't work. Like I learned in the first week that, you know, we had a, I think it was a gold sponsor was submitting some tasks and They're like, it's not really working. I'm getting this weird error. I was like, okay, send me a screenshot. Because apparently as QBee's master, I also have to do all the troubleshooting. So I was like, dang, downsides of this is I can't just send it to someone and say, hey, fix this. I was like, oh, it's me fixing it. But I was like, send me a screenshot. I loaded into the agent and I was like, hey, what happened? Why is this like, why is this person getting this error before we roll it out to everybody? And they go, oh, Uh, it's cuz you did not build in a timeout. So because this person was like, they logged in the first time and it just never timed out their login. They were just logged in for like 5 days at a time. When they went to go actually upload something, it broke because it basically timed out. Like it didn't reauthenticate them. And I was like, oh, again, not something I thought of. So I'm not perfect, do I put, I guess, um, not something I thought of. And so once I figured that out with the agent, we built in a, I think it's like 15-minute timeout now so that, you know, you sign into SashResponse. If you don't do anything for 15 minutes, it'll boot you. Um, so that every time you log back in, you're reauthenticated, then you can upload stuff. Again, things I learned from it actually being in deployment with people, not things I, I didn't even test for that. I was like, I don't know, because I was testing every task all at once when I was logged in. Again, not something I came across. Um, and so that was really important to see. What did and didn't break. And then also to see what people wanted to add to it. Like every week we got like new requests for like, hey, do you guys have email marketing copy I can use? Do you have, you know, uh, actually people, stuff like that, right? They were like, uh, do you have more information on networking? That's always a big question we get. So we've made a whole section for networking in QBee. Like we just kept adding to it, um, and adding to it. And then people were like, oh, I want to just submit my activations to QBee and not email you like I used to have to. So then we built a whole submissions and then we built a whole speaker submissions path where for the sponsors who had speaking specifically, you had a whole section of Saastr sponsor speakers and it would give you exactly what you needed to submit all of our speaker guidelines. But also, and this was magical, it would let you pick a time, which I know sounds really basic, but, um, QB would let you pick a time when you were submitting your speakers. And then when he saw that there was like 3 or 4 speakers at one given time, he would gray it out. So he'd say, okay, like, listen, all the sponsors are going to submit this task on the same day when it opens because they all want to speak on Wednesday. They all did. Okay. And so QB would say, all right, I've already had 4 people submit in the 9 to 10 o'clock slot. I'm going to gray that out so that the next 5th person who does that can't pick 9 to 10 because it's already going to be full. So like, again, things we had to do in an email back and forth that cost a lot of time that QB just did agentically because he's built that way. Um, so we added again, How Qibi is now is not how he was, and he won't be that way in a week. Like, now that we're getting closer to SaaStr, I've gotten more of these requests in the last week for even— here's the thing, now people want even more out of him. They're like, oh hey, could you add this? Could you add that? And I'm like, we're so busy right now, but yes, I will add all these other features and functionalities to Qibi because I'm glad that you guys are using it again a lot more than you used to. So, um, all right. Number 5, again, if you're going to reverse engineer something like this or build something like this for yourself with the vibe coding app. Is again, understand where you can leverage the data you're going to get out of this to be more agentic. That's basically how we turned Kiabi from a gen— somewhat generic, um, uh, customer success portal slash project management tool into an agent. Like, he still does those things. He is still a project management tool. He's still a customer management tool, but he's also more of an agent now, right? But again, he didn't start that way. He didn't get that way. Um, you know, it's, it's understanding where you can leverage things like this that are built this way that has different APIs that you can hook for yourself to say, okay, how can we get him to do things automatically? Even if you don't want to say agentic, just automatically, right? Like in this day and age, again, are you doing things like sending unique links to people when you're onboarding them? Or, or do you send them custom kind of things throughout the day? If you're like, hey, I saw. You're not using this feature that we just rolled out. Here are some tips on that. Like, let the agent do it. Um, are you sending reminders, workflow, or are you not sending reminders and workflows that you want to be when they log in or do, do or don't do certain things in your SaaS tool? Let the agent do it. Um, are your CSMs or aren't they checking in regularly with people? Because guess what? Kiwi can do it as many times as they slash you want. As Jason said earlier, that is one of the most magical things is he will do it whenever you want it to, or you can let them pick, right? Like, hey, how often do you want an update? And then maybe just force, force some emails or updates or however they want it. Um, if they have things to do, like honestly, QB this week has started doing things like collections. So I was like, yo, you need to, uh, get on this because Or humans are really slow and SaaS is coming up. So there's other things that you can build into it. But again, if you see something's not getting done regularly, the agent will just do it, man. Like it will do it without question. It will do it very quickly and it will do it for all your customers at scale. Uh, you know, we did this for 100 people. You could easily do it for 1,000, 10,000 easily. Like maybe you need multiple agents to do it or. Multiple QBs, but it can do it, right? I think too how we evolved QB into the name was, yeah, a little bit of the play on the QBR, but QBRs have always pissed me off as a customer because like, why am I waiting once a quarter for super generic stats? You're not telling me anything in real time. Everything in a QBR has already happened. Whereas QB, the agent, is in real time. It is what's happening now. It is not waiting for a magical quarter of a window to pass before I give you any sort of freaking insights. He is now, he is real time, he is live, he has access, he has all the data. Like, wouldn't you want that instantaneous versus once a quarter before you know this customer is going to maybe churn from you because you don't know? Like, I don't know, I feel like there's a lot of power in that of, you know, again, we, we were guilty of this too. Like, last year we waited till after SaaS Annual to check in on some things, and now QBee can just do that before the event. He's just a lot more proactive. And so again, it's not where we intended him, but that's kind of where he is now. Um, and I think that's really powerful to see with our, our customers, but it's— I'm sure will be powerful for your customers too. Okay. Um, agent hopping, I'll talk a little bit about more here, but yeah, you know, again, we don't store everything in the native Vibe Coding agent. We store things in different tasks through the API. I built out a custom Salesforce connected app Had I done that before? No. But I asked the agent and Claude how to do it. I was like, hey, actually our Salesforce integration kept breaking and I was super frustrated. Jason knows this. I was like, every day this Salesforce thing keeps breaking and then he forgets a bunch of stuff that I don't want him to forget. And so eventually I was like, hey, agent and Claude, how can I build this so that he doesn't keep breaking and that he has data more persistently? And so it led me to build out a custom agent, sorry, custom Salesforce connected app. Took me maybe 20 minutes, wasn't that bad. And now he doesn't break. So it's great. And obviously there's data in there, but there's not that much data directly in Qubi. So that works really well. Okay, last couple things and then we'll do any sort of last Q&A till we run out of time. The Agentic piece, I built out again, one layer at a time. It started as, okay, let's just do, you know, an email to me on who hasn't logged in or not. And then I'll add the rest of the team. Then I was like, oh no, we could do one for each sponsor. Then we started doing one for each customer and they got, you know, they're getting now, as they can attest to, if they're, some of them are on this call, I saw, um, as they can attest to, they get these weekly emails from QBee, um, that that David and I are copied on with the sponsor production team where it lists them out, everything they have or haven't done, what's due, what's upcoming, things like that. And so again, it just, it will just trigger certain actions. It will also tell me when people submit something. So this week, for instance, all of the booth graphics are due for everyone's booths because it takes time to produce those things. And again, we're like 5 weeks away. So it'll tell me every time somebody submits their booth graphics, which is great. I go through, I check them. Actually, previously I used to just check them myself. Now me and Claude co-work, check them with Illustrator on my computer, which is also very nice. That saves me a bunch of time as an aside. But again, for our internal team, it's just so much better visibility for everybody. We just share this with, you know, our production teams that need to get like a basic report. Folks who are new can onboard really quickly now versus very slow. So A lot of agenda pieces there, but I didn't build that all at once. We can touch on some of this, but anything else? I know, Jason, we're kind of running out of time.

I think this is great. One thing just to remind folks, if you can build your own version of this, hopefully that helps.

We've—

the slides are there. You've got how we built them. You do have to maintain these agents daily. We say this every single time we do it. but there is no set and forget, right? Whether it's little, whether it's things that will break, like the initial version of the Salesforce integration kept breaking, whether like today, like it sounds like 10K gave triple updates in Slack instead of one, which is a recurring issue. We know why it happens. Um, or, or there, um, um, you will see regressions. For example, this, if you haven't built software today, Amelia tested every button, every link, every upload in the app. Right? As your agent gets more complicated, as you add features, it may break those links and upgrades, even though you, you'd be like, why? I added a totally different page. Why did it break the upload button? Beyond the scope of today, these regressions will happen. It is life. And so you need to budget not only just like when we've talked about our AISDRs, have you have to review their emails every day for a long period of time, you have to do it when you build your own agents and your own apps. Every day you have to check in on how they're working, check in on their functions. And my hack is make sure you have them send you an email status update every single day. Have them build that because you'll see stuff break in the email and then you'll realize things are broken that you can't see. But just realize anything that's as good as QB, it is great. It has replaced many humans that would not do the work. It gives us 10 times the scale, gives us everything. And between Amelia and me, we're checking in on it every day. Right? Just like you're humans. And so just realize if you want to do this, do you need an agentic deployment expert as I've talked about before? Maybe, maybe not, but you need a human who cares enough to manage QB and 10K every day or it will fail for you.

Yeah, it's, yeah, it's really like, I know I put some before and after slides in there, but it's really been night and day. And that's why at the end here, I affectionately call him QB1 now because he is literally a quarterback. He is our, whatever you wanna call him, our AI VP of customer success, QB1, QB. Like this dude is what I rely on now. Like I, I can't see myself going back to how we were doing it. I wouldn't want to. But also I think just for the lift we've gotten from our side of the fence, but also as our customers and sponsors have seen where again, it's more tailored. Um, people can still talk to, and again, Maybe I didn't say this enough in the first 58 minutes. People can still talk to us. It's not like we said, hey, here's just QBee and that's it. We weren't like, we're gonna drop you totally into the AI and if you can't solve your problem, you're on your own. Like there's a reason we're copied on every single QBee message that goes out to our customers because when they reply to it, sometimes we reply. Sometimes it's like something nuanced of, yeah, to Jason's point, Today, like somebody was like, hey, I submitted everything to the portal, but it looks kind of funky. And I was like, okay, let's go take a look, right? Like, it's not like we're dumping them solely onto the agent to deploy, to be successful at SaaStr. It's something where we're backfilling. We're very supplemental. You know, sometimes they do want to talk to us through ideas of what they want to do at their booth or a lot of our Diamond Sponsors. We have still, we still do weekly calls with them. I still do weekly webinars for all the sponsors I get on. I probably was on 10 sponsor calls last week and I'll be on another 10 this week. Like we're still doing all those things. So I do wanna say that too. Maybe I should have said it sooner, but it's not like we drop everything in customer success. It's just we're doing a heck of a lot more of it with QB. And I feel like that's also true just from the fact that people keep asking us to add stuff to him, right? It's like, okay, so it's, it's not only helpful for us, but it's become more helpful for them because now folks actually ask for features and want to get more things out of it in a way that we previously wouldn't have been able to do and serve because we would just run out of time, or I just couldn't find people who actually would send 100 sponsors 4 unique links each week, um, and keep them on task, right? We just couldn't get that done. You're on mute, Jason.

Is one of the amazing things, guys, if you decide that there is a use case where it's worth your time, and again, it has to be above the line, it has to be worth— if you have If you can do this all in Notion and it works great, don't build it yourself. Don't— we can't, you know, the 90/10 rule applies. Buy 90%, vibe 10% at most. But if you decide to build something like this and you get it into production, you get it working, you invest time, the magical thing that Amelia said, it's so magical, is that unlike off-the-shelf software, when a customer asks you, hey, could you do this? Instead of saying no or asking your CS person if it's on the roadmap for the next 24 months, If it's not that complicated, we go into Replit and just say, build it now. We can ship it that day and we're not engineers. To say that is magical is an understatement. Anytime with, with off-the-shelf software, someone asks you for something that isn't obvious how you do it in the app, the answer has to be no. Unfortunately, we're stuck with what we got. So it's not a reason to start there, but if you get something pretty good into production working, your ability to enhance it for what your customers want, it is magical. That might be the biggest, the second biggest lift of all is the ability to vibe additions, not to vibe the 1.0, but to add things you didn't know you could automate before because your 8, your Replovable v0 can do it in an hour, not a second, and it needs some testing and QA. But if it's important, now you can just like anything you want that is SaaS, you can pretty much build. To, for your customers if they want it. So it's pretty, that, that might be reason number 2 to do it.

Yeah, I think too, again, we've, we have literally fought with human people who don't want to do this either internally, outside agencies. You probably fight some of your CSMs, um, and external teams on like how much coverage they've got. Like, again, maybe for maybe smaller starting up, you've got like 1 CSM for 50 accounts. That's never viable, right? Like unless you do it with an agent. Now I do think that's viable. With person plus agent, like then, then you have enough coverage. Then I think again, that's something where you don't drop it and you don't drop them just only into an agent, but it can alleviate a lot of tension between you and your customers and focus on the things that matter, like deploying things when you take the friction out of it, right? It's just the, Kube takes all of that friction away and he's gonna, he's gonna win every time because he's 24/7 automatic. Again, as often as they want updates, He can do it. Um, all right. Any— I know we're over now, but any, um, Jason, any questions you think we should hit in the chat or you want me to pick some?

Um, yeah, you can look for any, any, any, anything that we, that we missed that we didn't hit for sure.

Okay. Let me look. I go through all the note taker notes. Question is, did you create a QBeeIt SaaS for email? And will you do that for agents in the future? So QBee uses— we only said it briefly, but he uses Resend. Not perfect, but we hooked up Resend once, and so we use the same alias to reply to. It's like no reply. Sasr.com or something, um, but because QB again CCs me, David, plus our sponsor success alias, which is sponsor success@sasr.com, um, that's how people can flag to us, right? If they need more help. So again, because we're copied on everything, um, they, they can submit directly to the portal. People have rarely done that. They like to just reply to the email like QB sends them. They're like, I'll just reply to here. This is super easy.

We, I think we haven't, we only have one app where we've hooked up inbound email. So QB doesn't handle directly inbound email unless I'm wrong, unless you hooked it up, we could do it. Um, uh, it's a good question. We've done it once. Um, there's a tiny bit of complex managing inbound email is it's fine. You can do it in any of these projects. It's just more complicated than just sending an email. You have to monitor the email. There's a bunch of nuances. We're out of time, but, but, um, Right now, the humans just respond, right, to the emails that QBZ— it works. So it hasn't risen above the line to build, but we might build it. There's an argument to build it. But so far, it hasn't risen above the fold for Amelia. But certainly you can build that. You need an email platform like Resend or others that allows polling, that it itself runs real-time ingesting these emails, and then your agent has to do it. So, but you, I would actually would think folks with more scale than us certainly should have the agent respond by email for sure. That should be a core feature. Oh, for sure. We just haven't built it yet.

Yeah. Yeah. And honestly, that's just a time thing with SaaS. We're so close. Like again, the, because our customers who are our sponsors are focused on things like getting their booth graphics in this week. They're not asking for that kind of stuff, which again is, would I build it if it was over the line this week and mission critical on April 1st with 40 days to load in, absolutely I would do it. But I'm not because people just, they like this mix right now. It's working for them. Now they may ask for that in the next couple of weeks when they're done with all their shit and they're like, what else can I do?

They might miss it.

Yeah.

Right now, and it relates to another question we got, which is, do you tell folks it's an AI? I think across all of our agents, we don't tell folks it's an AI traditionally. I'm not opposed to it. I don't think there's anything wrong with it. I think our folks, I think Listen, we still— our average deal size is $100,000. So this is still a bespoke process. Amelia, David, even me to some extent is involved. We're not trying to take humans out of the loop. We're just trying to have the AI handle the 95% of stuff that can be automated. So right now our customers are fine getting an email that's a mix. They will reply back to the human if they want the human, or they will just do it themselves with QB. If your deal size was $1K instead of $100K, you would want to automate all the inbound and For sure. We don't have a chatbot on QB, but we have a chatbot on 10K or AVP marketing that could be added in an hour. So we could add a chatbot, we could do all these different stuff. But at 100K, people actually want a hybrid of humans and AI. And there's, there's kind of a meta insight there, right? People want to reach Amelia or David instantly, but they also enjoy the fact that for 90% of stuff, they can work with an agent that's better, right? They know. And I think there's a lot of our views are very 2025 or 2024. If the agent's really good, people are smart, like they understand the benefits of an agent versus a human, real-time access to data. And very— Amelia can speak to it, but very rarely have we seen that frustrate anyone. They like that combination and they get it. They get that. Well, I think they would be frustrated if they couldn't reach a human. For sure. They'd be mad. We all get mad when we log into a poorly trained AI bot and we can't reach a human. I want to blow my brains out. But we don't, we don't have— we don't need to do that at $100K price point.

No, I think the thing is too, it's in something I hadn't thought of till now too, is like with our very top tier accounts, and we're talking, you know, multi-six-figure deals here, um, like more than $100K, we go to where they want. Okay. Salesforce, no surprise, likes to work in Slack. Duh. It's their company. They have Slackbot. There's just so many things you can do. And so we were like, great, get in our Slack. So they're in our Slack. Like, you know, they have QB, they get the automated emails, they see them, they respond to them. They'll literally bring them up on calls. They'll be like, okay, Let's go over our tasks that we're missing. I'll be like, okay, great. I didn't have to tell you. Uh, QB told you. Uh, but also we're in, they're in our Slack, which I, I 100% come in our Slack. The water's warm. Like that's how they want to operate. The Google team likes to be on Google Meet and Gemini and Google Chats, and they also like to text me and that's fine. That's what they want. Right? So like, again, QB is not the end-all be-all. I should have said it sooner, but for our very top, for our, for majority of our customers, It is night and day. And for folks who want supplemental help, they can email us. We respond very— the thing is too, like, now we can respond so much quicker because we're not bogged down with all like the little nitty-gritty stuff that we used to have to answer ourselves that we can actually— it frees us up to answer the more important things when people ask us questions. And then, yeah, again, to the point of like, you know, we— I still get on dedicated calls with folks. I still do a weekly sponsor webinar and I for our top tier customers, I go wherever is easiest for them. I don't presume— I don't even, you know, QBeez is great. I don't presume— though, you saw a couple of my diamonds haven't logged in. That's fine, that's okay. Like, if that's the way they work, um, we'll figure out what works for them. And for them, it was like, uh, like one of our diamonds was like, okay, this is great that I'm getting all the emails, but I'm just gonna reply to you with like a zip folder of everything. And I was like, that's fine, but at least you got the emails of what you needed to do and I didn't have to tell you. So I think again, there's a mix there of, of part of what we've done too is it's, this has freed us up to be a lot more lenient on how we engage with our customers and the way they want. Whereas previously I would've been like, no freaking way. Like it is all or nothing. We're too crazy, we're stretched too thin, we don't have the bandwidth. Like they gotta just conform to us. And now we're going to where our customers want, which is what you want too, right? Like you want to go to wherever is easiest for them. So it's freed up a lot of our time to do that too. Going back through the questions now. Is the prompt shared somewhere? Yeah, that's in the slide deck, which I just relinked in the chat. We could throw it on sas4.ai after so people have an easy link to it. Can we see QBee in action? Yeah, I think we answered this, but it's just sas4sponsors.com. You can't log in, but you can go to— well, some of you are sponsors, so you can log in. Uh, okay. I think we hit most of this. Let me go to the more recent ones.

I think we hit most of it.

Okay.

We'll do, we'll, it'll take us a little while. Um, we'll, um, you can come build your own at Saastr Annual. The folks will, will be from all the live coding platforms will be there to help you. And we'll do, once we get through Saastr Annual, we will collect all the data and we'll do a part 2 of this and we'll show you what worked, what didn't, all the analytics compared to humans. So, and you'll see where we kind of landed on all of this. Um, and, uh, so I think we can break, but again, I would challenge folks. I gotta say, I've done a lot of investments. I work with a lot of high-growth B2B companies.

I—

very few people do this well. Very few people do this well. And when I meet a leader, especially chief VP of customer success, that tells me everything is perfect, they're always gone in 6 months, 100% of the time. They don't do this well. I think there is in every enterprise, this— there is low-hanging fruit here. There is something broken in your onboarding, retention, follow-up, customer education process. Find the most broken one or two things. And even if it isn't broken, be honest. Go through your flow. See where you have gaps in coverage. See who only— see where only 50% of your customers are covered or 30% or 20% or where there's gaps and think about what you would build to get that to 100% coverage. And I think it's worth just building that. What would get you to 100% coverage in a segment? And I think you can copy Amelia's prompts and copy our work or just go into the Leaders Like Reput and say, just copy SaaS for sponsors and see what you get. Have some fun with it because the software in this category is mediocre and it's difficult to configure and it's dated. It's dated. So I think that this is a low-hanging fruit for Vaibhav. So thanks everybody. We'll see you in May and we'll see you at the next one of these AI Workshop Wednesdays too. Thank you, Amelia, for all this. It's great.

Yeah, thanks everybody. Just to say last little bit here, to Jason's point, I'll just, I haven't really promoted this, but the agenda is up on SaaStr Annual. If you guys want to see what we're doing, all the vibers will be there. Anton from Replit, Amjad from, sorry, I said that backwards. Anton from Lovable, Amjad from Replit, Gene from Vercel is going to be there. On Tuesday, um, not to toot my own horn, but one of the top performing sessions is I'm going to do a version of this with our AI VP of Marketing. Um, but I'm actually going to have my FTE from Replit join me so I can show you guys how to do one in real time. So we'll do a version of this, but a little different where I rebuild an AI VPM. I'm not going to use our data. I'm going to use probably fake data. I decided was for the best for this because we're also going to record it. But, um, so yeah, that'll be a fun one too. You can come ask me questions there too. If you're coming in person, but yeah, all the, all the vibers, everyone's gonna be, Anthropic's gonna be there this day. There's a lot of—

Yeah, it's good. Our goal will become, if you come on Tuesday to deploy day, whether you fully use it or not, you can walk out with your own AI VP of marketing.

That's my goal.

That's the goal. It's gonna be awesome.

Yeah. It's the last session of the day for that reason. So you guys can ask us a bunch of questions, but that's my goal. So you can walk out with one. I'll see you guys there. Thanks for tuning in.