Media/Maven/YouTube

How to build a Github repo for your revops and GTM system

Maven/Michael Slawson & Benyamin Holley/

Maven/Aug 20, 2026
Transcript

0:00:00 Hey. Hey. Is uh my mic volume good? Your mic volume is great. Can you share screen? Yeah. Okay, great. Yeah. Did you want me to share something? Are you just asking? No, no. I'm just like I just want to make sure because you have your PowerPoint and you're not a host, you know? Yeah. Yeah, we don't want anybody to start sharing their PowerPoint. [laughter] You know those pesky parasitical PowerPoint sharers. Yeah, imagine zoom bombing like a meeting and just sharing [laughter] a different PowerPoint. Okay. Should we just start letting people in? Um, yeah, I think so. Is uh Kelsey or whoever the other Maven person, are they joining this

0:00:45 eventually? Yes. Okay, we got some folks in here now. Oh yeah, some very handsome folks too. Wow. I'm making you a host. Can you only have one host? Okay. Wow, we got Nick. How's it going, Nick? I hope things are going well on your end. Hello. Hey, Nick. Got Harris. We love all of you. We just only know some of you. Yeah, but we're going to get to know each other way better by the end of this meeting. I think we'll do a trauma dumping session at the end of the meeting. I think in case you guys you want to get your chest we'll share the recording out with everyone afterwards.

0:01:32 Slides also. Are we doing the slides? Are we sharing the slides? Yeah, we'll share the slides. Um, we could probably just get started. I'll just keep hitting the admit all button as we go. Cool. Um, do you just want to intro us first before I pull up the deck? Sure. Uh, hey everyone. Uh, I am I think that's a good that's a good uh icebreaker question then for for you is and I you don't you don't actually have a zodiac sign, right? What do you mean I don't have a zodiac sign? Doesn't everyone? Everyone, but you but you don't. I don't think you do. I do. And I'm Sagittarius and I don't believe in zodiac stuff. And people tell me that that's a very Sagittarius thing. So, I don't know. [laughter]

0:02:26 Um, hey guys, I'm Michael. Uh, this is Ben. We also have joining Alex from our team. Uh, Ben and Alex are both pretty new to apparel. I started apparel a couple years ago. And today we're going to be talking about the GitHub repo for RevOps and for GTM. Um, as when Ben joined a couple weeks ago, we were talking about what are some things we can just kind of talk about and where we think we have particularly strong mental models and you know, Ben's been talking about the using GitHub as a home base for running his RevOps and GTM for a long time and has done it multiple times. So, we figured this would be a great first lesson that we could lesson.

0:03:17 That's a funny word to use. Basically, a a session we could run where we could talk uh talk about some of the ways we see this see the world. And we'll do more of these if you guys end up liking it. Um but yeah, Ben, hand it off to you to do an intro and uh and uh I think Ben's going to run the slides uh today. Uh so, yeah. Cool. Yeah. Um, well, thanks guys so much. Uh, if I could uh could I actually just get in the chat like an emoji or something if you guys have ever used GitHub or uh or anything like that for anything pretty much. I'm just trying to figure out like everybody's fluency here. Um, okay.

0:04:08 Uh, well, I asked for an emoji. Sergio's emoji is a thumbs down. So I'm assuming that he has not used GitHub. [laughter] Um, cool. Uh, yeah. The reason I ask is because um trying to figure out how how deep to go into some of this stuff. Uh, I think that GitHub a lot of people are like a little bit intimidated about it. Um, but there's really nothing to be afraid of. uh like we're going to get into here. I think your AI coding agents and things, you want to treat them kind of like a genius and really lean on them. Um which is what I do. People are like, "Oh, wow. You're so smart.

0:04:56 You're so technical. Like how did you learn all that stuff?" I'm like, "I didn't really learn anything. Like I just know how to massage Claude to like do what I want." Um it's actually interesting. The people who are more effective with AI actually have middle managerial skills. um because managing agents and uh claude is you know it's very common to assigning tasks you know scoping a deliverable and making sure that it's um gets done on time. Um I won't really go into my background too much but um I was an SDR for a long time then I was head of growth at a very small IT services AI company.

0:05:42 Um then I uh run at AOPS where we where I was a G uh my title was GTM engineer and now I'm over here at Aparero um heading up the GTM efforts and things like that. Um let me see going to go ahead and share my screen. How's the uh readability here? Can everybody see this? Readable. Awesome. Um Michael, did you want to say anything else before we sort of get started? Sure. H well um I no not not yet but I will be intervening momentarily. Sure. So why do we use GitHub right? Um I guess that question requires a lot of context. Uh hopefully that's not a bad pun. Um

0:06:28 it's a good pun. Okay. Okay. So basically the way that I think about like go to market and revenue operations is that certainly you have your systems like HubSpot and Salesforce and they're great and you can build a lot of automations in them. Um Salesforce I I can do a whole presentation on the Salesforce CLI but Salesforce does allow you to like use natural language to code in functions and things like that. HubSpot not so much. A lot of people are on HubSpot. Um I hope there's no like HubSpot loyalists here. But what I'll say is that it's very limited when it comes to like CLI and programming automations. And really when you think about like what is a CRM um it's just like a dataware three data warehouses in a trench coat.

0:07:23 And in theory you should be able to pretty much get a coding agent to program any sort of automations into it that you want to. But if you've ever tried to do that, you know that's not exactly the case. And so um basically my sort of like little hypothesis is right that and this isn't new, right?

0:07:47 We're basically like inventing SAS again from first principles except uh consider like all the stuff that we're doing like micro SAS right um and what I can do what you can do hopefully by the end of this presentation is I am rocking Libre Office yet I don't have a Microsoft Office on here Michael doesn't pay me enough to uh pay for um Excel and all that stuff I'm just kidding but um is imagine any solution that you want to build into your revenue operations or CRM or things like that. Uh basically tell an agent what you want to do and be able to host it and deliver an outcome. And so that's what I'm hoping to show you guys. Um so first slide, right?

0:08:38 Like I tried to I tried to think about like wow what even is GitHub and like how does it fit in really GitHub is just a it's a repo, right? That's what they call it. It's a repository. It's somewhere where you store code. Um, what you do with that is more of an interesting question because you can store code anywhere, right? You can store code on your computer, you can run a home server, you can do all these sorts of things.

0:09:12 But if you're new or even if you're pretty experienced with this, I think a good way to think about it is right like your middle managerial skills will come into handy if you can properly scope a problem and solution and know exactly what you want to get out of this process um before you do anything, right? Um, and then the way that I'm doing this is I have uh Claude Code and cursor. I'm a big cursor guy uh just for reasons that we're not going to get into here. Um, and then that sends the code to GitHub. Uh, Cursor just launched something called Origin, which is basically like a GitHub competitor.

0:09:59 We could probably do a whole talk about that, which we're not going to. Um, and then, uh, GitHub's going to deploy those services, the code I should say, to Versel or Railway or wherever you're hosting um, your application, right? Um, yes, Harris. Yes. Um, can can I ask you a question, Ben? What? So, like GitHub in this slide, which I agree with, is it's in the middle of the thing. It's so like you know you are coding not in you're coding in a in your terminal

0:10:35 you're hosting the actual work that you're doing somewhere else like why is GitHub even a necessary layer here and like what do you think it is about it that makes people be so obsessed with it when it comes to GTM and like the stuff that gets built in GTM. Yeah. So, uh that is an interesting question because GitHub may not be necessary in the future. So, um if you came to this presentation, you're like, I'm going to learn how to do GitHub. I'm not sitting here telling you that you need to go uh learn something else now. But, um GitHub has version control. It's got collaboration. It's got security.

0:11:20 It's got all these other things. There is actually like a social element to GitHub as well. um which maybe many of you are familiar with. But um one thing that's interesting is that there's a lot of open source like GTM repos like oh for instance um we're building our new website and I wanted to have SEO AEO best practices in there and there's literally like free GitHub repos that help you like refactor your site uh to make sure that it adheres to those principles. So there's a lot of like open- source code out there that's stored on the GitHub network. I don't know how to phrase that. Um, yeah, that's a good Vicos.

0:12:08 That's a good explanation. That's why people generally use like GitHub is for version control. It It's interesting because like the way that like I do things as a solo sort of person is like I don't worry about version control. I don't read code. Like I don't do any of the things that GitHub is probably actually good for. And um I think it'll be very interesting to see like where coding repositories and coding storage kind of goes in the next year or two. Um so I don't know if that answers your question, Michael. It's not probably not a good answer, but yeah.

0:12:50 Well, that's a good answer. I think I think also it um well it will you're you do a lot you've done a lot of your stuff solo, but I do think a future is coming. There's probably a few businesses now, but there will be more who have built a brain and uh have built a brain and have multiple engineers working in and around the brain and and there does need to be version control because you know we're we're like updating code that is driving outbound and driving account scoring and driving you know all kind of the core elements of of a classical GTM motion and you're going to need to have a way to track changes uh as multiple people are interacting with the same brain.

0:13:42 Yeah, for sure. I know the um Octave folks have talked about this a lot and it is interesting to think about like version control for your ICP or for messaging or things like that. Um, and there's something we're not going to talk about too much in this presentation, but um, I do have like self-healing, self-learning stuff and Claude Code that um, learns from messaging that worked.

0:14:10 And it'd be very interesting to see like an application. I don't have like a wellbuilt out process for this, but it would be interesting to see like um how how those things evolve over time and like how do response rates and like conversion rates change over time with like different messaging and if you stored that in GitHub like instead of just in your brain, right, which is how all sales teams operate, they're like, "Oh, well like six months ago this messaging worked or that messaging worked and it's like all vibes based and totally untrustworthy, too." Um, so there's an interesting way to think about like how could you optimize those sorts of things through like version control. I don't know. It's a very off label use.

0:15:02 Yeah. Should we next slide it? Um, uh, yeah, real quick. Just like the live SAS piece here. We'll get into like what what that is, but um, one thing that I think a lot of people are maybe stuck on or maybe get maybe get stuck on is like, okay, well, cool. I've like coded a dashboard or something and it's like that's great. Uh but you know that you can actually code like a full cycle SAS application that's hooked into like any API or system that you want too that can actually like go do stuff. So we'll get into that.

0:15:44 Um yeah, this is kind of covered by the first slide, but and I just did want to mention we could do a whole class on this. uh we're not going to this time, but there's a lot of upstream stuff about training your coding agents that is worth looking into. And um yeah, so look into that. Um I think I covered this too a little bit, but like I will just say that I've talked to a lot of people about AI coding and things like that and they're like, "Oh, do I need to like know Python?" I'm like, "Listen, if you want to go learn Python, have at it." But really like we the peeons of uh the engineering world right on this call that's a layer of abstraction that's too granular to really worry about at this point.

0:16:41 Coding is kind of a solved problem. Um I don't want to get you know there's the cliche about like taste is the mode or whatever. There's there's many moes taste I'm sure is like one of them. um the jobs to be done in revenue and GTM engineering are way more important in my opinion to know than like how to structure a data warehouse right because with very few exceptions the kind of work that we're doing and I'll show you the work that I've been doing it doesn't require data warehouse expertise it's like what what outcome are we trying to achieve here like one thing that if I could just leave you guys with one thing is like go to market engineering revenue operations these are really like product problems like we're really building like products for the internal sales team.

0:17:39 And when you think about things like that, version control, GitHub, accessibility, engineering, accessibility for the rest of the team um to collaborate on things like I just invited Alex to a repo today, right? Because he's going to start working on it. He doesn't need my permission to work on it.

0:18:00 Um he needs full autonomy to like go do stuff like within the framework and this uh the system that we're building, right? Like a revenue organization. Like this is something I'd really tried to hammer home at AOPS. and try to get everybody into the growth team GitHub so that we could collectively start solving problems and like building out like true tribal context and it does take like a level of sophistication there from the team that I think a lot of companies maybe maybe they don't have that level of sophistication or maybe they're afraid to let people run uh with it you know and and my philosophy like dude I was like giving SDRs like access to the growth team GitHub and giving them all the API keys and it was actually working too.

0:18:55 I know some of you guys are like, "Oh, I don't don't want to give my SDRs API access." But like, if they're good, if they're smart, I would just do it. Um, so anyway, then, have you ever found exceptions to the rule where you actually did need to loop in like a software engineer on one of your teams to like talk through a problem as you were building out uh building out your your stack? Yes. Uh, yes, I did. And I did not get that support from from this one project. I'll just very short story. So we had this lead magnet that I built at AirOps. It was very successful.

0:19:41 It was an AI search readiness scorecard. It would go run a bunch of prompts about a person's website. Um it would analyze the schema, all that kind of stuff.

0:19:53 If anybody knows anything about AO, I'm not going to like explain it, but um uh and I don't know like where they got the idea that like I was like a real software programmer, but like the CEO just like handed me like this half done like repo and he's like, "Hey, can you go make this work?" And I was like, "Yes." I mean, that's what you say, right? You just say yes. Um and in that instance, it would have actually been better for me to do all the MVP work and for a real engineer to like figure out all the kinks. um you know for obvious reasons I think that was like a project where there was a level of sophistication where we needed real engineering resources to support it past like uh a v0.9.

0:20:49 Um so I just think it depends on the level of complexity. Um yeah. Yeah, that makes sense. I mean I that's that's what I would expect like you know you can you can get a long ways but eventually you need to loop in people who actually do this stuff for a living right

0:21:11 yeah I think like um there was uh so these are probably like software engineering 101 problems that I ran into at that time but there was like concurrency rate limiting uh so that just means like how many people are trying to pull a scorecard at once I had to do things like prioritize external polls so that like prospects could do it um instead and like if an SDR was requesting a scorecard to do outbound like their uh scorecard would be dep prioritized in the queue and like that stuff's probably really obvious to somebody who's ever built an application but it's just like I didn't know about any of that stuff and then people were like why is my scorecard taking 36 hours to run and I was like well uh I don't know let me

0:22:07 like go figure it out and um I ran into so many problems downstream of it like but it was a great exercise and I learned I learned so much from doing that. So, if if your CEO in a u sends you a Slack message at 2 am uh that sounds like he's having a schizophrenic breakdown and sends you an unfinished GitHub repo, I highly highly encourage you to just say yes and see what happens. Um because I think I became a way better uh engineer from that. Um uh Michael, do we want to take questions now or Mike? We'll take I think we'll run we'll run through we'll we'll have it probably at least 15 minutes for questions towards the end here. So uh

0:22:59 yeah for sure. Um these slides there's only what like four more five more slides. We'll get but yeah let's talk through some specific examples. Cool. So uh network mining is something that I've worked on quite a bit. I worked on it quite a bit at AerOps to a pretty good degree of success. Um, hey, and you can actually just always drop question in the chat too that we'll go back through like towards the end.

0:23:30 Um, so LinkedIn does not want you to actually use their platform. I'm very convinced that LinkedIn does not want you to use their platform, right? Uh so this uh problem that I'm working on at APA right now basically is like hey you know me Alex and Michael all have networks that we can leverage you know how do we actually leverage them right um at Aerops uh I did something a little bit different I guess I'll talk about that basically at Aerops was a much more mature business we had you know CRM with like 200,000 people in it I mean thousands and thousands of companies um and when I rebuilt the TAM one of the things that I I wanted to do was see like where do we have warm opportunity ities for introductions.

0:24:25 And I actually wanted that to be automated. I never actually got to be truly automated, but my uh plan was um if an account executive gets assigned an account into their book of business that we have this app that automatically searches up this repository of LinkedIn data that's been downloaded from people to see who in their who in our network, excuse me, is connected to their book of business. And I re what I wanted to do was actually just send them a Slack message and be like, "Hey, this we enrolled this person in a limb list campaign for you uh with this copy. Uh do you want to approve this?" Um and do you want to put them in like another different sequence or something like that, right? Like I wanted to be totally automated end to end.

0:25:22 Never quite got there, but that is very easy. Um maybe it sounds complicated or maybe it doesn't sound complicated but um those are the sorts of things that like you can start doing when you start understanding like how to use GitHub, how to use APIs, things like that, right? So for our purposes, you can sort of see the uh the the chart here. Um we get our networks, we ddup the people. Um I I actually have something interesting that I'm doing. I don't know if this may be just like too whimsical and unnecessary, but so like I have um [snorts] it's basically just like a little SAS app and it's got a list of all the people.

0:26:11 It's got a queue of people it thinks are our ICP and I either approve or disapprove them and it's training a a model on the back end. It's really just a prompt. It's really just training a fancy prompt. But like the more people that I approve or disapprove, it's using all their different fmrographic data from like Lead Magic. Um so it's reading their like description. and it's enriching their company information. It's like figuring out like what um I'll I'll post it in the cafe.

0:26:46 I'll post a screenshot in the cafe uh later. Um uh and it's going to hopefully my my plan is that this information and this training that it's doing is going to like help us apply this prompt uh to future endeavors where I'm like, "Hey, here's a list. How many of these people are actually our ICP?" you know, and it's got very detailed information like I disapproved an executive assistant, which I don't know, Michael, is it should we be uh emailing executive assistants? I kind of feel like they might know a lot. I uh I'm glad that you disapprove of so many people.

0:27:29 Okay. [laughter] Yeah, it uh it um scrapes their political opinions. [laughter] Well, we we ex um I I actually just want to really quickly interject actually and say like to me this is uh this is one of those things that's like from from first principles if I'm like sitting down with a new client and like they're like hey I guess with GTM you can do like email and cold call and LinkedIn or really if you're if you're really if you're like jumping into a new GTM role in any business like I think that leadership usually like has this idea of like the channels that you're going to push your your stuff through.

0:28:16 But to like sit down and say like the first most obvious thing we're going to do is like mine the networks of the people who work here for like warm intros to people who are an ICP fit. It's like the most obvious thing that to me that that you should do uh that you should do this. Like this is this is like a play that makes total sense to me and and like get you like a lot of quick kind of uh lowhanging fruit, quick win energy, too. Uh because people are likely to reply when you're a firstderee connection to them.

0:28:58 Yeah, absolutely. Um it's the most obvious, lowest hanging fruit. And it's like when I used to run my outbound agency, this was a long time ago, but um people would be like, "Oh, we got to send cold emails. We got to send cold emails." I'm like, "Okay, I will send cold emails for you, but like do you think that there's like some other people we could talk to first?" Um yeah, I'm not sure what you mean, Alberto. Like, uh I have a bunch of different folder structures I could share if you want to be more specific. Um, but you know what's uh interesting is actually inspired if anybody uses Midjourney um they help you personalize your profile by selecting a bunch of pictures like hey do you like this picture?

0:29:52 Do you like this picture? It gets like a sense for your style. And I was like oh we should just do that for your ICP. We should just have like um like uh you know you should just swipe left or swipe right for like hey this is your ICP. This isn't your ICP. And eventually it'll start start like learning what it is without you having to really like explain it because it's got all the fmrographic data about these people. So, um, anyway, uh, and then at the end of the app, basically, like once all the people are enriched, you can just go ahead and put them into a limbless sequence. I have like a different I should show you the app. I just, um, don't have it up right now.

0:30:46 Um, so anyway, that's like a thing that I've worked on. Do you want any questions about that before we move on? No, but I think one of the questions that's accumulated that we should show at the end is like we have to give the people what they want and they just want to see your folder and subfolder structures. Um, dude, I don't think it's going to be that impressive. Um, [laughter] well, I just I have a code folder on my desktop and like there's nothing to really like the the folder structure. Okay, there is stuff to talk about the folder structure because I have like put work into that. Um, let me just share my other screen for like two seconds.

0:31:37 I really don't know what this is going to show you guys, but like hope hopefully that there's not um a terrorist manifesto in here anywhere, but um yeah, the I mean this is this is all that's in the the folder. I don't even want to click on this stuff. I don't know what data is in here. Um so yeah, like imports, investor sheets, logs, stuff like that. I mean this is like honestly this is one of those things where like I don't really read it. Uh here's just all the HTML files for the app. Um, okay. Maybe maybe let's maybe let's put it this way. Like, yeah,

0:32:22 I would say it No, keep keep that up. Like that's actually good to keep up. Um,

0:32:28 I think that there's this idea that like the folder structure really matters when it comes to like that's the whole like there's this kind of uh impression in our in the mind that the point is like to have this beautifully organized folder structure. And I guess I guess it's like this is like a square peg question like trying to reach round hole mind of Ben like are you kind of saying like h you know the folder structure is like it's just it's just the structure like you know and like it's actually not as important as you'd think or or do you or would you say no it is actually important I just don't like I don't really know how to show it in a way that like can be be uh uh like instructional.

0:33:24 Okay. So I'll put it like this. for for applications that we're creating. Close that out. Um, okay. Um, for Claude Code, it matters what the folder structure is. For these apps, I don't think it matters. I That's why I want to make a delineation between like for training agents, the folder structure does matter. uh for for applications I don't I think it's a layer of abstraction that's too immaterial for us to worry about um for for agent training the the folder structure matters because there's like certain context certain API documentation certain training material I guess you'd call that you want referenced at a particular time and I'll I'll explain this in like I'll try to explain this in like two seconds I have what I call a referential memory system in Claude Code.

0:34:19 Uh all that means is that when certain subjects are mentioned, it's going to go pull uh certain information. And I I equate it to like when you're cooking an omelette, you don't want to be thinking about French history because then you're going to mess up your omelette. Um and so like I kind of like treat agents the same way. I'm like, "Hey, when we're when we're doing cold outbound email, I want you to reference the cold outbound email uh folder, when we're thinking about design, I want you to reference the design folder." I don't know, that sounds obvious, but there's a way to set it up that makes it a little bit more useful. Um, let me get back to the presentation. Um, so anyway, let's move on from the the network minor.

0:35:14 Um, unless anybody else has any questions about that, they can throw them in the chat. Um, so I have this idea that like your CRM should really kind of be silent. Like you really shouldn't have to log into the CRM. And me and Alex and Michael have been working on this thing called Mother Mother 6000 named after the computer from the Alien franchise. Um, you should all go watch Alien. Um, but basically I mean it's it's kind of simple. Anybody can go set up a Slack agent and you can ho the where it becomes interesting is you host the code for the agent on GitHub, right? And from there you can put plug in various APIs. So basically it's sort of like this idea.

0:36:06 I do run Windows. Um Macs are just expensive. Michael doesn't pay me enough for me to afford a Mac. Uh Michael has a has a Windows machine also. Um so basically like you have a GitHub folder with uh all these different connections to all these different different applications. Uh, right now mother's hooked up to the CRM hooked up to some other stuff that I can't even remember. Um, I just got email bison and scaled mail set up. So, I'm going to have mother hooked up to those things as well. And so, like the simple thing that we're trying to accomplish is we want to be like, "Hey, mother, what deals moved this week?" Uh, hey mother, how many email replies did we get this week?

0:36:58 Uh, hey mother, what campaigns worked uh and didn't work? Um, let me see. John, let me actually let me let me actually answer that question. Um, so Versell reads the code off of GitHub and launches a SAS app. So yes, Versell deploys the app to a website uh that's not indexed or anything. I would just as a note I would highly recommend that if you're going to deploy stuff on Versel, make sure that it's either password protected or buy the updated upgraded Versel subscription that allows you SSO so that only people from your organization can log in because uh these websites are not indexed but that doesn't really it's like the thing where Claude exposed all of those sharable files the other uh sharable conversations.

0:37:50 I don't know if anybody saw that. That's what I'm trying to avoid. Um okay. So, I actually used to work at a company called Syft, which sold into MSP and cyber security companies. And now I've become so paranoid about like cyber security stuff, which is actually a good thing. Um, but could be a different discussion.

0:38:13 Go look up some cyber security basics because anyway, uh, mother is not a Hermes agent, but basically that's what it is, right? And I'm actually looking into like seeing if Grockbot or something else could help us like agentify this stuff more because like what I want to do is be like I want to enable the team to be like hey we just onboarded this new client can you go ahead and start an onboarding flow for uh let's say they want to run cold email or let's say hey mother can you start an onboarding flow to mine all their calls and like start developing an ICP documentation or things like that right um the sky's is the limit on that kind of stuff [snorts] Um, Harris, that's a good question.

0:39:07 I'm gonna answer that later, though. Um, I think we could talk more about Mother, honestly. But it's really the the simple way to put it is that like you can just create an app that on Versile has all of your API keys for whatever tools you're using and you can go make conversational commands in Slack. I think from like a more 10,000 foot view type of thing, what we really want to do is like make AI and agents multiplayer. Um, so that the whole team is sort of like collaborating on things.

0:39:45 I I haven't quite figured out how we're going to do that. That's my pie in the sky sort of like reality. I think there's somebody who's not muted. I don't know if you can mute. U Michael, do you want to Is there anything else you think we should cover there before we move on? Um, I think that if we don't show it today, we should show it next time. Like what is like what uh something Oh, who is that person that's talking? They're very They're very hard water. Yeah, that's not I found them. I found them and I I silenced them. We shall only hear from Ben today. Um like can we can we show like the raw uh series of prompts or or or of of how this is actually working right now or like one of the flows of of mother.

0:40:44 Um you can tell that we didn't prepare uh pre pre-prompted questions. I'm just putting Ben on the spot. Yeah, there's nothing like there's not a lot to show there right now. We've not really used her too much internally. Um what Okay, what I'll tell you is that um we had this same setup at AirOps where basically the rev our RevOps director could basically like just probably check out for the day and like have his agents like handle everything. And basically like all these dumb little things and if you're an SDR, I apologize in advance, but all these like silly little asks people would do when you're a Rav Ops person like, "Hey, can you close this deal?

0:41:33 Can you uh make sure that um this duplicate deal that I created because I'm not fastidious about my work, can you make sure that that gets deleted?" Like you just tag um Arthur was the name of the bot um at Aerops and it would just go do the stuff. It would just uh it would build um HubSpot campaign lists for you. It would do things like that. anything that you could do with the HubSpot API, it could just go do because it had all the documentation, it had all the endpoints, it had all the permissions to do that stuff.

0:42:15 And what happens over time is like you get more and more asks for like various things. It saves all this stuff as context and you start to understand from a much more like programmatic view like what does your revenue operations business actually need to do? Like where are the areas of weakness in the business? Um, and what should we be trying to automate more thoroughly? I was going to show you guys just like the stuff I was doing in cursor, but there's like a bunch of API keys in here that I don't want to expose. Um, cursor for GTM would be a good class. Ben Ben or Mi Mitch had a question which is, have you considered exposing mother to Claude through Claude tag and dropping that into Slack channels?

0:43:09 Yes, great question. Um, I actually have on my road map for mother to have her hooked up to cursor. This may be silly, but it's just my thought. Um, we should be prompting like our internal GitHub repo from Slack and we can just be sitting there conversating about it. And I think you can already do this with the cursor tag. Um, but I'm going to see if cursor like has an API or something that we can plug in where we're like, "Hey, mother. Um, hey mother, uh, can you go add this capability to yourself?" [laughter] Um, so that you can go like do whatever. Like somebody might be like, "Hey, can uh like let's say Alex is like,"Hey Ben, I really hated that email campaign that you put out. Can you edit the copy on that?" And then I would just tag mother.

0:44:08 And be like, "Hey, can you just um go edit the copy on this based on Alex's feedback so that it's all like in Slack?" Um some people don't like Slack. They're like, "It's a way, you know, it's a time waster." But I actually think that using Slack as a service for like GTM work and revenue operations is um I don't know. I like it. [snorts] Uh, let me get back to my screen here because we only have a couple slides left and I think uh Whoops. We'll get through these last ones pretty quick. Um, so there's this one workflow that I did at AirOps that generated us a lot of pipeline.

0:44:54 I was pretty proud of it. Um, short story because uh, we only have a couple minutes left and I want to get through all the questions that we can. But so one time I got into a internet beef with the CTO of user gems because I posted something on LinkedIn. I was like, anybody can just go vibe code user gems. Like why do why does this software even exist? And he was like, well, you know, you can't just vibe code user gems blah blah blah blah blah blah. And uh, what I will say was he was correct. You cannot just go vibe code user gems. I did vibe code their minimum viable product.

0:45:42 I know they do a bunch of other stuff now. Um, but former customer campaigns and job changer campaigns was like just like so much more complicated than I thought it would ever be. Um, and there's probably like a midw version of this that I just uh didn't do. But anyway, um, so for former customers, this one we might need to zoom in on. For former customers, we basically had like a bot that would post something in a Slack channel be like, "Hey SDR, this customer is in your patch. they change jobs, do you want to add them to this limbless sequence? Yes or no?

0:46:25 Um, very simple, right? Uh, the problem was like ownership and all this other like minutia that again that you don't consider when you're starting it, right? You're like, "Oh, this will be easy." And, uh, then it turns out not to be easy, right?

0:46:43 Um, the cool thing that I think I did that was interesting that actually booked us some huge meetings, um, was if they were a former customer, and to be fair, former customers, you could probably throw a potato at them and they'd book, but, um, I think adding a little bit of something to show that you care, uh, is important, right? Um it would go read their metabase usage uh like what workflows did they run when they were customers and be like hey um you know when you were at so- and so company you guys did this competitive analysis a lot um we've actually launched some new features with our new agent captain Quill I'm out here shilling for air ops still they they programmed drink yeah remember where you work Ben

0:47:38 yeah yeah um you know uh are you guys doing any of the same stuff over at your new company. Uh because it looks like you ran this workflow like 50,000 times. Um don't know if you guys just like want to chat and so that was the extent of like the personalization, but it did make a difference. I think I think it makes a difference when you make people feel seen and heard and all this other stuff, right? Um so you can see the uh this is a really simplified version of the chart. Let me see if there's a more complicated version here. Um, so basically it would we would have a web hook from common room. There's probably a bunch of other sources that you could use for this. Uh, common room worked fine.

0:48:33 Um, we had our ICP gate, right? Uh, is it an existing deal? Um, making sure that we're not This is what this is what became complicated, right? Because like there's all these things that you have to account for. If the new account wasn't in HubSpot, I had to send it to a clay table to get scored. If it didn't get scored, it wouldn't get sent back into the workflow. if it did get scored, we had to make sure that it was sent to the right SDR in the right segment.

0:49:11 Uh things like this and um you know, then it would divert between a former customer and a former prospect campaign. Um and the reason I'm showing you guys this uh and maybe more practical advice, um is like it is easy to canvod.com. Yeah, I haven't seen that, but I know what it I know what it is. Um I'll check that out later. is you I just talk through this stuff with cursor. Uh I just put whisper flow on and I'm like, "Hey, this is what I this is what I want to achieve." And we just work on it together until until it's done. Um it it really kind of is that simple.

0:49:58 Um so anyway, those are like the three workflows that three practical application things that I would want to talk about. And um I think that kind of covers it. So the pre presentation's kind of done. If we want to start going through questions, comments, complaints. Yeah, lots of complaints. Okay.

0:50:18 Um, uh, yeah, guys, if you want to just do like the hand raise thing or you can, uh, try to just unmute and cause chaos as well. Uh, that either way works, but we'd love to if you have any specific questions, uh, we'd love to chat through and we can stick around. I think we set this for 45 minutes, but we have some time afterwards if anyone wants to just talk through any specific problems. Um, and then yeah, Ben added some nice QR codes because we know everyone's on their phones during this sort of a thing anyway, right? So, uh, whip them out and, uh, feel free to add us and if if you have anything you want to talk about offline, we're happy to do that as well.

0:51:12 I'm looking at some of the questions. So a wall uh a wall to the answer to answer your question about GTM engineering should focus more on critical thinking and judgment. Absolutely. Uh I think if you're really good at sales and marketing and revops um adding coding to your skill set is great but it's hard to learn and I have a very short attention span. So um yeah, Ben's got his hand up. That's what's going on.

0:51:44 So yeah, what's up? Uh just a second. Yes, thank you. Thank you for presentation. It was actually interesting to find couple of insights. But there is a question where actually you you find a problem to execute with agents some part of the job or some some part of the funnel etc etc. It would be interesting where is actually um you spend more time or it nearly impossible to create the app automation or I don't know anything that could be related to that.

0:52:18 So are you Yeah, go ahead. So, so you know, in in the spirit of trying to say something more interesting than than it depends, one thing I can tell you is that like a lot of the campaign, a lot of the efforts where we have success with sending emails, we actually have just like templates set up for emails that we think convert and and there may be a little bit of uh filling in filling in with some, you know, specific snippet of information, but I think like trying to And like maybe this is like commonly accepted at this point, but like trying to get an uh an agent to like write a whole personalized email um I think is often uh is often actually less effective than just writing a very good email and and

0:53:14 cycling through different different templates and maybe doing a little bit of spin tax on it uh to make sure that you know you're you're you're increasing the variability of the copy. that um you could spend a lot of time trying to build an agent to write the copy or you could just like write good copy, you know.

0:53:38 Yeah, I definitely agree with that. like um I think that that my philosophy on this is that as I've become more and more AI pill you know over the past 24 months I really think that like AI is good at science engineering mathematics coding and I really think that there's a diminish point of diminishing returns on it actually trying to do like good creative work um because in some ways it is [sighs] it is uh AI I takes on the properties of the people who designed it. Uh their their racial, their political, their intellectual, their philosophical interests is modeled after the people who created AI, which is a very small group of people in California with a very particular set of properties.

0:54:29 And for one reason or another, I just don't think it's good at creative work like that. Um I I don't think we should be anyway. I won't go down that rabbit hole, but like yeah, I don't know if that answers your question at all. Um, let me keep going down these questions. Anybody feel free to Oh, cool. Yeah, thank you. Thank you very much. Yeah, no problem, man. Um, anybody feel free to unmute while I'm just going down through reading these. Uh, for Sana, why is the network miner on Versal? Uh, why not run this as agents for each founder? I'm not sure I understand the question. Uh Mark, yeah, how's it going, man? Go ahead.

0:55:17 Hey guys, thanks for this as well. Um I thought the LinkedIn network graph was really interesting and something that I've been trying to figure out was doing a similar thing but for our current customers. So obviously it's kind of easy to get your employees to export their LinkedIn connections and get their network that way. But it's kind of harder to get customers to do that for you. And what we were looking to do was see who's our uh current customer like LinkedIn network and match that against our like target account network and see if there's any introductions that way. I wonder if you've looked into that at all.

0:56:02 Um are you just asking like can you get people to download their data like for you? Well, it's kind of hard to go to a current customer and be like, "Hey, can can you like export your, you know, your LinkedIn data so that we can match it against like our prospects?" Do do you mind do you mind if I ask u how what do you do and like a general range of count of uh of what uh of how many customers there would be to to to get this data from.

0:56:40 So I sell into health systems in um so in healthcare. So we have like six uh health systems and a lot of these people know people that work at other health systems, right? So they've, you know, met them over their years of working in healthcare. Um and it's kind of hard for us to know that we currently work with somebody and they know somebody at one of our target accounts. And I was looking at tools like there was one called the swarm, but um it didn't map like first connections. It only mapped based on like employment history and stuff. And I just wondered have any kind of thoughts or

0:57:26 it's it's a I think it's a I'm sure Ben has some thoughts. Honestly, I think uh it it's very likely that you won't be able to convince them, but uh I would ask them and I would maybe like offer them some sort of incentive to to do that and I would maybe try to package it as uh I I would try to package it package it up as like a uh so so so that it feels quite professional. professional and you know maybe like I don't know I'd probably think about because you're are you who is the customer is it like a chief something officer in the health system

0:58:12 yeah directors yeah stuff like that yeah interesting I don't know have you tried asking any of your customers [snorts]

0:58:20 um the CEO's been adverse to doing that they they would like a way to just you know um without having to ask them to go into their LinkedIn or anything. Just figure out a way to cross reference their connection list with a target accounts, you know, employees. I mean, the thing is, health professionals aren't typically that active on LinkedIn, but you could you could uh you could create a a list of Yeah, you could create a list of like u all the customers who you want to track and then just look at wherever they've commented or engaged with other people's posts or companies and then use that as the basis of a graph of who who they may know and then ask them for warm intros that way.

0:59:13 that you know in healthcare spec particularly LinkedIn is like a secondary or tertiary thing for for a lot of them. Um so it might be hard to get like a a high a high collect rate doing that. I don't know Ben you have other thoughts there. Um yeah this is a very difficult problem to solve. I would recommend the swarm actually. Um, and I would I would second Michael being like, "Hey, listen. Here's a contract for a referral partnership agreement. Part of the contract is you download your LinkedIn connections and we map it against our TAM and anybody who we close based on this agreement, you know, you would get 15% of or whatever. Um, I really don't think that there's a that I know of an easy way to do that." Um, I want to keep going down this list of questions.

1:00:12 Uh, Harris, you said, "Would you put demand gem revops knowledge?" Sorry, Mark. Sorry. Did you have something else to say? I'm sorry. No, just saying thanks. Yeah, that was actually a good idea with the engagement doing it that way. So, yeah, for what it's worth, I've tried to do the same thing and it is difficult, but a couple of people did say yes. Um, so you just never know. Um, would I put demand and revops knowledge in the same repo? Um, I really think that's a very contextual question. Um, I really don't know how to answer that. I I haven't gotten that. My thinking hasn't evolved that far yet. So, I'm just going to say I don't know. Um, Sana, I know you had a question that you were trying to ask earlier.

1:01:07 Yeah, I I I'm a little So, I'm just wondering the whole network miner thing. Why would I Why not just hand because it's three founders, why not just hand them the repo rather than like deploy an app on Verso? Um, just for whimsy and frolicking. Just for fun. Yes, we're fine. Um, and you get non people. Well, I have 10,000 connections. Michael has 6,000. I don't know how many Alex has. Um, and there's our time is very limited. Um, so we got to go by No, I'm just kidding. Um, our time is very limited.

1:01:47 So like we only want to reach out to people who are like have a high propensity to be in our our network. And I think part of the exercise was like kind of I mentioned like nailing down a prompt or some sort of sort of training model on like what is our ICP so that I can then apply that prompt to future campaigns and things like that. So there's like a bigger picture version of it.

1:02:18 Also, if any of you guys are interested in this, part of what we want to do at Apero is become like a software factory, right? Um and spin out these like sort of mini SAS apps that we can roll out to our clients. So, if we have a founder who's doing founder le sales and they come to us and they're like, "Hey, we want to do cold email." We're like, "Well, hey, have you like actually mined your network uh yet?" And they're going to say, "Well, no, not really, probably." We're like, "Okay, download your connections, upload them to this app.

1:03:00 We already deployed it for you and um it'll automatically enroll it in the limb list campaign for um in the Limless account that we set up for you." So, potentially it's something that we can like roll out to clients. Um so, I hope that answers your question. that does because I I just did something similar but using locker as an orchestrator and it does not directly um send the message the final variant to hey reach. Um so I think that's what happens with play and flocker and all the other guys. So that's why I guess it makes sense to keep this this particular type of thing outside of your orchestrator entirely in your agents or with a versal app I guess.

1:03:51 Yeah. I mean, it's it's it's semianual, right? But that's kind of the whole point is like we have the ability. We're very technically sophisticated. We can spam 10 million people, but like at this stage that we're at right now, we want it to be like um semianual. Um uh before we take any more questions, Michael, did you have anything you wanted to add to anything or or answer any questions that you saw? Um well, I mean, I just wanted to double on that, SA. Like I literally run a GTM advisory business and even I when Ben made the offer like can I just mine this for you? You know, just download and I'll I was like great because even I'm kind of busy, you know.

1:04:43 So like I think just having it done for you is is um is a is helpful. Uh gen generally beyond that I would just say uh we can keep like we can stick around for a while but we would love to chat with any of you about anything that you that that you're thinking about GTM land and uh and I hope you can find us where where we are which is on LinkedIn and and in the GTM cafe as well. Um and if there are any topics like I think a few kind of came out of today. I know it's a big ask and I mean this sincerely to be like, "Hey, that was an okay class. Like, here are some other things that would have made this more interesting." Like, we would love that feedback though.

1:05:40 Like, we will we will honor your name and we will make a Slack emoji in our internal Slack celebrating you if you send us some ideas of of uh of topics that you would like us to discuss uh in in these open sessions. And and this was our first one ever. So, we'll even, you know, we'll we'll be experimenting with the amount of time of Q&A and just doing office hours and, uh, we want to be very accessible here. Uh, do we have time to take David's question? And he's had his hand up for a bit. Yeah, of course. Cool.

1:06:22 Thanks. Ben, I think my question is more around, you know, we're in a in an era where just because we can doesn't mean we should. There's also a source of um tools and it's pretty to overengineer things. So maybe like in your experience what are some of the it's two ways of framing the question. what are the principles that that as GTM maybe the more technical side of GTM we should follow to determine when GitHub and this sort of infrastructure is necessary and then more uh just like anecdotically when was the point where you naturally like kind of hit a a wall or a bottleneck where you said okay uh having this isolating my you know local system or my cloud setup or whatever is not enough I need this GitHub repo of course I

1:07:18 know people have been mentioning and iterating on uh importance of version control and stuff but um yeah from a more strategic standpoint what what's that what are those common bottlenecks or like yeah transition points where you say okay this is where go to go to market starts needing GitHub um I think uh yeah I think from now I obviously work at a much smaller company but when I was at AOPS the the purpose was that like when I leave that these things will stay on and so like what I built out lives me, right? Um if I die in a plane crash was the joke I was always making with my boss, right? We were making that joke so much, but um maybe I was having a premonition.

1:08:11 Um but basically like you know two things like there's security concerns um IP concerns and for when somebody smarter than me joined like my goal at Aerops was to was to hire another GTM engineer that was more like data and coding focused so that they could like sort of refine all of the slop apps that I'd made. Um, so it I don't know. It was more like I just wanted to build something more longasting and when you're doing it all on like your local Claude Code and agents on your PC and stuff like that. Um, it just makes it harder to to build on top of things. So I don't know if that answers the question. U maybe there was something else there.

1:09:02 Ah, completely. I think it does. And and regarding the the IP side, I I think I've seen you promote a lot GitHub as a from a like a portfolio standpoint. I don't know if you want to add anything on that. If you've seen what's a like the go to market space, what's what's it moving towards? What are you seeing? Are you seeing people talking about the GitHub repos actually taking decisions based on that? What's a maybe some anecdotes on that side? Oh man, I got to be careful about how I answer this question. Um, I think, uh, if you're a GTM engineer or RevOps pro, uh, and you have projects that you've worked on that you're allowed to share with people, you should open source them. Uh, a because PE, and if you don't know what that means, you can look it up.

1:10:01 But, uh, a because people are going to fork it and build upon it and then you can take that back and be like, hey, somebody built something really cool with this. It's actually like way better than my version. For my part, what I'm gonna start doing maybe with certain things, I'll have to chat with Michael about this, but like uh maybe start open sourcing some of these tools that we're building um as like a lead genen source so that people can say, "Oh, wow." Like um these guys actually know their One of my engineers looked at your slop vibecoded app and said that for slop it was pretty good and solve a business problem.

1:10:49 Um so I think yeah, I saw a post on Twitter the other day. It was funny. is like I don't open I don't uh all my GitHub repos are private because I make stuff that makes money and I was like yeah that's probably true. Um so I don't know what we're going to do with that. Um my my portfolio like on my website if you go to benignamine.ai there's a portfolio section but um that's all like mermaid charts and like descriptions that doesn't have any of the code on it because a lot of it's stuff that's like proprietary to Aerops that I probably shouldn't be putting on the internet. Um so that's my answer to that question. my very careful answer to that question.

1:11:40 Cool, man. Cool, man. I appreciate it. Uh yeah, that's it on my on my end, but thanks guys. Cool. Uh anybody else any other questions that we didn't get to that were in the chat that we totally ignored? Go ahead. Go for it, man. Okay. So, um I dropped a question on the chat. um was essentially asking um if you had used uh trigger.dev as opposed to building an app on like Verso. Um I do use trigger.dev for certain things. Um but there's not like a good reason for it. I I really don't know like like I've used Trigger like I know what it does but I I don't actually know how you would host an application on it.

1:12:30 Like I don't know if it hosts code like I assume I thought it was just like a place to kind of make API calls and like scheduled jobs and stuff. Maybe maybe you know more about trigger than I do. Uh no no not really. So I was actually explor exploring it. Uh thought it was like a good alternative to anything. Um, I haven't really gone deep into it, so just wanted to get your thoughts on that. Yeah, for sure. I mean, I think, uh, Trigger is a great tool. I have it plugged up to, um, Mother actually to, uh, run scheduled jobs, but that's really like I don't even know if I need Trigger for that. Honestly, I don't even think I do. I think um, it's funny. Uh, Claude recommend or cursor recommended that I use Trigger for

1:13:28 for mother. So, I'm like, okay, you cursor knows best. I don't know. Um, yeah. All right. All right. Thanks. Wait, is there one more? There's one more. No, no, that was all. Somebody asked Azie has a question. Sorry, I have a question. You keep saying you're talking to cursor like cursor has its own like uh because I use superet uh so you're talking to cursor cursor has its own uh I guess model is that what you're saying or you're talking to cloud or or codeex and cursor cursor is just your ID.

1:14:06 Uh I just use cursor for coding. Um and I I exclusively use the agent view which if uh if you have a small brain like me I would recommend. Um, and I use Whisper Flow, so I literally just like sit there and ramble for like five minutes about whatever nonsense is going through my mind. And uh, yeah, but the L you're talking I mean the the models are talking to the usual ones like cloud and codeex within cursor. Um, well, I'm I'm I'm using gro 4.6 right now because I'm out of uh usage on all the other models. Um, is that the question? Yeah, I just was just trying to understand what you were talking. Makes sense. Thanks.

1:14:55 Yeah, for sure. I'm using the most redpilled model there is, I think. I don't know. Oh boy. Um, is based on who you're selling to, though. [laughter] Yeah, no comment. Talk about that all fine. Um, I actually have to do do have to hop now. But, um, thanks for thanks everyone for joining. We'll do more of these. Um, and we will see you next time. Cool. Thanks everybody. Okay. Chowo. Chiao. See you.

First public Apero Maven session, on the Apero YouTube. Ben's pitch: a CRM is three data warehouses in a trench coat, and you should be able to tell an agent what you want, host it, and get an outcome. HubSpot is limited once you try to program it. Salesforce is a different story because of the CLI. GitHub is just a repository. What matters is the path: scope the problem the way a manager would, write it in Cursor or Claude Code, store it in GitHub, deploy to Vercel or Railway. People who are good with agents are good at assigning work, scoping a deliverable, and checking it. He did not 'learn to code' so much as learn how to massage Claude into doing the thing.

Notes

  1. A CRM is three data warehouses in a trench coat. In theory an agent can program it. In practice HubSpot fights you and Salesforce's CLI does not.

  2. GitHub is the home base so the automations outlive the person. Store the code there, not only in a workflow UI.

  3. Treat coding agents like a direct report: scope, assign, check the work. Middle-manager skills transfer. He says the people who are good at agents are good at that.

  4. Path they diagram: Cursor or Claude Code, then GitHub, then Vercel or Railway.

  5. He will say he did not really learn to code. He learned how to massage Claude. The point is still a small piece of software for your own RevOps: describe the outcome, host it, deliver it.

Watch on YouTube

Get new episodes in your inbox

Let's talk shop.

We try to give away as much GTM alpha as we can on a first call. Tell us what you're thinking about, we'd love to help.