Media/Appearances/YouTube

How To Build A GTM Campaign in 60 Minutes

CyberMarketingCon/Benyamin Holley/

CyberMarketingCon/Dec 10, 2025
Transcript

0:00:00 Okay, honestly, this looks like a quorum, so we'll go ahead and start rocking and roll in. Uh, so good morning, guys. How are you guys doing this morning? Very enthusiastic after uh 4 days of uh partying and debauchery and uh karaoke and stuff. Um, so yeah. Uh well, my name is Benyamin Holley. Uh I work at a company called Syft. Um I'll go into them in a little bit, but uh the concept of this talk is how could you build a go to market plan in 60 minutes with uh with AI? Uh how could you do it agentically if you just came back from the last talk?

0:00:44 Um, and it's something I've uh I've thought about a lot um over the past couple of years because the way that people generally build campaigns, marketing, uh lists, lead building, it's all still like very manual. And um in some ways it really still needs to be. Um, but in other ways I wonder I'm like, you know, when you really think about like what humans do uh in like for instance like Zoom Info or Apollo or something like I I think the robots could do it way better and uh way more efficiently and uh you know complain less. So um anyway um before we get started I have a serious question. I need to volunteer from the crowd. Uh this is going to be tough.

0:01:36 Um, who has a company, nobody's going to answer this question, but who has a company that they work for that they really didn't like? Because I want to use them as an example. I want to build a campaign for them. Um, does anybody want to volunteer tribute uh for a glass door review right now? I need one I just need one URL. That's all I need. Yes. I don't. Okay. Intelligence. What What's their uh what's the URL? Intelligence with an e.com. Okay. And uh we're going to use them for something. I'm going to show you guys something here in a minute. But uh it takes a while. It kind of has to like run in the background. Is it intelligence

0:02:26 like that? Yep. Okay. And uh you'll sort of see this running in the background. Hopefully it works. It works like 90% of the time. Sometimes it'll like get tripped up and I'll sort of explain Clay and what's going on here in a second. But um I want to get back into the uh thing. Um okay. So just a little bit about who I am. Should you listen to me? Well, if you ask me if you should listen to me, yes, I think you should listen to me. Um as a salesperson, I think if I put any doubt in your mind that you should listen to me, u maybe I wouldn't be doing my job. I don't know.

0:03:15 Um, so, uh, real quick background, I've worked all sorts of jobs. Uh, I got into SAS a couple years ago. Um, I used to run recording studios and, um, I didn't know that that was actually my first sales job running a recording studio. If I had known that that was a sales job, I'd probably still be doing it. Um, I think that's the, uh, conundrum that most artists run into, which is they don't know that they're actually in a business. And there's a lot of talented people out there who don't know how to make any money.

0:03:55 And, uh, so I I think I was maybe one of them. Um, I I worked at a one recording studio that was actually like a music education company and they had like a software product, but I didn't know anything about that stuff when I was doing it. And so that was actually my first startup and I had no idea. Um, today uh I work at a company called CIF.AI. Um, we work with MSPs. Uh, so many y'all probably know what MSPs are.

0:04:30 Uh, we sell what we refer to as a service intelligence solution to those guys. Um, and I won't really get into all of what we do, but uh, the principles that I'm going to show you guys helped us raise $5 million. They helped us recruit uh, the CTO and the VP of product from we poached the VP of product from a, you know, a big company in the space and uh, it was like a big win for us and uh, things like that. Um, so we've had a lot of success doing what I'm going to show you guys here today. And yeah, now I'm the head of growth. So I mean, you know, head of growth is so it's like I'm at a 10erson startup.

0:05:22 So head of growth really just means like uh absurdly overpaid SDR as far as I'm concerned. But um so one of the things we're going to talk about here is uh the context issue with AI, right? And um what happens like when a person who's never used and this may may not make sense right now, but it's going to make sense in a bit uh about why I'm talking about this.

0:05:51 The problem with AI, like if you get your grandma to use AI and she's like, "Okay, this is cool." Like your grandma will actually probably find it pretty cool. So this is maybe a bad example, but if you get somebody who's like never used AI before, um they're going to have a very different experience than like if I'm using it, right? because I have thousands of interactions, hundreds of documents, it has all of my data. Uh it's a huge security liability. Um you know, it knows everything about me. And so like when I ask it a question about what should I do with X, Y, or Z, um it can answer in a very different way.

0:06:40 Um and the same thing happens with like sales tools that use AI. Um essentially, you know, it's like you have to rebuild the context window every time. Um, and it becomes like very cumbersome. And uh, the tool that I'm going to show you guys today, Clay, uh, many of you are Can I just get a show of hands? Like who's actually familiar with Clay? Who's used it before?

0:07:08 Who knows anything about it? Okay, so this is a pretty relatively um, educated audience, but um, there's there's limitations with Clay. There's limitations with all of these like workflow tools like NAD and stuff is that they don't have a context window and you keep having to like refeed context into the next piece of the workflow and it becomes really like cumbersome that's it's not like the end state for AI um right now we're like still early like messing around with these tools and um you know these days like I do a lot of stuff in cursor I do a lot of stuff in Claude Code um you know a lot of the sales stuff that I do is in those you know um and if uh if somebody had asked me to do this talk like a few weeks after I'd been asked originally this would be a talk on like cursor and Claude Code for sales um but uh anyway Um, and so yeah, I sort of got into like what does the traditional ABM playbook kind of look like?

0:08:22 It's like you kind of pull a list, uh, you know, filter by vertical, send them a bunch of emails or whatever. There's there's a lot of really big problems with this. Um, first of all, the verticals like they don't really tell you anything. The workflows that like I've built and what you're going to see here in a second, let's actually see if it's working properly. It is. Um, it's like the average company has like, you know, thousands and thousands of companies in their in their CRM, but they don't really actually know anything about those companies past like a URL or something.

0:09:05 And what I've done like at SIFT and I've worked with, you know, in the past, you know, consulting clients with this, it's like, let's take all of the companies in your CRM and like have an agent go to their website and like read the website and like, is this company actually somebody who we want to talk to, who we want to sell to? um the amount of like garbage data that the average sales team has in their CRM uh it's it's absurd and um you know becomes a graveyard and so like that's what we want to kind of try to avoid and so what I'm going to show you guys uh has a lot of use cases uh CRM enrichment like mass CRM enrichment is kind of one of them right um and so like for us uh here at Sift one of the first things that I actually did uh even before I started working here was I did a lot of research on the uh total addressable market because you know these guys asked me to come work for them and uh you know I'm like okay well I'll go work for it's just it was a preceded stage startup at the time so was you know kind of a big bet right and I'm like okay well let me actually see if the the market is able to support what their what their uh ambitions are right which is the same ambitions probably that every ventureback founder has which is you know we want to become a unicorn and sell our company to service now or something like that right um And so I I uh went into Clay.

0:11:00 I actually went into a whole bunch of different sources to get all of the MSPs in the world, but you know, right now we're just selling to the Americas. Um and said, "Okay, if we can capture like 1% of the market, how big could this company be?" And uh you know, I gathered about 65,000 MSPs worldwide. I think there's probably more. or most of them are really small. So the kind of type of tool that we sell is not super relevant for a lot of these smaller shops. Um but I think like as sellers, as marketers, a lot of times it really benefits us more to think about things like a financial analyst or a or a or an investor than like somebody writing copy. I guess I I don't know.

0:11:54 That's how I've sort of thought about things as far as like the data layer of my work on the GTM side goes. Um, and honestly, I think this is a good general principle for like anybody who's like thinking about taking on a new role because people are like, "Oh, um, you know, come work for us. You know, it's going to be it's going to be great, whatever." Um, and, uh, I think like most people don't really take the time out to analyze the market that they're going to go work in and see if these goals that these companies are, uh, trying to hit are are realistic or or anything like that.

0:12:40 But you can do this stuff on your own with a little bit of research. Um, and this will be a piece of that, right? Uh, anyway, so what I built here, um, it was, you know, I guess it was a pretty simple clay table, right? And the there's a couple of processes that, um, are sort of, uh, simplified just for the purpose of this like demo, right? Um, but basically, uh, the reason I wanted a URL, let's see if it went ahead and worked. Um, so this is Clay. Uh, and I guess if people don't know what Clay is, it's basically I don't know what you'd call it. It's like a no clo no code workflow builder for sales. It's got a bunch of APIs plugged into it.

0:13:34 Um, mostly like surrounding data enrichment, but um, you can also hook it up to a sequencer or your CRM. Like um, at Sift I have Clay hooked up to Salesforce and our sequencer and outreach and all that kind of stuff. Um, and you can just like sort of execute long form workflows like uh you know I have uh this probably isn't like super impressive but maybe somebody will get something out of this right like uh one of the workflows I have is anytime somebody fills out the type form on our landing page it uh we don't ask them we ask them for their email address right and um you know it just go the Clay goes and finds their LinkedIn determines if they're a good fit uh you know adds them to the CRM if they're not already in there enriches other sorts of data.

0:14:34 Uh I think this is kind of like table stakes for like sales teams these days or should be but I know a lot of folks are still kind of operating on a uh I don't know more oldfashioned type stuff where they you know get a lead and I don't know hope that it's fine um and don't do any research before they call it and uh I was talking to the uh he's not really like an SDR but the kid who works for me and he's like yeah I didn't even look at the URL before I called these people. I'm like, dude, like, come on. I'm giving you a lot of ammunition to know exactly who you're talking to when you call them. You should use it.

0:15:26 Um, so anyway, so I guess like how do I put it? Like I've always had this concept that, you know, you should be able to build a campaign just from a single URL. And it's uh and in this case it was intelligence, right? And so what I did in this first table is I had Clay go do a bunch of research on these companies on this company, right? And I don't know anything about this company. Uh I'm going to assume that most of this information is accurate. It's usually pretty accurate. Uh so intelligence's ideal customers are primarily high growth SAS and technology companies and cyber security firms blah blah blah blah blah. So this is like pretty good research.

0:16:16 Do you know if this is accurate? Yeah. And so that's one of the things about AI is um these days I think it hallucinates a lot less uh than it used to, but it is still um not perfect. And uh these are all the different sort of like data points that you can pull just from like kind of looking at somebody's website um you know other sources. And the reason I think this stuff is important is because if you want AI to work in sales, like what I find myself running into over and over and over again is like you get uh generic AI slop answers from Claude or Chach or whatever.

0:17:02 And it's not because like the AI is bad or something. It's because it doesn't have the context that you have spent years and years cultivating. Um, and so in each one of my interactions, I want to hand it this amount of context so that it actually knows what to do, whether it's qualify a lead or write an email or uh score a deal or uh, you know, determine next steps from a call and all that kind of stuff.

0:17:35 It's we do a lot of like homebrew stuff which is like totally not scalable and in two years it's going to be like a a nightmare to manage. But uh right now it's really cool and fun and uh novel and uh you know analyzing call recordings and stuff uh and uh sending them to a data lake is you know kind of the stuff that we're like we have fun with. But um and so I what I had to do is analyze the ICP uh you know figure out who the typical buyer titles are.

0:18:14 Give me a little synopsis of the ICP. Um and then what I had to do is go find me just five companies because we're going to extract contacts from these companies. Um, and this part is actually something I really, really dumbed down for the purpose of this demo because there's a better way to do this actually, and I'll describe it to you. I wish I had like a good visual, right? What this piece right here where the AI goes and grabs companies, you could just as easily structure this query as like a JSON and use the Apollo API, for example, to pull all of the companies in this company's TAM, right? And then you would use um AI to basically score each one of those companies, say is this a good fit, is this a bad fit?

0:19:11 And you would have thousands and thousands of companies. Now, so why is that important? Um so first of all, uh I was talking to a buddy of mine who took on a revops role at a company. And um classic uh you know it the stuff that like ventureback startups mess up, it never ceases to amaze me. And um they were like they he's like they hired me and they have these revenue targets and we don't even know if the market's big enough to support it. And I'm like dude you got to do this research ahead of time man.

0:19:52 You can't take a job when you don't know like what's going on. And um he needed to expand their total addressable market. and like if you don't know what that means, everybody here probably should, but uh expand the total addressable market that they are aware of, right? And so um something like creating a JSON query that just goes and uh uses the Apollo API to go grab all of that information without you having to like sit there and click through a bunch of, you know, uh drop down menus is actually like a way more efficient way to do things, I think. And um the robots are way better at that stuff than we are. They're going to recognize uh things that we're missing. Um they're going to know certain things.

0:20:46 And I mean there's always like caveats to this. People talk about like, oh, we want a human in the loop. We want a human in the loop. And uh I don't I don't think we actually want a human in the loop. Um, I think we want to give the robots enough context to go do what they need to do, what they're actually good at, what they're better at, what, you know, some of these things that they're actually better at than us. And, uh, you know, selecting a bunch of items from a drop down menu in Apollo to get the right leads is actually something they'd be way better at than us.

0:21:33 Um, if they know what we know, though. That's the caveat, right? And so, this is how we're going to uh, give them what we know. So anyway, that's how I would have done this if I had wanted to go more in depth, but I didn't want to, you know, or else I didn't want to be here for three hours. They only gave me two hours for this talk. I'm just kidding. Um, I didn't want to spend that much time on it. So, these are companies that the AI thinks would be good for intelligence. Are is this like semi accurate? Like, could you go sell to base 10 or Postthog?

0:22:18 Yeah. Um, and so it gave me the, you know, URL basic stuff, a fit reason, sim grep, that's a funny name, is a developer friendly appsac SAS platform. I don't know what any of that means, so I'm just going to have to trust the AI here. Um, and so what we did there, uh, is now I have an accounts table that these accounts were sent to, right? And uh you know, for this purposes of this demo, I just wanted to grab like five of them. But you know, you can imagine if you had like 5,000 of them, uh you know, you just have a lot more leads.

0:23:03 But um for in this instance, we're we're just trying to keep it really simple, right? Um and so this is this is the part about context that I sort of wanted to talk to you guys about, right? So from the preceding table in order like this is the problem with AI tools like this as we use it right now.

0:23:28 This is why I'm so bullish on like using cursor and clawed code for GTM applications, right? Because you can store like a markdown file with all of the context that you're ever going to need about your business, about your customers, and it's going to be able to apply that to every like piece of a workflow. Um, but like if you want to do this with more like traditional tools, uh, I don't think you can do this at all really with like Apollo or Zoom Info or any of the like I don't know uh I don't even know what you call it, more legacy sales tools, right?

0:24:12 And um, often times what I find uh, I've demoed a ton of ton of AI sales tools over the past couple of years. Um, and what I find about most of them is that really functionally they're just they're a clay table with a UI on top, right? Um, and I think that we're we're really limiting ourselves like if we accept somebody else's default settings because like when you think about like software, how's software designed? Like software is designed with the software engineers and the product managers and whoever they give you default settings and that's how they envision the product like being used. Like that's their philosophy on whatever process uh they're helping you with. And I don't believe um I this is going to sound weird, but like I don't believe in their default settings.

0:25:08 Like I don't want to use their default settings. I want something with like true malleability, true flexibility, and no code stuff. Clay, nad, like we're sort of getting there, but I really really strongly encourage you guys to go check out cursor and Claude Code. Um, I actually built this presentation in cursor. I'm not a developer. I'm definitely not a web designer, but I wanted it to look like a Windows XP. And so, um, that's what I had to do. But, um, anyway, so like I would feed, uh, the next piece of context to the next table, right? And so I fed all of this stuff to this table and then I asked it to go find buyer contacts.

0:25:58 Okay. And so again here you could have repeated the sort of a Apollo workflow where instead of like hey you know AI is going to go get some people off of LinkedIn, right? Uh hey we think this person or this person would be good. But there's a way better way to do this. Like I said, structure a JSON query, feed it to Zoom Info or I don't know if Zoom I I I think you can do it in Zoom Info, but their API you have to their API starting cost is like $15,000 a year or something. So, um which is a little bit too lush for me like at this stage.

0:26:44 U maybe series A I'll get the uh Zoom Info API. Um but you want to feed it this data before it goes on and does the next thing, right? Um, and so let's see, it pulled three contacts. The head of engineering, the head of forward deployed engineering, that's a new one, and the president/GTM operations um at base 10. And so it's proposing that yeah, we go find those people, right? Um, okay.

0:27:14 So there's contacts. Um, we can take a look at the contacts, but uh, right now it's actually I think it's doing some analysis because I have another, um, section here that asks it to do some stuff. Um, real quick, I I I don't think this is like revolutionary or anything, but if you're pulling contacts, I hope that you're validating all of the email addresses with something like Lead Magic or a whatever million verifier. Um, I know for our own email campaigns, it's super important for us to know who's got a catchall email, who what uh what MX provider they use. And in this case, everybody here is on Google Workspace, which is a pretty weak security gateway. Um, you know, usually the folks that we're prospecting into is like Proof Point, Mcast.

0:28:09 They like absolutely, you know, kick my ass when I try to email them anything. Um, so I love it when I see a good uh Google Workspace because uh I know that I can uh send them a message. Um, and that's the sort of like optimization that I like to do on emails, too. Like I don't you you shouldn't just uh this isn't an email security talk, but uh you you if you're doing a proof point MCAT customers, it's going to be a lot harder. And so there's like particular ways to structure emails like um anyway I won't get into all that but the other thing that I do with this contacts table is um send these folks to the CRM and so Clay has a Salesforce lookup.

0:29:02 Uh none of these folks are in my Salesforce, which isn't surprising because I don't sell into any of these people, but um if they were, you'd be able to see like the CRM link, all that kind of stuff. This is the fit assessment. This is sort of like a one of the problems that I ran into as an SDR is that like my revenue operations team would, this is way back, way back in the day when I was at Panda. worked at Panda and uh I remember I was on the outbound team there and the revenue operations team they would give me like a list of people to call and I would look at the list and I was like this list was not created with any sort of logic or rationale.

0:29:57 It's like literally somebody just pulled some variables in like a lead finder and like sent me a list. And you can do that and you could gaslight your SDRs by giving them something like this. So they might actually call them.

0:30:14 Um, and what I mean by that is like what we want to do is give people like a conviction around like why should we reach out to this person? What about this company says yes, I want to talk to these people. Um, and I'm not saying you should gaslight your SDRs, but like if when I was an SDR, if somebody had given me like, you know, 500 words of like reasons why they want to call these why we want to call these people, I would have probably been like way more gung-ho about it than like, oh yeah, here's a list and um, you know, good luck. Uh, so you can sort of take a look at what what we have here. Ship Fox's product Premiere faster, cheaper, GitHub actions, blah blah blah blah blah. So that's kind of a lot of data, right?

0:31:13 Um, so the next piece of this workflow, uh, you know, sort of, well, just to go over this real quick, like sort of similar, we want account I want account notes on accounts in Salesforce. Like, uh, I don't like having blank account fields in our Salesforce. Um, for most of like the larger, uh, prospects, I have like a bunch of open job listings that we've pulled. I have a bunch of uh recent news like did they acquire a company? Did they get some kind of funding like I do uh a lot of very automated homework on these companies that all goes into Salesforce and um there's actually it's more than just like I want data for an SDR to use or something like that because think about this right Salesforce. Okay.

0:32:07 So, one thing is a lot of companies um larger companies, smaller companies too, but like a little a lot of more more forward thinking AI companies who have like a little bit more of a technical sales uh team um are moving out of CRM into data warehouses. And um I got uh I got a funny question uh a couple of weeks ago from some of the folks on my team. They were like, "Why do we use Salesforce?" And I had to sort of like explain to them my master plan for it or whatever. And they're like, could we just like do this in Superbase? Like if we're gonna you because everybody on the team uses Cursor, even the sales team. I I do all of my Salesforce work out of Cursor.

0:33:02 Um they're like, why are we doing this in Salesforce? The UI sucks. Whatever.

0:33:07 Couldn't we just do this in Superbase or like Snowflake or something? I'm like, okay, yes, we could. And one day I probably will uh mirror the CRM inside of a data warehouse so that we can uh do stuff like, you know, manipulate finance data and product data. uh you know we use metabase for like user analytics and stuff like that and um eventually we'll do that but right now it's in Salesforce but the reason that I want things like job listings M&A stuff etc etc even if those things aren't like 100% current Salesforce whether we stick with Salesforce or we go to Superbase or whatever wherever we uh maintain our uh customer data um and our prospect data all of that stuff is going to be used as the data that like feeds the AI brain of our company.

0:34:05 Okay, so that's sort of like what I'm trying to build up to with all the things that I'm doing here, right? Um, inside of Cursor, I can just ask questions about what's going on in Salesforce. I can just say like, "Hey, um, uh, what's the biggest what's the most common blocker in our deals over 100K?" You know, things like that.

0:34:30 What why is this deal not moving forward? um a perfectly natural language, but uh if there's no data in there um well then you know it doesn't really do us any good, right? Um and so the same thing for companies that we maybe haven't talked to or whatever. We could be like what are the you could ask a natural language query like what are the companies in Salesforce that have the most employees or most technicians because we can do deep research on that in claude. It's not super reliable to get uh because we sell the tool that we sell at SIFT. It's you uh MSP technicians, MSP service desk managers, like those sorts of folks use it. So, I don't really care like who the CEO is really.

0:35:23 Um you know, things like that. What I want to know is how many technicians do they have? How many heads could we possibly sell to at these companies? Um and it's uh something that we take into account when we get a customer. We say what's our uh average level of account penetration? And that's a field in Salesforce. And so you could you could ask things like um I don't know what's our average level of account penetration. What's the biggest account with no with the lowest level of penetration? Um you know where where do we need to spend our time expanding accounts that uh we currently have?

0:36:08 Um and so all of that comes back to like what's the data in Salesforce? What's the data inside of the CRM? Um and so uh you know uh growth signals, things like that, right? like uh do we have a customer and uh do we have a customer who has recently acquired a company right well we need to go talk to that person because they just acquired a company with 40 technicians which could be quite valuable for us um and to that point I actually haven't implemented this 100% yet at recommend a workflow like this to quarterly uh refresh your CRM because a lot of every a quarter of your CRM goes stale per quarter order.

0:36:56 Um, and so, uh, what I do, what I'm doing right now is assigning a number between 1 and four to each account in the CRM inside Clay. And I actually have our entire CRM inside Clay um, in a spreadsheet kind of like this. You know, it just pulls the uh, Salesforce account record and, you know, enriches all of the data. And, um, you can set up a quarterly enrichment. say I want to run uh you know job postings and news and all that stuff once per quarter on a random number between one and four and at the end of the year your entire CRM will have been reenriched with all of this data. Um so I know that was kind of a long tangent but uh I wanted to talk to you guys kind of where I was going with this.

0:37:54 Um so Salesforce now this is what would actually just go in uh for like an account executive or somebody to read on the account level. a strong prospect. Base 10 is a high growth AI inference platform. Yeah, I I interviewed at B 10 before I got the job at Sift and uh I mean I'm pretty happy where I'm at, but um that would have been a cool opportunity because they definitely have it going on.

0:38:25 Um so yeah. Um, so this is sort of like I guess this is the last piece of this workflow and all of these concepts that like I'm talking about they like so I came from I used to run a lead genen agency right and basically what that means is that I spammed people for money um I don't know you guys have done that um but like uh sending emails writing personalized emails is like sort of the last piece of this and I don't know how they turned out because I haven't looked at them yet because we're doing this live but uh I had AI I write some emails and I'll show you the prompt that I use and all that kind of stuff.

0:39:16 Um, one thing I want to talk about though is that like AI right now it how do I put this? We optimize AI for a dopamine rush a lot of times uh rather than something that's actually useful. And I think the people who are building these tools to some degree for consumers are also optimizing them for a dopamine rush rather than something useful. And so like before we look at these emails or whatever, um I guess my point in saying this is like whether these emails are good or bad, we'll find out here in a couple minutes. But a lot of times what comes out of like that chat window is not useful at all. Um it literally it's just like AI slop and um without context, without like direction, um you know, a dopamine rush is not useful, I guess, is my point.

0:40:16 Um, one of the other things that that I've noticed about AI, uh, and these are general principles that u, you know, aren't necessarily my original thoughts, right? Is that the people who are really effective using artificial intelligence and doing stuff like this are also the people who are really effective at managing their own teams? Um, because when you think about it, like what does an agent actually need? There it is. Agentic. Um, like what does an agent actually need? It needs context. It needs direction. It needs what does good look like? like these are the things that I try to implement into my prompting um to make sure that I get something that's not just AI slop.

0:41:05 We'll see how good I did, but um and the other thing that this is just something I want to mention uh it may or may not be related to this talk, right? Like AI in a lot of ways is kind of like biblical. Okay. And so like what do I mean by that is that AI is made in our own images. Um, and for better or for worse, uh, the people who created AI and the majority of content published on the internet, uh, was generated by, um, you know, white 20 to 35year-old, uh, liberal capitalist men. Uh, so, and I'm not, uh, saying that is a good or bad viewpoint to have.

0:41:52 Uh, I'm just saying that that's generally like this how AI was trained. And it's and it biblically sort of reflects them. And in that way, I mean that like you know, you know, they say, you know, we were made in God's image and all this stuff.

0:42:11 And uh so the way AI acts, it acts the same way that we do really. And it's funny because like all the faults that we sort of attribute to AI, we're like, "Oh, AI is hallucinating." I'm like, "Oh, and you're 100% connected to reality. I suppose like you never hallucinate." Um they're like, "AI is just making stuff up." I'm like, "Okay, well, we should all take a look in the mirror before we like get upset with the AI for making stuff up." But um let's let's take a look over here and I'll show you what I fed to this table to give it these emails to write.

0:42:56 So I gave it everything, right? I gave it the uh well it doesn't really need the MX record, but um it's in there because this is the table where we would uh then uh put it into the sequencer for sending, right? Uh I gave it the reason that the company was a good fit. I gave it the pain points that intelligence addresses. I gave it the IC ICP synopsis. Um the strategic objectives of the company which was a another field we had it do research on like what's this company doing? What you know what what kind of initiatives do they have going on right now? Um targeting notes. Um which I guess is like I don't even remember what that is supposed to mean but it looks like uh just notes for targeting.

0:43:51 So I'm missing context on the stuff that I actually did. Then I have the product overview like what does Sgrip do? And so when you think about like uh when I think about like the emails that I've sent and have actually written by hand I'll tell you a quick story. One of the biggest um customers that we booked since we worked at Sift, one of well one of the biggest customer that I booked since we worked at Sift. Um what I did to get that meeting was I I used OBS, which is like a screen recorder, and I went and played back a webinar that the guy who I wanted to talk to did, and he's like the head of AI innovation at this big MSP. Um super cool dude.

0:44:45 Um, I won't say who it is, it doesn't matter. But, uh, and I, uh, then uploaded that webinar recording to, um, a transcription tool and I got the transcript of, uh, you know, the webinar. Um, and then I put it into Claude. I This is way too much work for one email. Don't do this. It was It is a really big customer, so it worked in this case, but 90% of the time that I put this much effort into stuff, the email probably goes straight to spam. So, I don't normally do this and I tell people not to do this anyway. Um, and so anyway, I I did all this analysis on it. And what like what we kind of want to do in an email is we want to like enter the prospect's mind and the conversation that they're already having with their teams today.

0:45:45 Um, which is why I look for recency, uh, news, jobs, hiring. Job postings are actually the best place to get data about a company. Um because they tell you they tell you all this stuff that they would normally like guard with their life like um you know what data lake they use and like what email security stuff they use and there's there's so much information out there about companies that's just like public on the internet that you can ascertain by like getting you know using the header on their you know an email reply or something.

0:46:25 It's it's crazy. Um, I I'm actually like like one of the things about email prospecting is it really really resembles social engineering. And when you think about it, all marketing is kind of social engineering because you're getting people to do things they wouldn't normally do so that you can have money. Um, and so I I sell to, you know, like a lot of y'all probably do, sell to people who are cyber security professionals. And uh, anyway, you have to have a great product and you have to keep that in mind. like, you know, people will get upset and all this kind of stuff, but uh at the end of the day, I'm not trying to hack anything or do anything un illegal or anything like that. So, anyway, um well, moment of truth, right?

0:47:21 We'll see how these emails look. Um so, the first thing that I noticed that I don't like is the re it keeps doing this for some reason. I've been testing this out for a while and uh send sending an email to somebody and making it look like it's a reply to something is not good. Um, but when you start working through this kind of stuff and like seeing what the AI actually does, like you can sort of see why a lot of the faux paws like happen with um, uh, email marketing because like a lot of people are totally relying on AI to do this stuff and um, AI. Yeah, it doesn't really want to do the right thing. It requires a lot of direction.

0:48:13 Say, so your team at Simgrip runs apps tooling over Google Workspace. Teams at your scale often juggle false positive triage bottlenecks and developer handoff friction. I don't know what any of this means. I don't know if this is a good email for intelligence. Um, we help similar security and engineering groups reduce review time by about 20% by tightening triage logic and remediation workflow across the platform. Could be off, but if these issues pop up, open to a quick 15-minute run through later this week. So, I would probably rate that at like a six or seven. Um, it could be a lot better, but for the purposes of this demo, I needed something that would work no matter what random company we chose. Um, so like things that could be better.

0:49:07 Uh, I always try to do things that are presumptuous. How do I put it? We want to presume that we have a viewpoint and an opinion on somebody's business and and that we've actually done some research and we're like taking the time out. Now, you may say, "You didn't take the time out. You had AI do all this." Well, yeah. Yes, I did. I took the time to make this clay table and uh that's what I took the time to do. And um but you want to at least appear that you know something about them. And like I think from a a human perspective, what happens if somebody like replies to something like this?

0:49:54 Um I pro I don't know if I would reply to something like this uh personally. And that's really the that's really the gold standard, right? Like if I wouldn't reply to an email, then I shouldn't send it. And so when I take time to send emails by hand, um there are things that I would actually reply to like the guy's webinar that I downloaded the transcript and used his own verbiage in the email. Um I don't remember what kind of verbiage he was using, but like that's kind of the power of like AI and uh sales like if if there are figures out there. Now, not everybody does webinars, right?

0:50:40 So not everybody's going to have that kind of data to go out there and grab. Um but there's other stuff. There's cultural I don't want to say memes. There's like cultural artifacts on websites about companies on on people's LinkedIns that like you know uh val like for example like values that a company might have might be on their website or they might be on a job posting and it's like like that's the kind of language that we want to grab to get somebody's attention. Um, because as much as like product market fit matters and all that kind of thing, this this company that I'm actually talking about, uh, who I sent the email, like we had a lot culturally in common.

0:51:31 Like this guy had a lot of the same thoughts about AI that we did. And we're like, you know, we're like on the same page with a lot of this stuff about where we think it's going. Um, you know, how the implementation of AI in the MSP industry is probably going to evolve. You know, that's why we think and for an early stage startup, a lot of times that is kind of how you have to get people in the room. You have to say, "Listen, we we might not have a fully baked, mature, enterprise ready product, but culturally culturally we're aligned and um we think you should talk to us because we want you to have access to this when we, you know, change everything, right?" Um, so let's take a look at one of these other emails.

0:52:28 Yeah, I don't like the quick followup. Uh, like I don't know what that means. I specifically asked, let me actually show you the prompt. That would probably be more helpful. So you are a skilled SDR and direct response copywriter who moonlights as a private detective. Uh you will generate a succinct fiveline email to the contact specified using only factual information available in the referenced data source and you can see what data was actually referenced. Um, all of this, right? Um, keep the body properly capitalized and punctuated under 150 words and include a single minor humanizing error either in the subject or the body, not both. Um, so I you guys might not like this, but I really do believe this that if you're going to use AI for communications or posting content or whatever, um, you have to humanize it somehow.

0:53:26 And uh I kind of like the uh Torque guy, which whose talk I love, by the way. Um I love Torque. They're awesome. They were they're in the same industry as us. They serve MSPs as well. Uh think about like think about your inbox, right? every message is properly capitalized, is properly punctuated, has a nice signature, uh, you know, is, uh, verbose enough to, I don't know, explain whatever the product that these people are pitching to you is. How could you be different? Like, how could you do something different? How could you get someone's attention?

0:54:06 Well, you could at least make things appear to be human by adding some spelling errors or punctuation errors. um you know and and there's other ways to do that, right? That's just an example that I'm using here. Um so, and anyway, the the story I was going to tell is uh I had this client one time when I was back when I was running my agency um who I did this for and and when I worked with clients back then, I did not ask their permission to do stuff like this.

0:54:44 And uh you know, better to ask forgiveness and permission and uh usually it worked. for these folks it didn't and they didn't really you know anyway they they weren't we weren't like a good fit because they didn't have a truly differentiated product kind of like um you know they were saying at tor at the torque talk right the torque talk um in order to do marketing in order to do sales you do have to really have a truly differentiated product you have to have something amazing like if somebody sends me an email um into my inbox in order for it to get my attention it has to be literally I'll I'll put it like this when I worked at Panda I asked the CEO and the CTO Oh, and uh like the VP of HR or whatever.

0:55:41 I was like, "Hey, um because I was a new SDR and I kind of wanted to know this stuff." I'm like, "What would it take to get you to respond to a cold email?" And they said, "Well, Ben, you would literally have to appear to be reading my mind with whatever we are dealing with at that very moment." And uh so that's kind of why like I do what I do. Um because I do believe that like when I think about myself, there are a lot of things right now that I'm dealing with at CIF where if somebody emailed me and I'm going to solve your um Salesforce adoption problem, right? Uh classic problem. Um I'd be like, "Okay, well, all right. Let's let's chat about that.

0:56:34 What do you got?" Um and I remember, you know, whenever I make cold calls to people, anytime I book a meeting, people are like, "Oh, well, you know, I don't usually take cold calls. I don't usually book meetings with people who called my cell phone." Well, of course you don't. Like, nobody just takes calls with people who call their cell phone, right? But if I literally like the way I view like cold outreach is if I'm looking for people who are like trapped on a set of train tracks and I want to push them off of the train tracks and if somebody pushes you off the train tracks, whether it's like, "Hey, I noticed you guys aren't SOCK 2 compliant or whatever, you know, um you're going to want to chat with them." I don't know. I would think so.

0:57:32 Just out of like courtesy, right? um you know, if somebody like throws you a genie, a lamp and a genie, you're going to want to talk to the genie a little bit, right? So, that's kind of how I view cold outreach. And, you know, a lot of people, um, you know, it's it's hard. Uh, most people aren't going to respond to you. Most people are going to be annoyed, uh, to to a certain degree. Some people are going to be rude, uh, which is fine.

0:58:08 But um the people who you actually connect with and who when you start a relationship and get a customer and like actually solve their problems um you know all the other stuff doesn't matter right so we can keep going down these this prompt um so factual reference only factual details subject line formatting subject must be entirely lowercase and feel like an internal memo without sounding spammy and so what I mean by that is like I I think uh I want to say this in a uh politically correct way. Some of my more experienced colleagues do still send emails like as a group uh rather than slacking me. Um but being a a spoiled millennial, I'm like why would you email me? Like just send me a Slack message.

0:59:01 Like um but that still does happen and especially a lot of bigger companies, email is still a more like prevalent way to communicate. Um, and so if you had something where it was like, oh, ops bottlenecks I or like you don't really know what an internal memo kind of looks like. So, you kind of have to do some guesswork. But, uh, that's the kind of stuff that I want to do. Do not say, "I was impressed. I I noticed your LinkedIn or any subjective admiration." Uh, when I talk to my uh, Eton, who's the kid who works for me, I always tell him like, "Do not glaze people in emails." Like, we do not want to be glazing anyone cuz he's 22.

0:59:52 So that's the verbiage that he's uh familiar with. Um and so here you can see an example output. You always want to again the same principles apply with management as they do AI. Uh if you tell somebody to go off and do something and they don't know what good looks like, well, uh good, you know, whatever results you get back are kind of on you, you know. Um and sometimes it is fun to like give less experienced colleagues like a task without explaining anything just to see how they'll do. Um, and then they come back and they have like some Frankenstein monster thing. And so anyway, the same kind of thing happens with AI sometimes. It is fun to like experiment. Um, there's other stuff you can do in Clay like they have document context.

1:00:48 This is um I know we only got 5 minutes left, so I I'll try to hurry it up. Um, and we're actually done anyway, honestly. But, um, I'll sum it up, right? Uh, that context uh, let me go back real quick. Actually, this context thing, this is like sort of the problem with Clay with all these tools is like this document context. I don't want to upload a document context. I literally would have to do that for every AI column. And you can make a template to and like speed this stuff up, but I don't really want to do all that. Like I want, you know, I don't run a legion agency anymore, so it's just SIFT that I'm using this stuff for.

1:01:39 Um, but like I want that context every time. And so with a cloud project, with a Claude Code, that context exists natively. You don't have to like retrain the AI on stuff. Um, okay. We already talked about that. Talked about what Clay is. Talked about that. We talked about Clay's big limitation. Doesn't have memory. Um yeah, I think uh I was talking to um a friend who was a early investor and like shareholder at Clay and he was telling me that like he had a very specific pointed conversation with their leadership about how cursor and Claude Code and these like coding agents are going to start replacing these sort of like workflow builders in that regard very soon and that Clay should kind of take notice of that.

1:02:32 Um because I think like the RevOps teams of the future are not going to log into Salesforce. They're not going to log into the CRM. Like they're going to operate everything from a chat window. Like I don't want to log into Salesforce. I know you guys don't either. I tried to uh set the uh logout time to infinite in Salesforce. And it turns out that they will not let you do that. The longest time that you can go without uh Salesforce logging you out automatically is 12 hours. Um and honestly, it's a terrible security practice to have a forever open Salesforce window. I'm pretty sure that's like not sock 2 compliant or something. So, um, let's see. I have a Google doc with a link to this table that I'm going to send in the WhatsApp group.

1:03:29 I meant to have a QR code up here, but I I guess it didn't work. Um, so anyway, listen, uh, we got two minutes left. I know we ran through a lot of stuff. Uh, I really appreciate you guys. Been a great audience. Um, if you have questions about this or anything I talked about, get with me after the session. So, um, I think we'll we'll probably wrap it there. So, thank you guys.

A live, unrehearsed session at CyberMarketingCon. Ben asks the room for a company URL he has never seen. Someone gives him intelligence.com. Clay starts running in the background while he talks, and he tells them it works about 90% of the time and sometimes trips. The table researches the site, drafts an ICP, finds lookalike accounts (Base10, PostHog), pulls contacts, and writes a fit note an SDR can actually read. He used the same motion at Cyft before he joined: about 65,000 MSPs worldwide, most of them too small for the product, enough to test whether a pre-seed could support a raise. Those principles, he says, helped raise $5 million and recruit a CTO and a VP of product.

Notes

  1. The demo starts from one URL he has never opened. Clay researches the site, drafts who they sell to, finds lookalikes, and writes why the account is worth a call.

  2. He tells the room Clay works about 90% of the time on a live run. When it trips, he explains the table instead of pretending the demo is clean.

  3. Before joining Cyft he gathered about 65,000 MSPs to see if the market could support the company. Most shops are small. The product is not for all of them.

  4. Head of growth at a 10-person startup, in his words, means absurdly overpaid SDR.

  5. Vertical filters do not tell you if a CRM record is a buyer. He sends an agent to the website and treats the rest of the CRM as a graveyard until someone reads it.

  6. He scores Salesforce accounts 1 to 4 inside Clay, with the whole CRM pulled into the table, then writes the note an AE would actually read.

  7. He does not want a vendor's default settings. ZoomInfo's API, as he remembers the quote, starts around $15,000 a year, which was too much at that stage.

  8. Clay and n8n have no persistent context window. He keeps the business in Cursor and Claude Code so he is not refeeding it at every step.

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.