# OpenClaw use cases

Channel: The Next New Thing
Video: https://www.youtube.com/watch?v=GfZpmcStxWA
Duration: 43 min
Language: English
Words: 10608
Transcript page: https://viewrankai.com/tools/youtube-transcript/GfZpmcStxWA

---

[0:00] I'm running an eight-figure business. We're looking to scale it to the moon using OpenClaw. Look, OpenClaw is literally creating a video right here. Let me show you how this thing actually drives customer. No AI theater, actual customers. Let's do a screen share. Presented by Zapier, the AI automation company. Andrew, so good to have you on here. You run Single Grain marketing agency. You do marketing for so many other companies. Can I see how OpenClaw

[0:23] has helped you grow your business? Yeah, I will show you everything. Uh you stop me cuz I tend to talk very quickly. I'm going to jump all over the place when you have a good thread pull on it. So, um I'm going to start with what I have on the screen right now. So, I want to show the receipts. Um obviously, for my business, because we're more of like a tech-enabled marketing agency, um reach matters quite a bit. And so, all these X articles that you see on my screen right now, 99% of it is it's AI created. And so, um you know, you can

[0:53] see this one over here, 95,000 views. The next wave of billion-dollar companies will run on two things: 40 deployed humans and AI agents. That's number one. Second one that you see on the screen is this 368,000 view Car Parties autonomous AI can make you 701 times faster. It's a future business, which we can talk about later. Um but my point is, this is actually getting me reach. And not only that, the reach is

[1:14] nice, but it's actually driving leads. There are, I think from the X articles alone, there already a handful, I want to say four or five, um nine-figure companies that reach out saying, "Hey, we want to do this stuff." And it wasn't like I was actively saying, "Hey, come buy our stuff." Um it's just the thought leadership alone, and it's just me working with OpenClaw day and night and saying, "Hey, what looks like a good X

[1:36] article that we should put out there?" Cuz X getting is giving a lot of reach right now. So, I'll pause for a second there. This is just the the first part. Okay, I've got questions about this, but give me a couple of other examples, because this is more than just a writing assistant that you've built for yourself. Exactly. So, the other thing too is like I I've been seeing these charts on X. I'm like, "Oh, these charts are really cool." So, I open up my

[1:55] OpenClaw. I I have Whisper flow open. I'm literally just talking to it the whole time. I'm just like, "Hey, this chart looks really cool. How can we make one for ourselves based on all the work that we've done?" Um and so, um or or can we make something that's a little different from like an AI angle that would interest businesses? I don't just want the reach. I I want people that are that are interested in business. So, it creates this dot chart that you see on the screen over here. So, how companies use AI. So, it pulls a it's it's a little dated like a McKinsey state of AI

[2:20] post, and it they did research on 1,300 companies, and you can see the number of companies uh that are using chatbots, but also the run the ones that are running autonomous AI systems. It's less than 1%. So, okay. Great. Cool-looking dot chart. How did it perform? 56,000 views on it. Drove a bunch of email sign-ups. Um and this is completely automated. All this does every single week is it's called a chart factory. So, it ingests all my content and everything that I've talked about from my all my YouTube, my podcast, everything. There's

[2:45] an ingestor that runs every night. Okay. Um and then it will spin up four charts or like six charts for me every single week. And then you decide what to post. Yes, exactly. So, and then by the way, it's gotten to the point cuz it's hooked into my Buffer, um which is what I use to post. It's hooked into Metricool, which is our social social media like uh analytics tool. Okay. Um and then it just says, "Hey, here's the captions for you. What looks good?" And then I'm just like, I'll hit reply on it. I'll say, "Hey, this one looks good. Can you make some changes over here? Uh the caption,

[3:10] just make some changes over here. It's good to go to post." Boom, it schedules it. It's it's up for the next day. Okay. All right. What's What's another thing that you did? So, I've been um so I I'm sitting on the couch. I'm like, I see all these Hormozi highlights popping up. I see this channel, okay? So, this is Alex Hormozi's channel, and it's it's just his highlights. I'm like, "Wait a minute. What the heck is going on?" 41 minutes ago, 44 minutes ago, 47 minutes ago, 58 minutes. Wait, he's published like five, six, seven, eight, nine, 10 el- in the last hour? Like, how is he

[3:40] doing that, right? And so, you know what I do? Uh in fact, I can open my Telegram right now. I can I can walk you through my process. I'm sitting on the couch, and I'm doing this with with my with my OpenClaw. You want to see it? Yeah. Okay. So, and by the way, just so everyone can see on the screen, too, um

[3:54] not all these videos are going to hit. He's going to have some that really take off. He's going pure volume here on these. So, um let me pull up my Telegram. of these are doing nothing, and then he has one from 18 hours ago that did 1,800 views. Okay. Exactly. Okay. And is it all social that we're going to be looking at today, or are you looking at different No, we're going to go into the the business stuff. So, I want to go into like Slack and how we're actually using it for like the team for a sales standpoint or an SE like how my marketing agency uses it,

[4:19] basically. Um so, check this out over here. This is my Telegram. This is no different than what your guests have shared. But, this is me on the couch over here, okay? So, how has this channel grown over time? Is this a viable way to grow? And you're just basically sending it the channel, and

[4:33] you're asking it to analyze it for you. Yeah. Yeah. That's it. And I'm like this this is hooked in with my YouTube API. So, Alfred's my chief of staff um bot, so it has more connectors to it. Um so, there's YouTube APIs hooked into it. So, it's like, "Okay, um channel Hormozi highlights created six months ago, 58,000 subs, 8.2 million views, 1,000 videos." And you can see he really jacked up the volume starting in February, 502 here, 341 in March. And you can see the number of views. And so, it it calls out the patterns, um and then it's like, "Hey, I'm like, is this

[5:03] viable for growth?" Yes, with caveats. So, it says all these things over here. It's like, and it's calling out the strategy. Volume for him is lottery tickets. He's just like each post is a lottery ticket. You don't know which one's going to take off. Um so, I'm like, "Okay, well, this is cool. Like, can you explain what you mean over here?" And then, what it does that's really cool, it just keeps going. I'm like, "Hey, do this for me." He First, it was like, "Let me do 3,000 for you." I'm like, "No, no, no, no, no, no,

[5:24] no. Just do do an MVP for me. Okay. Do do do like a small sample size, which I'm going to show you in a second. But, what it does is it it creates these titles over here. It finds the right moments from my Marketing School or my Leveling Up channel, and it cuts it up. So, not only finds the right moments, it titles it, it cuts up it puts it into a Google Drive. This right here already replaces a lot of what your content team

[5:47] is doing. Wait, but how how is it doing it? How is it even actually what access did you give it to go and read your videos and understand what's in them, and then what tool are you using to edit? Yep. Great question. The the real answer to what tool am I using? I have no idea. But, I will tell you what it's connected to. So, it's you this is what I need in order to do that. Yeah, so it All I All I gave it,

[6:09] Andrew, was you the YouTube API. Um and it So, here's the pipeline over here. I'm going to transcribe them using Whisper. I have no idea what Whisper is, okay? Second, LLM reads the transcript, and then all this stuff FFmpeg, I have no idea what that is. So, it just got it just does it. I see. It tells you, "This is what I'm going to be doing and using it." Got it. Whisper is the open um Open AI tool for analyzing audio. Got it. And

[6:32] this is what This is impressive. Okay. So, you don't even know. You just said, "I want you to figure out what I need what you need, and then tell me what it's going to cost." And it does all that for you. Yep. And And then it's like, "It's going to cost you five to 10 dollars a day in API costs." I'm like, "Whoa, whoa, whoa, whoa, whoa. Let's not do that yet." And then, I want to show you the output um because I was like, "Okay, this is all nice. This is all theater, though, if you can't show me the output." Which is actually the challenge I gave to my team, which

[6:54] actually right now my entire team is doing a hackathon right now. Um so, While we're talking. Okay. Yeah. I have This is literally the MVP over here. Okay. So, these are the six clips. So, um let's just pick one of them over here. Um you know, I'm getting chills as I see this, cuz I know how much of a pain in the neck this is to create, but I also I also know that you cannot do this easily with Claude code, that you can't just sit and chat. Okay, let me see. Yeah. And by the way, Andrew, I think I can give you the process for this, too, if you want to

[7:20] roll it out or we want to share with the audience. Um but it's it's like every time I make one of these, I'm like, "Give me the Google Doc. Give me the like And then I want you to organize all of it in Google Drive, cuz I know I'm going to be reusing these things." And if we if you make an app or whatever, you want to plug it into whatever it is you're using. You want to make sure this

[7:35] is repeatable. Meaning, you at the end of each one of these processes, you say, "I want you to create an SOP so someone else can do this." Or Okay, got it. So, this is what it created. Yeah. So, this is this is the These are the actual videos, okay? So, I don't know if my audio is sharing.

[7:48] Let's see. if it's not. It's probably not.

[7:56] Okay. Yeah, I know it's not, and it said it's basically part of your video where you're actually walking people through how to do something. Yeah, this is pulled from one of my webinars, I think, which I didn't ask it to do that. Um and then I think this one over here, let's see. This is your conversation with Neil

[8:11] Patel, your co-host for your podcast. Okay. Yep. Yep. And so, but like the the the the the titles here aren't aren't bad, because all it did, Andrew, was it just looked at cuz you you look at Hormozi's highlights channels over here. It's it's all kind of the same thing. Helping an SEO agency get to 14 million a year. Helping a nail salon build a franchise one. It's all the same thing. Um ver- worse is these are I'm I'm like Actually, these are not these are not that bad for AI generated. Multiple The $500 dinner that landed Lyft as a

[8:35] client. Perfect. Yes. Yes. Or like um it's actually multiple seven figures from a two- one one 200-person event that I did. So. Okay. All right. So, it figured out how to do it. It put it together. Can I see the Google Doc where it uh wrote out its process? So, it made an X article that I literally just published before this. I can find the process for you and send it to you. But, this is how it works. It's like literally here's the entire story,

[8:57] and then there's a prompt down here. Like, if we kind of went through all this stuff already, like the process. But, um the prompt down here, I think is I'll find I'll find the prompt for you, but it's it's all in here. But, look, it generated a seven-page uh you know, X article here. Um and then here's all the steps that

[9:15] you need to follow to make it happen. Let's Let's pause for a second, because one thing that stands out for me is I know when I look at your X articles, you use one of those text boxes that actually stand out nicely in an X article. How does OpenClaw know these things about your writing style? Yeah. So, I I ported over my I have a Claude skill that has all of my um I call it the content repurposer skill. So, it has my best-performing tweets, my best-performing LinkedIn posts. It has everything in there. So, I ported that skill over to OpenClaw, and I said, "Hey, use this this content repurposer

[9:47] skill, and you're going to use this. Whenever I want an X article, you're going to use this skill." Um and then also also the the thing you call that, Andrew, um I don't know how you pronounce it. Is the ASCII diagrams or whatever. So I said, "Hey, these diagrams look like they really stand out. Make sure you have a multiple

[10:01] diagrams in in each of these cuz yeah." Right. Right. No, even as a reader, I find myself actually enjoying those diagrams for understanding the process and and they stand out nicely in an an X post. Okay, so it's just a skill that you give it, it writes the post. How far along is it? Like how does it get you 80 90% of the way there towards publishing something on X? I I think it's it's so D if we're talking about the X posts and by the way, I I found this the SOP for you over here. So I I'm happy to share this afterwards, but so here's the

[10:31] entire SOP if you want it. It looks like it's 17 pages. get it and we'll put it in the show notes for anyone who wants to have this SOP. if we go over here, I think it's it's 95% there. The only thing I need to change Andrew, so when it when it pops it out, so I do make the images using Gemini and I I do have literally I have a loom of me making like four of these articles and so I know a machine can just copy me. So I think that's a that's the thing that people should take and and use for the future so they don't

[10:55] have to keep repeating this, but the key thing is Gemini will make the image. I like these stick figure images or I have these black and white images. I only have two forms. So I have a Gemini gem that just repeats that over and over. The Gemini gem will look at the article and it'll just make this. Okay, I'll just like do

[11:10] the black and white or do the charts. And are you personally going into gemini.google.com to do this or does OpenClaw have access to Gemini? I'm sure I could give it access. I think I've like been hesitant to just let go and let it YOLO, but I'm pretty sure I can. Okay. All right. So I'm getting an understanding of how this is done. Let's zoom out for a minute. Um Well, did you want me to finish answering the question on what I actually do here like the 1% I do and

[11:36] then the the rest I do not. Okay. So I really only focus on the headline cuz sometimes like you publish it and it gets like one like in the first hour. So I do put a lot of time my human time into the headline and really the first I would say the first paragraph or the first five lines. As long as that hooks them in, I know the rest will just keep

[11:54] them going. So that's been my strategy to to getting more views on these X articles. Okay. Um What what are you doing differently about the way that you handle OpenClaw that allows you to create get good content out of it? A lot of us are are just not getting great content. I'm probably spending honestly I'm probably spending 12 to 14 hours a day on it. So there's a lot of business strategy that goes through it or like you saw that how I just was I'll look at a channel and be like, "I want to do that." Or you know, I'll I'll look at how um I'll look at a channel, come up with a

[12:23] strategy, I'll look at the dark charts. I'm just in constantly scanning things that make sense or someone might publish something and say, "Hey, why don't you the Andre Karpathy thing?" I'm like, "Okay, well, how do we make that?" You know, Alfred which is my chief of staff, you have all my goals already. You know what what our team's trying to

[12:37] hit. You know what we're trying to do. How do we make this for ourselves? Okay, boom. I love this. Okay, I think you should make something called auto growth. And then I work with it. So it's basically my thought partner. So I'm constantly talking to it the whole time and I'm just constantly asking, "Hey, what would be interesting content?" Like before talking to you, I was like I'm like, "Oh well, I want to make this good for Andrew." I'm like, "Hey, use the

[12:56] YouTube API. Who is Andrew talk to?" Like and and knowing what I talk about, what is actually going to stand out? So everything I have tabbed up right now was already pre-planned with AI because based on everything that you've done. Okay, one thing I'm noticing you do is is do this thing that I've heard which is to tell it what your goal is and have it know what your overall goal is. What else are you doing that you're feeding it that's allowing you to

[13:17] to make it work? So it has I mean it has a a lot. So Alfred, here I'll I'll give you some examples. So it has all of our goals. It has all of our financials. It is connected to my leadership team. They fill out these KPI docs every week still manually. So it knows how we're trending and it also by the way, it goes into my leadership channel. It's like, "Hey guys, here's how we're trending right now." And it tags people, "What's going

[13:36] on with this metric over here?" Wow. Yeah, and it's like it's very specific too. "Hey, you said you were going to commit to this. Where is this at right now?" And you can't get mad at it. It's a machine. Right. I found it it's like it's good for performance management too. So all that to say is it has access is that has a bunch of API access to a bunch of things that we have. We actually have a an API gateway that we built internally cuz if my team tries to do all this stuff and they start to just connect, you know, client stuff and all

[14:00] that like we need to make sure that we have control over that so that API gateway kind of protects us. But the short answer to your question is it has access to all the goals and has it's you know, it has its own MD files too and then it's constantly scanning Gong. It's connected to my granola as well. So everything. Okay. By the way, that is something that Wade Foster, the founder of Zapier, our partner here, has said if he has a Zapier AI reach out to people with suggestions for how they can improve first of all, it's going to be unbiased and it just give them feedback

[14:32] number one. And number two, people also don't take it in a personal way like they might from another human being who they think has it out for them or soft pedaling or whatever. I I got to tell you this like it's we had someone on our team. She was not very receptive to the feedback. It wasn't until the AI just called it out over and over and it has very specific examples. I was like, "Hey,

[14:51] what do you think about this feedback?" Like I haven't even read the feedback yet. What do you think about it? And I actually didn't read it yet and and then she's like, "It's fair. I'm I'm going to do better." I'm like, "Okay, what are you going to And and then like it's it's it defuses the situation and makes business a lot less personal and it makes business a

[15:06] lot more fun. Okay, what's the next thing we're going to look at? Yeah. Cold email is not going to be dead yet. Okay, but nobody wants to spin up all these different domain names and check the availability and figure out what naming convention works and then figure out the right sequences as well and then layer on my auto growth tool that we're going to

[15:24] talk about in a little bit on top of it. I had OpenClaw do all of that. So we can talk about that and I'm going to show up my Instantly first and then I'm going to move over to to Slack right afterwards. And you you can you can you can lock me down whenever you want to. Okay, let's do it. Um So this is our Instantly account. So what we're doing right now, we're actually um What is Instantly? Instantly is a think of it as cold email infrastructure. Okay, so you can make a different, you know, campaigns. So you know, for warm or cold outbound and it will actually help you buy

[16:00] domains as well. So if you're going to send a bunch of cold emails, you have to have infrastructure for it. Am I making sense there? Yeah, because if you use your same company domain, it's going to start to get blacklisted which then keeps your regular messages from going out. If you create just one sub do- one domain that again will happen and so you need to keep moving. I see. Okay. What beyond that beyond the domains, what else goes into it? Yeah, so the sequences to the copywriting obviously it matters quite a bit cuz everyone's getting spammed by cold email all the time. So the sequences matter

[16:28] and it's been proven by We used to work with the cold email company. They said that I'm going to test this a little more, but they said Latina female names usually convert higher in terms of open rates and things like that. Interesting. Okay, so you want to test all of these things. Exactly. So I'm just like I

[16:43] don't want to do any of this, right? Like I can have my team do it, but they're not exactly they're still catching up on the AI side. So like I'm like, "I'm lazy. Can you just do it for me, OpenClaw?" And I'm literally in Slack talking to it. I'm just like going back and forth and it basically makes these sequences over here. I'm going to I'm going to give you an an example over here. So it says, "Aaron, I'll analyze Configuris presence across ChatGPT, Perplexity and Gemini at no cost. Here's what you'll get." Detailed whatever and it says all these things, right? Now, am I going to say this is the best sequence

[17:09] ever? No, but I'm going to I'm going to just let it run cuz it's it only started running on on Monday, right? For these cold sales ones, but I will say it's already working from a recruiting standpoint. So basically the whole idea here is it'll it'll make, you know, like an eight-step sequence and I'll continue to follow up with them and then over time what I'm just going to ask sequence or are you trying to hire using

[17:32] it? Got it. This is a sales sequence. Yeah, it's already starting to work. So the the hiring one is is working. So we're looking to hire we're looking to hire like a senior executive right now and it's already getting I think the the positive response rate on it is about 3% or so. Generally with cold email, if you're getting like 1 to 5% positive response rate, you're already pretty

[17:51] good. So it's in a good spot already. Okay, so you're using this both to recruit new employees for the company and to get customers and OpenClaw did all of that. Yes, correct. It did all of it. And it has access to this because it has API access to Instantly so it could control it. Correct. Yeah, so that's that's the big thing because I can ask it to do all these things. All all the things I want it to do. So I'll say the sequences, buy me these domains. Oh, I want to use you know, this I want to use these you know, this these names. I want to use this type of

[18:22] setup in what with our naming convention. So like I want to have single grain and with that with whatever else like a modifier on top of it. So it's single grain growth, single grain life or whatever and it'll just make a bunch of them and it'll check what's available and what isn't. But what I also do and I think you might like this Andrew. So So I want five sequences. Okay, so I don't want to write the sequences. Make me five sequences. I want you to create an expert panel, okay, of the of 10 legendary copywriters, persuasion experts, email copywriters and I want you to recursively score each sequence 1

[18:53] through 100 until you score 90 or above. I don't want to see it until it's 90 or above. Okay? So then it forms each of these. Go ahead. No, I'm just I'm just marveling at it. That's smart. So basically you're saying to it, "Not you, Alfred, are going to write this for me, but I want you to have a panel of experts who are going to be writing and evaluating and until the panel of experts says, 'This is a great sequence,' I don't even want to look at it." And it's just iterating and going through it. Okay. Yep. 100% And the cool thing is I'll obviously I'll I'll check,

[19:20] right? As I'm doing this, I'll I'll check to see what it's like and then I'll continue to give it feedback. So as I'm going through, I'm like, "Okay, these look pretty good, but the emails are too wordy. I I want you to have the emails under 100 words. Each one needs to have a call to action and then it needs to be written in my voice too. So take the content re-purposer." So it's pulling from all these sources and then I didn't tell it to make one the first sequence is like focus on like pipeline waste. The second one's on the future sales. The third one's on pilot results, speed,

[19:44] challenger, all these things. And I'm just like, "Dude, let's just test it and see how it goes." And then we can continue to iterate and you know, eventually like auto growth starts to run on it once we're sending enough volume. We're not sending enough volume right now, but once we're sending enough volume, my auto experimentation tool is going to start to iterate on it, and whatever it takes off, it's going to just make that the winner. And almost like think of it as AB testing that's

[20:05] running on its own. Okay, I'm with you. This is exciting. Yep. Great. So, I I'm I have By the way, I'm happy to share this as well. So, like And this is your skill, right? This is This is not even like I This is just uh An SOP. This is an SOP. Yeah, I

[20:18] said give You know what I tell it to do? I don't tell it to make me a skill necessarily. Um I tell it I say, "Hey, I want this to be a plan, and I want you to store this in a in Google Drive." And then some of them like like yesterday literally asked I was like, to your point Andrew, I was like, "Dude, I feel like I've made so many of these plans and SOPs, like which one of them do I use enough where they should be a skill?" And so, I maybe I just ask that once a month, cuz I think there's a lot

[20:39] of skill theater that you see on Twitter. Um so, I want to make sure whatever I make a skill that I actually use. You know, the problem that I have with this is that it does not know Open Claw does not know to go into Google Drive and get its plan out, but it does know to access a skill. And so, if I tell it I use Notion, it just forgets to go into Notion. It's It doesn't have

[20:58] that habit. What do you do about that? Yeah, so I do have limited skills I So, I find myself constantly using the uh content repurposer and the X article one. Um and so, I I think that that's that's what I do. So, I'm like cautiously adding them because I don't want to have um I don't know. For me, maybe it's a limiting belief, but uh I just haven't had like I haven't had the need to invoke like when I log into Claude or or or Claude Code, I don't invoke a lot of skills on my own. So, I have a bunch installed like I have the the the Gary

[21:25] Tan ones. I have like the uh the the Super Marketer ones where you have like a 30 or 50 of them. Um and I just don't use them that much. I have Like what do you do about those? I One my One of my problems with having skills is that they're hard to access remotely. I have to keep asking it to share it with me. I would love for it to be in Notion or even Obsidian in a way that I can access it remotely, because then it becomes like what you're looking at right here. A plan that makes sense to you to look at on a regular basis,

[21:50] but also a skill that it knows it can it has to follow. Okay, but I I got it. I It seems like for some reason you're not having an issue with it going in and accessing these plans. So, No. And you Andrew, you just gave me an idea. I'm like I might just ask it to go one step further. It's like like let's uh let's test making these into a skills for like a couple of days or something like that, right? Like I I might test

[22:08] that with one of the agents that I have. Yeah. Okay. Okay. All right. So, what's next? Let me go into our Slack, because this is what I think is the most exciting, and this is where I think we're getting a lot of usage. Oh, this is fun. So, um this is my my brand team over here.

[22:21] Look, maybe let's just start over here. I'm going to jump around. I'm going to jump over to the sales team. I'm going to jump over to recruiting as well. I'm going to jump over to the chief of staff part two. Um and hopefully this will give people some ideas. So, um By the way, the thing I just showed you Andrew with the the long like the um the Alex Hormozi highlights channel, those are more like 5 to 10 minute segments, okay? So, this morning I was like walking back from the gym, okay? I'm literally walking back from I'm like, "Hey, that Hormozi highlight SOP, why don't we do it for short form, too?" And

[22:51] then it clips this so it makes a short form um folder, and it clips this one over here, and this one's only like 30 to 60 seconds. And I'm like, "Hey, like this is already pretty decent." And then this This is our short form guy over here. He agrees. "Where can I find these clips?" And by the way, when I'm This is like an outsourced team. When I'm showing them all this stuff, they're like, "Oh my god. Holy shit." They're like, "This is crazy, right?" Um and and they do this for a living. Um so, that's that's like a little preview on like you can make like skill almost like the

[23:18] Hormozi skill that I have right now, the Hormozi highlight ones. Uh I'm making like baby skills that are spawning from it. So, um What's that feedback where you said we are missing the uh we're missing the payoff here? That's one that they That's feedback to another team that we're offboarding right now that they're they're in this channel as well. So, um

[23:35] and then this new team will pick it up. So, that's more like human feedback. Got it. Okay. Um yeah. So, let me go over to sales operations here and see how it's working. So, okay. We have a deal We have multiple workflows here. Um in fact, I might just ask it the workflows to give people some ideas as well. So, let's as we're doing this. Uh what sales

[23:55] workflows do you have running right now? Okay, so Arrow is one of your Open Claw agents, and Alfred is your main one. Okay. Yeah. Uh Alfred's the main one. Arrow is for sales. Uh Cyborg is for recruiting. Uh Flash is for social media, and so on. So, um let's do this over here. Look, they're By the way, they're doing a hackathon right now, and um I'm going to

[24:16] start with the the Resurrector, okay? So, the Resurrector is obviously on the on the agency side or really any product or service. You're going to be winning deals. You're going to be losing deals, too. And so, um we have a deal Resurrector here, and again, there's there's pipeline that you can recover, but typically your sales team is too busy to get on top of that because they're like trying to close new deals, right? Um but oftentimes a lot of these I think a lot of these deals that we've lost in the past oftentimes I think my team comes in when they lose. It's like we're number two. And so, what's to say

[24:48] that someone signs with an agency they want to work with them forever? Maybe you know, something something happens. They turn off the agency, or maybe it just wasn't the right timing, whatever. Lots of reasons, right? 99% of people aren't ready to buy. So, this will surface revival targets, and it will craft the drafts as well for the team to get on top of. Um and so, the team can just hit approve or disapprove. And then what's happening right now is um I'm I'm still we're still fixing the piping here, but it's it's going into this um this dashboard

[25:14] that I showed a little earlier, I think. Um but the whole idea here is I'm talking to the sales team. They're going back and forth with every day. They're like asking questions like Look, you can see Sabrina from my team. "Hey, I've been uh I've been deep in deal work this week. Please get me up to speed on high priority activity tasks from this channel and curling work that you've been doing this week. Summarize. Don't give me a novel. Um how impactful do you think um you have been this week?" Okay, so let's see what happens here, and let's see what we what it's asking Arrow, your Open Claw agent,

[25:41] what did she do? Okay. Yep. [snorts] And then sometimes by the way, you you kind of have to like kick it like it's an old car. Um so, so it's like Vic comes and says, "Please act on this." So, here's the week I had at clients. Um here's what I shipped. Here I I did email reformatting for you. I did email quality rewriting. I didn't uh I I did enrichment backfill. Uh dashboard ingest client." I did all this all this stuff, right? And here's the thing. I went to Vic and Sabrina yesterday. I was like, "Hey, you guys are using Arrow quite a bit." Like I I I I I hadn't thought to

[26:07] ask this question. I'm so like I'm like, "God, I'm so stupid. How could I not ask this?" I was like, "So, Vic Vic's like, 'Yeah, this is probably saving me at least 20 hours a week.'" I'm like, "What?" And then with Sabrina, she's like, "This is at least 5 grand a month, if not more in value." And when I went to Vic, I was like, "How much do you

[26:22] think this is worth per month, right?" He's like, "At least 5 grand." And I went to the like someone else on the sales team. They all said 5 grand. I don't know why they all said 5 grand, but it's saving 20 to 40 hours a week. And what it's doing is it looks like some of the busy work, and also some of the hey, pay attention to this work that

[26:39] a sales manager might do. Yeah, exactly. And the thing is here's the thing. This Arrow is connected to HubSpot. It's connected to Gong. It's connected to like it's it's limited on what it's connected to, but the team knows that they can do data pulls from this. Not only that, this will strategize with them whenever they get stuck. Um and not only that, it can execute on things, too. If we needed to execute on on sending emails and things like that, it can cuz it can hook into the Instantly API. So, this can become a full stack salesperson. Meaning right now what it can do is it can say, "Look, this

[27:07] conversation that you had with this potential client, it's uh no one's responded to it in a while. I think you should send a message." The person might say, "What can I email them about?" And it will look at Gong, which is your transcription service for uh sales, and say, "You know, they talked about this issue. It's coming up this week. You might want to reach out to them about Exactly. So, so like with um with this one over here, I remember I I remember this one was like a while back and and crypto company, right? Um I'm like I look at the email, and it's like, "Oh,

[27:32] um you know, we haven't spoken in uh you know, since we last spoke in in 10 months. It looks like uh some things have changed over here, and you know, we last talked about this. Um since then like here's what we've been working on. Does it make sense for us to reconnect?" It's not like, "Hey, come buy our

[27:45] stuff." Does it make sense to reconnect? And some people will say yes. Some people will just ignore you. That's okay. Okay. I I don't think we saw the the dashboard that you just mentioned, the one that's on Vercel. Do you want to click on that? Yeah, I'll come to the dashboard in a second. Let me just um let me do Let me Let me show you how we're using it on the SEO side. Um so, so okay. This is Amy, our SEO director working with it. She's like She's like, "Hey, are you connected to the Leveling Up with Eric Siu YouTube channel? What

[28:10] capabilities do you have on YouTube?" So, this is Oracle, our SEO bot. Um so, she didn't have the the the permissions for it. Um and then so she they kind of start going back and forth, but I I give her the powers later, right? So, I'll I'll I'll talk about how I do that, but you can just see like Look, Amy starts asking, "Is this channel wired for YouTube analysis?" Okay, I'm looking to understand this over here. Here's the calibration. Okay, here's my POV. Are you connected to LinkedIn?" Okay, so anyway, so she's like she ends up getting like I'm like, "Hey, did you get what you needed?" She's like, "I

[28:38] actually did." I'm like, "Okay, well, let me help you with this." Because again, I don't want to just let my team yellow on everything. Um so, then I'm like, "Okay, YouTube API powers. I'm I'm okay get granting that access." I had to go to Alfred, chief of staff, the manager of all these agents, and then grant the YouTube YouTube API powers, and then boom, she's got this. And then

[28:56] she's like, She meaning Oracle, your AI. Okay. Your Open Claw. And then my my hope is like she she's kind of been going back and forth with it. Um but my hope is the rest of this team eats it up, and you start seeing like, you know, 20 to 30 like thread conversations. And the only the other thing, too, is is Andrew, as the bot starts to help people more and more, it it'll help people level up, but it'll also help performance manage, too. It'll help it'll help people ask better questions, but also show you it it'll help you stack rank. Who's doing well, maybe who

[29:23] needs some help. And that's something that, you know, it's it's a machine doing it. So, so there's no, you know, hopefully there's no offense. But on the on the questions people are asking? Uh it can tell based on the questions people are asking how well they're doing? So, my take on this is let's say uh Vic for example, going back to sales for for a moment, or Sabrina for example. Um the types of questions that they're asking, are they even utilizing it um enough from like an AI fluency standpoint. What types of question are they asking? Is it just very surface level? Cuz you and I know,

[29:51] like, a lot of people that use AI, they just want it to they ask very lazy questions, right? It's more so they're not trying to learn. It's more so they're just asking surface level things. So, if that's the case and they're not utilizing the best, well, the AI can see and the AI can also give feedback on that, too. So, that all folds up for us cuz we grade AI fluency from a performance standpoint, that rolls up into the performance management

[30:13] conversations. Okay. Yep. Okay. Now I can move to the dashboard. Um, so let me move my screen. And is this this screen? Good for our audience. Okay, great. I mean, these are real business use cases. We're not just talking about playing with it for fun. Yep. So, okay. This dashboard, I mean, you've talked with the the mission control dude. I think everyone has a mission control now, right? So, this is what our mission control looks like. Um, but my mission control, like, I like my dashboard, not just the mission control, but I'm having

[30:41] it focus on sales and recruiting first. Why? Because sales is very like, by the way, I don't need a lot of permissions to help you generate more leads. I don't need to hook into your your CRM and things like that, right? Um, at least in the beginning. Um, recruiting maybe a little more and then all the other stuff, maybe I need to hook into, you know, some things and then enterprises

[31:00] won't allow us to to easily connect. There's going to be procurement, things like that, security concerns, SOC 2 compliance, all that. So, I'm like, our thing with this is if you look at sale on the sales side, um, Arrow, this is this is fresh right now. We're still connecting everything, but, um, you can see there's a pipeline here, okay? We can actually move people through this Kanban board over here. You have a Kanban board for sales that Arrow is helping to close, but your people are

[31:25] actually doing the closing, right? Yeah. So, the whole idea here, Andrew, is, um, so where where we're kind of, um, we're working through this right now, but we want all of these, um, Slack you you seen the Slack instance already in the sales operation channel. So, every day it gives, uh, five cards and like, here's who you should be reaching out to, um, on the cold email side. Here's who we should be reaching back out to on deals that we lost, deal revivals. Oh, by the way, it's connected to all my LinkedIn connections. So, here are Eric's warm connections. Here's the angle that we should be reaching out to. That, by the

[31:56] way, that workflow booked me a meeting with Google that led me to being invited to speak at a Google conference, okay? So, that's a whole 'nother thing. Um, but we have these different workflows that are running every single day. And then the whole idea with this is that there there are cards that the team can approve. And once they approve like three cards in a row, this more or less becomes autonomous, right? We call them these calibration loops. I'll pause for

[32:17] a second here. say when you say it approves cards, what's an example of a card? This is a card right here. So, let's let's say, for example, I mean, this is this isn't filled out right now, but one. Yeah, it's a fake one, right? So, let's say we hit approve over here. So, like, you know, So, it's saying, go ahead and now I want you to send to what is this to reach out

[32:35] to Acme Corp test. the whole idea is we we actually have been using a recruiting tool that that does this, um, where it you you basically spec out what you're looking for from a recruiting standpoint and then it gives you three profiles to look at. Um, after you approve three in a row, then it starts to autonomously, um, reach out to people for you. It dumps them into a sequence. That's been great for us. And so, we're trying to do that for sales and recruiting at the same time. Okay. Um, because that's valuable for anybody. And like, outbound works for us. It was like, but we want to move

[33:04] to this infrastructure. And it's already been working on the Slack side, but for clients, we know we need to have a dashboard. If you're going to sell this from a business standpoint, you can't just have a Slack bot, in my opinion. I see. Okay. All right, so you want to manage these agents through essentially a Kanban board or some kind of mission control that allows you to see how they're progressing to help you close sales or to help you hire. Correct. And and by the way, I I think this screen is probably the easiest one. I I I think I just made this like yesterday or the day

[33:30] the day before. And it's like, okay, you have 100 total prospects over here. Here's the pipeline value. Here's the average score. Here's how many you have in queue. And you can literally hit on your keyboard like a like a card, right? Like almost like, I don't know, those those like dating swiping things. So, you can hit approve and enroll. I have no idea what this is going to do. Okay, great. This is what happens. So, it's like a video game. You kind of just keep going through it. Um, but the whole idea here is again, once you have three of them approved, it's going to be more or

[33:53] less autonomous. And the thing that I showed you earlier, Andrew, that I think every company should have, um, is if you have an agent doing work autonomously for you, I think you should have like some type of trust meter, okay? So, the whole idea is that if you approve three, okay, good. Maybe this fills up completely, but over time information changes, right? Your product might have changed. Um, maybe you have a different offering or people might have moved on to your company. So, the trust meter decays over time on its own and then you have to build up context again to let it go at fully autonomous. That's how I

[34:23] think things these things should be built from autonomy standpoint. And the trust meter goes down based on time or based on activities? It's based on time and activities, yeah. Got it. So, the longer it goes on its own, the more activities it does on its own, the less trust you have because now you need to go and manage and make sure that it's up to speed. Okay, I'm with

[34:41] you. What's next? Yep. So, I think maybe we can go over to Beat Claude, which will then lead into the, uh, the auto research thing. I think those can be the last things here. So, um, okay. So, um, Anthropic, by the way, this is not an original idea. Anthropic has a Beat Claude challenge, okay? They posted it to X. All I did was I took the GitHub repo. I just dumped it into Open Claude. I was like, hey, I think we need to do something like this. How does this apply to our goals, right? And then you start really going back and forth with it. Okay, ask me

[35:08] clarifying questions. Oh, okay, let's plan this out. Um, those are some prompts that I use. Um, but literally I I like, Andrew, the first person that came through this that Beat Claude, and maybe I should explain what this does, but the first person that that beat it, um, we we made an offer and that person just joined the company. So,

[35:25] Claude? let me explain. Yeah. So, Beat Claude is it's our hiring challenge for AI native professionals. So, we are only looking to hire strategic people who can think better than AI. So, they're welcome to use AI all they want, that's not cheating, okay? But most people, as you know, they use it as a like if you're if you're if you don't have good judgment, you've seen the two-by-two. If you don't have good judgment, you become a slop cannon. We don't want that. Um, so, when people are looking to join Single Grain now, if they're applying, they have to go directly to this page and they have

[35:55] to Beat Claude, okay? They look for a challenge that's relevant to whatever role, um, they're applying for. And basically, time commitment is it's going to take one to two hours. Submissions are capped at, uh, four pages. If I click on this one for like, let's say, marketing strategy, you go in and you read the brief. There's a scoring rubric here. Here's a baseline for how Claude will typically score. So, we're revealing everything to you on how we're going to do this. And

[36:19] we give you a mock situation here, okay? Here's a go-to-market strategy for an AI customer support tool. You're advising a Series A startup that is building an AI-powered customer support tool. Here's the company snapshot. Here's how much they raised, all this stuff, okay? Then it says, here's your task. Okay. Give me a 90-day go-to-market strategy

[36:35] to get them from zero to 50K in MRR. Here's what your strategy should include, channel prioritization, specific campaigns, pricing and 90-day milestones. Here's all the constraints you have. Again, here's how we score over here. Maximum four pages. Now, you might say, "Eric, well, this is very restrictive and this is taking so long. Many people don't want to do that." And we say, "Good." This has eliminated so many people that we like this pisses off so many people. "Well, you know, if you're going to do this, um, you know, you should you should pay." And I'm like, "Fine. Like that's what you think, that's okay. Like the like if you want

[37:04] to apply, this is what you have to do." And here's the thing. It has given us the most amazing like I just saw this this morning. Um, this guy, he literally made a Notion document saying how interested he is to to work for Single Grain. And he went through the Beat Claude challenge. He made a Notion document. He went above and beyond. He showed me all the examples that he's working on. I'm like, this is someone we want to take a deep look at. So, this has been a an incredible filter for us for hiring AI native people. I see. And what you've got there is a test. And

[37:32] when you say Beat Claude, it's is Claude also doing the same thing? And then you're ver- I see. And then you're comparing the two results. What does Claude natively create as an answer to this? And what does this person we're thinking of hiring, who is better? Dude, let me show you. Okay. So, this guy, okay, he Beat Claude. So, you can see this is our this is our internal dashboard over here. So, we have an API

[37:52] gateway. This lives in our API gateway. So, Beat Claude dashboard. So, you can see here, this guy scored an 86, um, versus Claude scored a 77. So, verdict is he beats Claude. And it shows, um, you know, his evaluation up here, like what he answered and how Claude answered down here. Um, and then if I scroll down, like, we we also have a dinner test candidate. Does he pass the dinner test? Um, you know, here's the red flags. Let's look at the top three improvements for lowering score submission. Um, I don't even know what that means. But you can see on the left

[38:19] side over here, here's his submission. And then on the right side, here's the head of talent from Claude's side. Um, and he clearly if you if you go through it, it's like he clearly has thought through it and he's added a human element to, you know, the the AI system that he used. Yeah, I see. Yeah. Okay. All right, so now you're getting to see what he's like, how is how does he compare to the best that AI

[38:41] can offer us right now? Exactly. Okay. What's next? Okay. So, let me set the the context for this cuz, um, for for viewers here. So, this X article, I would recommend that you read Karpathy's, uh, auto research, um, his, uh, GitHub repo. You can read his post on it. But when he released auto research, which is basically how you can compound your coding experiments faster, um, overnight, how it runs like five experiments or, sorry, an experiment every five minutes or so, um, all these people started to build it

[39:10] from like a coding standpoint, right? The moment I saw that, I was like, "Whoa, this instantly applies for really anything where you're running a lot of experimentations or or iterations." So, I'm like, this could apply to sales. This could apply to marketing, like we're talking. Yeah. Um, you know, anything, right? But you do need heavy volume for this, which I didn't realize until actually I started to do it. Um, so, my thesis here is like, if you're a good marketing team, maybe you're running like one experiment a week, maybe 52 in in in in a year that your entire marketing team is running as a whole. Sure, you're like, individual

[39:39] teams are running more stuff, but I'm just saying, you can probably be running a lot more experiments um, because if you're talking about running 100 experiments a day or so, you do the math over time. Okay, that's over 36,000 in in in a year, probably even more than that, right? And so, if you have a system that you build out kind of built off of the the auto research engine, but you use it for marketing and sales, that means you should be able to move a lot faster. And really business to me is is about compounding. It's it's all about

[40:05] iterations. And if you can speed up the iterations, then you should be able to grow a lot faster as a company. So, that was my thesis. I think that made a lot more sense frankly in marketing than Well, I guess for coding it does too, but the idea that you're sending out so many offers by by email going back to your cold email example. You want to keep testing and iterating and you can only do so much as a human being, but if you have it in a systematized way, you can do many more. Okay, but you do not yet have Open Court set up to do this, but that's

[40:34] the plan to eventually automate it. So, Open Court is running on it right now. So, actually if I go back to the dashboard over here, there's there's this auto growth section, okay? So, in this auto growth section, which will will will let it load. So, you can see what's happening over here. So, these are all the email campaigns that we're running, but there's not enough volume right now. You can see the 469 cent. I

[40:53] need to jack up the volume to way more. So, I think we're going to try to jack it up to like 1,000 emails a day or something like that and then hopefully it can it can get through this quickly and then we can see what happens. So, like it's it's saying it's going to

[41:04] This one has a countdown. By the way, I just built this like 2 days ago. This is scoring in 10 hours over here. And then you can see what's winning, what isn't winning. And then it's going to it's going to start to test. And so, in my mind I'm like, I'm not going to do this cold email BS if I'm not going to have it also test auto growth. So, I'm trying to knock out two in one here. I see. Cold email and at the same time use

[41:23] this process for iterative testing. Yeah. And I I think you have to either be sending high volumes of emails or you have to be spending a lot in ad spend to test kind of your creatives or you know, your different offers and things like that. I think that's required to make auto research or auto growth work. All right, you're spending hours a day on this. It's super fun. It's super exciting to see it. Overall, what would you say it's adding to your business

[41:46] bottom line? So, let's start with the what the sales team said yesterday, okay? So, they are they've revived around I want to say around $2 million in pipeline, okay? So, that is significant. And then you just kind of back that into like a conversion rate there. Like, you know, that's at least a couple a couple hundred thousand

[42:05] dollars there. That meeting with Google um we're talking about being invited to speak at a Google conference and not only that, me getting very helpful intel. I'll just leave it at that. Plus the pipeline it's driving from the X articles or the the dot charts it's it's putting out right now. And we're talking these are not even like they're they're nine-figure companies, right? And usually our our

[42:26] ICP is nine-figure, 10-figure or so. And then the conversations it's leading to. Like we if if I didn't see you creating stuff like I wouldn't even know what you're doing right now. You're seeing what I'm doing. Like like everything kind of is coming together like that. And then when I go to the sales team, 20 to 40 hours saved per week and you're saying each what like each agent adds $5,000 in value. And now, by the way, the hackathon, I I sent a message this morning. I was like, guys, like don't make don't let's stop hiring humans to do robot things, okay? So, I want you

[42:54] guys to think about how you can make everything agentic and I want you to work with these agents to make solutions for you so you don't have to do this stuff. And I was kind of mocking like Monday. I was like, guys, nobody wants to freaking do keyword research. You want to spin up ads. You want to Nobody wants to do that right? I was like, guys, free yourselves, right? So, I I kind of have to play like, you know, this Messiah right now, but I also have to push people. But anyway, all that to

[43:16] say is that's kind of what we're seeing. This is we're talking real pipeline. We're talking about real time saved. We're talking about real revenue generated. So, yeah, that's what I would say. It's exciting. We'll leave it there. Thank you, Eric. Thank you, everyone. Bye.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt