Leveraging AI

284 | How To Use Claude Code To Get More Leads (detailed playbook) with David Yu

• Isar Meitis, David Yu • Season 1 • Episode 284

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 42:14

SECURE YOUR SPOT FOR THE 
MULTI-AGENT ORCHESTRATION AI COURSE: https://multiplai.ai/multi-agent-orchestration-course/
AI BUSINESS TRANSFORMATION COURSE: https://multiplai.ai/ai-course/ 

What if your cold emails could feel like they were written just for each prospect—because they actually were?

The reality is most outreach fails not because of lack of effort, but lack of relevance.

In this episode, you’ll discover how AI can transform your entire lead generation process, from identifying ideal prospects to crafting deeply personalized emails that speak directly to their pain points.

Instead of blasting generic templates, you’ll learn how to build a scalable system that researches, qualifies, and communicates with precision—helping you reach the right people with the right message, at the right time.

In this session, you’ll discover:

  •  How to define and identify high-quality leads (not just more leads) 
  •  The step-by-step AI workflow for qualifying and enriching prospects 
  •  How to automatically research companies and extract real pain points 
  •  Why personalized cold emails outperform templates—and how to scale them 
  •  The exact system behind achieving 20–30%+ reply rates 
  •  How to structure effective cold email sequences that get responses 
  •  Tools and platforms to automate outreach without hurting deliverability 
  •  How to continuously improve your outreach with AI-driven feedback loops 

David Yu is the Founder & CEO of Futureman Labs, where he builds AI-powered automation systems that help businesses scale lead generation and outreach with precision.

He specializes in creating end-to-end AI workflows, from data sourcing and enrichment to personalized communication—helping companies turn cold leads into real opportunities.

👉 Connect with David on LinkedIn: https://www.linkedin.com/in/davidyu-ai/ 

About Leveraging AI

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Isar Meitis

Hello and welcome to the Leveraging AI Podcast, a podcast that shares practical, ethical ways to leverage AI to improve efficiency, grow your business, and advance your career. This is Isar Mateis, your host, and we have an awesome episode today. if you've been listening to the show for a long time, you know that I say that a business can. Live without marketing. A business can live without operations. A business can work without hr. A business cannot work without clients because then you don't have a business. And in order to have clients, you need leads. In order to have leads, you need to be able to do a lot of tedious things to actually do this correctly and to be more specific. You don't just want leads, you want the right leads. It's actually relatively easy to get leads, but most of them are gonna be junk, which means you're gonna spend a lot of time and a lot of money and a lot of resources not getting new clients. And the goal, again, is getting clients. And to get the right leads, you need to know, first of all who they are. You need to be able to define your exact ideal client persona or ICP. You need to be able to know what their pain points are. You need to be able to research potential leads and know if they are actually aligning with your ICP. You need to understand how they talk, what they say, what their needs are, what are their pain points, what's gonna provide value to them. All these things take time and effort. And it used to be lots of human labor. That's how we did everything. And now you can actually build a system that. Users, AI to do all the research and all the qualification of leads for you, and even help you create personalized communication, at least in the first step to approach these people. And so now you're getting the right leads. You can do it at skill because it doesn't matter how many people you run through the system, you can run 10 or 10,000. It's gonna take you the same exact amount of time, zero. that is a very, very attractive thing for any business because again, that's the bloodline of any business. Our guest today, David Yu, has been building automation systems for clients in the last two years or so, and he's an expert on building exactly these kind of things, and he's gonna show us today exactly that, how he has used or how he's using Claude code in order to build a funnel that goes through a list of potential random leads and finds the Niles in the haystack, the, uh, diamonds in the rough, call it whatever, whatever, phrase you want to use, but finding the ones that have the highest likelihood of actually turning into clients and how to approach them at skill. he is the founder and CEO of Futureman Labs, which is a company that does exactly that, provides AI automation services for other businesses. So he's not just talking about this, he's actually doing this for a living, so he knows what he's doing. I. No, this is gonna be extremely interesting for all of you. He is really, really good at explaining what he's doing. He has amazing processes and systems, and so I'm really excited to welcome David to the show. David, welcome to leveraging ai.

David Yu

Thanks a lot. Iar. I'm really excited about this topic too,'cause I'm being, getting really good results with the call email. Uh, so, so yeah, I'm, I'm we'll jump into it.

Isar Meitis

Yeah. I, I think that's, you know, I, I, I totally. Forgot to mention the fact that it's a cold email system, which I think makes it even more interesting. I think most businesses, uh, especially small businesses, are afraid of cold email outreach. Uh, because you don't really know how, and it's kind of like this mystic thing. Like you, there's, there's the people who know how to do call outreach and the others won't even try. And once you build a machine that can do this at skill and do it effectively, it's a whole new universe that opens up to you of being able to get more clients.

David Yu

For sure, uh, because I, I used to do it, like the old fashioned way where people copy templates and all that stuff, and I got zero replies and I thought, call me email didn't work. And I think we're really lucky in this internet section where AI can do a lot of heavy, heavy lifting for us. Um, so, so yeah, I'm just gonna show some results first, uh, and, uh, and we'll jump into it. so this is instantly a core email software that I'm using. Um, so the past two, campaign that I ran on there actually had 20% reply rate and then 31% reply rate. and that was really surprising to me because I, I thought the market reply rate was around, uh, two to 5%. so, so yeah, I, I'm not an whole email expert. I actually just started studying it and actually try to understand how I can use AI to. do this effectively. And, I'm just gonna show you the system that I use.

Isar Meitis

Yeah, that's fantastic. And again, I do think these are crazy high because if you think about, I think newsletters gets open around 30 to 40%. Right? And these are people who actively signed up to a topic they know they want to consume. and I'm, I'm talking about open rates, not even response rates. And so those numbers are really, really high.

David Yu

Yeah, I, it, it is crazy for me too. so yeah, I made this little presentation for, with Gemini, um, and here it goes. Um, so we're gonna talk about the different phases, and there are different phases to, Doing a successful co email campaign. First thing to gather the data, right? And I'm just gonna go through this, uh, quickly and maybe some of you already know the sources, to get leads, uh, but in case you don't, uh, there are, uh, things like Apollo and air skills. These are the two that I use most, more often. And based on the niche, you can, go on Gini or Google, where whoever you like, just search the industry and then, uh, lead source. And then you would come up with the different, sources that you can find. and for my niche, I, I tried to target e-commerce and, uh, the sources I, I tried with is the store census and built with. Those are two platforms that you can, uh, buy leads, uh, and you can pay for the subscription. And if you're don't want to keep on subscribing to it and you can just cancel the subscription after you export your leads, you can always save money that way. Um, and there, there are all, also another way where you can try to scrape information or use a, third party scraper to get you the information. another thing is called, uh, appify. Appify is, uh, aggregation of, bunch of APIs where you can get data programmatically. And that works perfectly with, uh, things like Claude code and AI agents just because they understand API. and then you just give it the access. so if you're targeting like brick and mortar stores, uh, this will be perfect because you can. Store, scrape Google Maps, um, and then you can find like local businesses and offer them, uh, whatever your services is. and lastly, which is very specific to Claude code is because Claude code initially was made for just coders. Uh, but I think it goes way beyond, code now because any marketer can use Claude code now. And, um, you just tell it what you want to scrape and then it will tell you if it's possible or not possible and you can actually run the scraper for you. so underneath it will write code and then execute the scraper for you. So

Isar Meitis

yeah, I, I, I'll say something that connects two of these dots, in most cases. So I already have, and it's actually really cool to see. I have several different, similar data sources. Already connected to Claude because of previous things that I've done, right? Yeah. So I'm connected to Appify, I'm connected to a bunch of other tools like this, even data for SEO, like different stuff like that. and every now and then I, I ask Claude to do something and it just goes to the right. API. So once I don't ask it, I don't tell it, but it's starts like, oh, I know this kind of data exists through this tool that you gave me four weeks ago. Yeah. So I'm gonna go there. I'm gonna check if the data exists and if it exists, I'm gonna run the API and do stuff like that. So the cool thing about the Claude ecosystem and any, it, it doesn't have to be Claude, but both, uh, David and I use it a lot. It's just really, really good at understanding what are the tools that it has at its fingertips. And he knows how to, and sometimes like, why are you going to this API like, oh, they have this thing and then they can do that. I'm like, okay, go. and, and you know, I, I know it's gonna cost me a few cents to test it out, but I don't care. and so it's beyond the fact. The one time thing, like once you connect these tools to your favorite AI coding platform, uh, if it's an agent environment like Claude Code, it knows how to go back to these tools because now it has access to it and it can help you or suggest it in the future when you're doing other stuff.

David Yu

yeah, for sure. I think the most powerful thing is it has access to internet, right? So you can access, access all the documentations, it all knows how to implement it, and then it can simplify it and then tell you exactly what it does, and, um, you just tell it to do like where whatever you want to do, pretty much.

GMT20260406-163401_Recording_avo_1280x720

Sold is brought to you by training courses by Multiply, which is my company. We currently have registrations open for two different courses. The first one is the AI Business Transformation course. It is a course we have been teaching since April of 2023. So for three years now, at least once a month, thousands of business people have learned AI fundamentals through this course. So if you are looking to build a solid foundation and learn multiple business use cases as well as tools and exactly how to apply them, including data analysis, writing proposals, creating videos, images, and so on, basically solid fundamentals across the board for AI usage in the business world. This is the right course for you. I am the one teaching this course over Zoom. So it's actually me, not an AI avatar. And it's not a recorded session. It is me. You can ask questions, you can interact with other people like you. And the coming cohort starts on April 20th, and it goes for four weeks in a row. Every Monday, two hours every single time. And in the middle of the week you can come and join us for our AI hangouts and ask questions in between the sessions as well. It is probably the best course out there to give you solid AI fundamentals if you want to learn how to apply AI in your business. The other course is our Multi-agent orchestration course. This is a more advanced course that teaches how to build, like the name suggest multi-agent orchestration solutions using cloud, cowork and cloud code integrated with other tools integrated with your entire ecosystem and pick stack. And so if you are a little more advanced and you wanna start your way into the agent era and be able to automate literally any digital work in your business, build entire teams of employees, come and join that course. We have launched the early bird registration last week, and it is almost filled up to early bird sessions when I was planning to do just one. But the demand was absolutely crazy. So you might make it into the early bird session, but if not, we are opening another regular cohort just the following month. You can find out more information and register for any of these courses, either one or the other, or both if you want to, because you can start with the basic course in April and then join the other course later and you can find all of that information in, links in the show notes that will take you to the specific courses. And now back to the episode.

David Yu

so yeah, like now we're going to jump into the, the, the terminal. Uh, but, uh, fear not, it is really just really simple, uh, like step by step. Um, and, it's layout here. The six different steps, uh, we have to set up, and we verify the email. because sometimes emails, they, they might not be a hundred percent valid, and if you send an email that's not valid, uh, they can hurt your, uh, email, domain reputation, and it will increase your bounce rate. and then the third step is the most powerful thing you can do with Claude Co or any AI agent is because you can do deep research. Uh, so you can have the leads, for example, e-commerce store. Usually you will get the store, uh, URL as well, so you can research into the store information, and then you would do, a review scraping, uh, in, in this case, I, I do, uh, trustpilot review scraping, and then it gets all the reviews, especially the negative reviews, and then you summarize the pain points of that business. Uh, so when you, uh, go into personalizing the email and proposing, uh, offer, um, you can. Tailor it to exactly what they're struggling with. And then they, they, I actually get a few, uh, thank you emails just because they, they, they're struggling with that problem. They got. Thank you, David. Like, I, I was actually thinking about this and I'm glad I got your email. And so I, I think that's a game changer because that you are starting form their POV of, uh, I, I have this problem. and not just because you're trying to sell something, but actually trying to solve their problem. uh, after, um, deep research personalization, then we import to instantly, uh, you can choose any other call emailing platform, but I choose instantly just because they have a API access, which goes well with, uh, Claude code. It can just, uh, import that data directly into the, the platform. Then you build the e email sequence. uh, you can have, usually at least a two step email sequence, but you can do like three or four more if you want. but usually a two, two step email sequence is good enough. Um, so I'm just gonna show you the terminal real quick. I ran a test. I, I created a new project and imported a few leads. Um, so I'm assuming at this stage you already have a lead. Uh, and then usually that lead list is in like CSV file. And then just tell Claude code where it is and then tell it to, just to run the process and generate the code email. Um, so in the end I will, I have, I'll have a link where you can download all the. The code email Claude skill. So you can actually import the Claude skill into your own Claude code. Uh, so you don't have to worry about setting up the, the, the Claude MD file or the, or the setup of the file folders and everything. Uh, you can actually just, download the Claude, uh, Claude code skill, and then just import it into your Claude code. And then, when you ask this question, help me run a process to generate code email. You will just pick up that skill. You can see that it, it loads the skill and then it runs through the different steps. And then these are the, the six phases that I mentioned. and then it just runs it for you. and, uh, each steps, uh, if it needs more context, he will ask you, uh, for example, if he doesn't know, like what's your offer? Or what's, like if he sees a problem with the, with the email source, maybe he thinks like the email source is not good enough. It's not personal enough, it will tell you, oh, maybe you should find like a different source or, um, find a different point to, to mention. Um, and here I, I kind of just quickly talk about what I do, for my business, and then it continues and it builds through step by step, into phase one. And you can see that he, he writes a bunch of code, uh, but actually, usually I just, tap on Yes, yes, yes. Because usually that, that they just get it right the first time.

Isar Meitis

So I, I, I wanna pause it just for one second. Yeah. Uh, so a few things kinda like for those of you who are not using Claude code or are afraid of the concept that it's writing code and so on. So there's, there's three main ways you can use Claude code. All of them are relatively simple. Some of them are a little more geeky, but they're all still simple. So the easiest way is you install Claude desktop on either a PC or uh, Mac, and there's three tabs on top. There's chat, there's cowork, and there's code. That's probably from a user friendliness, probably most user-friendly way to use this, uh, because you don't need to know anything. You don't need to install anything. You don't need to do anything. You just have Claude on your computer and you go to the code tab and you can use Claude code. and so that's option number one. Option number two is you can run Claude code inside a terminal. What is a terminal? It's like the. Most basic way you can communicate with systems, inside your computer, and you can open it in the same way inside of Mac or a pc. Uh, you just go to your little search thingy and you type terminal and it's gonna pop it up. and then all you have to do is install Claude code, which is it's one command and it's going to install it. and that's it. And from that moment on, you just type Claude and it runs clo for you and you just tell it what to do. Uh, the third option is to run it inside a development environment. Something like, cursor or rep it or replicate or, uh, whichever, it doesn't matter. and it just runs a terminal inside that development environment. Again, you don't need to be a developer to do any of these things. Uh, the, the benefit of running it inside an environment is that it has. More of a human style user interface when there's buttons and menus and windows and, and you can see all your files and stuff like that. Where in the regular computer terminal, uh, you don't have all of that. So that's the benefit of running it, uh, in, in, in that kind of environment. As far as the skill in. The AI universe now, it's actually a claw invented it. anthropic invented it, but now it's available everywhere. You can create a prepackaged set of instructions and reference files and so on, that these tools know how to then reuse as needed. And so when David is saying, oh, I'm just telling you to use the skill, is because he took the time to explain to Claude what he wants to do, what are the steps, how to do the research, uh, where to find the information, all these kind of things. And then he packaged it as a skill. So now Claude, when you tell you to go and do the thing, say, oh, I have a skill and I know how to do this, and then it's gonna, but the skill is something very, very simple. when David is saying. MD files is a markdown file. A markdown file is basically text like readable English. and the difference between that and like a Word document is that it doesn't have the fancy formatting, it's just text. And because it is just text, it's a much, much lighter file and hence it's a lot more efficient to use. And that's why all these tools, wherever you go, everything today is do.md files, which are basically simple text. But what's in the text is a set of instructions on how to do the entire process that David's gonna show us. and, and Claude just knows how to use it. So it reads the file, says, oh, I know how to do this now. I know kung fu from, uh, the metrics, right? That he suddenly knows Kung fu. he then Claude suddenly knows how to do this process. So now let's go back, uh, to your story and I'll, I'll bring everybody back. David uploaded a CSV file of a list and he told Claude, okay, go and work on it, and it knows how to do all the steps. So now we're in step one.

David Yu

Yep. Exactly. Yeah. Yeah. Thanks a lot. Uh, yeah, I, I think I went a little bit too fast in terms of, uh, uh,

Isar Meitis

you, you know this, so, so you don't think yeah, yeah. Perfect.

David Yu

Yeah, it's, yeah. It's so good. so yeah. Um, it, it run writes a lot of code, but you actually don't need to learn how to write code just because that it will write it for you. and after the step one phase is complete, we go into, uh, the phase two. Where is, so what's

Isar Meitis

the output of step one?

David Yu

So the output of, uh, step one is, it, it just checking, it's just collecting all the information it needs and then kind of set up setting out the project. Yeah. So it would ask you, uh, what's your offer? what's your company? Usually you can tell it a little bit more, you know, like maybe like what's your tone? Like what's your voice of your brand, all that stuff. And then you can, you can work even better if you give it more context like that. and so yeah, then we move on to phase two where it is the email verification. so like mentioned if you don't have a verification step for your email, then your email get bounced and then now her, your email reputation. Um, so in this case I use a API called Milliam Verifier. Um, you can use any other API that you want, but this API. I, I find it to be very cost efficient. Uh, you pay for the amount that you use. Um, so you just, what you do is that you, before, um, it gets to phase two, you want to go into, the file, let's say in the file that it creates. Um, so you will have a, in example, that just means the, uh, the variable that it needs to run the APIs. so a very important note to, to, take here is that you never want to upload, uh, any of the credentials online like GitHub. Um, so whenever you put in, so the notices is not, not real value here, but like when you, whenever you create a new, actual.in file, you just want to make sure that those files are not uploaded onto GitHub, uh, or any, anywhere on the internet. uh, you just want to create,

Isar Meitis

so again, not, not to get too geeky on this, but it's a, it is a very important point. You are using GitHub, which is what most people are using to upload their code or their processes into a, a. Reusable environment. So if your computer craps tomorrow, you still have everything you built so it's good best practices. You do not want to upload your API keys, your passwords, and so on. There is a hidden file on your computer that tells GitHub what not to upload, but you need to verify that. And the way to verify that is to tell Claude code because you don't need to know about that file, and you don't need to know how it's written, even though it's simple English. you can tell it. I would like to verify that when I sync this folder into GitHub, it will not use my passwords, my usernames, my uh, API keys and so on. And then we'll go and do this and it will verify it for you and then. Go and tell it to sync your whatever folder, the GitHub. So it's really good point from David.

David Yu

Yeah, it's, it's very important. Um, yeah, so, uh, after you have the dialing file, you should have the, instantly API key and also the mailing verifier API key. Um, but at phase one, you actually only need the avail mail and verify API key just for the email verification. Uh, but in the, in the later phase, you would need the instantly API key for the instantly, um, campaign creation, all that stuff. Um, so let's go back. so in here, it goes through this process of, verifying emails. You actually look at the different emails and see if there's any like, broken emails and you try to fix it for you. and. You notice, like, oh, there's a, there's like a certain number of, uh, personal emails and there's like company emails and things like that. you kinda categorize them for you. you, you can take that data however you want it, uh, but usually, what you can do is like if you spot a pattern or, or just like a high array of, bad emails, then you should just reconsider where you get the leads. Um, and then probably just, uh, not use it Spanish. yeah. And, but,

Isar Meitis

but again, here, what, what what, what it's showing right now is, again, step one will just set up, like it set everything up to be able to run Step two, it verifies the emails and it found 12 valid emails across six different companies. It tells you what the companies are. So again, you are not doing anything. It's going behind the scenes, doing a third, using a third party. Tool that knows how to verify that your emails are actually real emails, which is going to keep your email score high, which means you're not gonna get punished later on when you try to send additional emails. So this is a very important step, but again, you don't have to do anything. It just goes and verifies it for you.

David Yu

Yeah, exactly. Exactly.

Isar Meitis

Does it add columns to the CSV with like the correct, uh, information? Like what's the actual output?

David Yu

s you mean like from this step? Um,

Isar Meitis

yeah, yeah, yeah. So, so you have you had a CSV file with, with emails,

David Yu

yeah.

Isar Meitis

Does it add like, uh, the correct email or the, a good or bad signal? Like what's the actual practical output?

David Yu

Yeah, so, so the approach I'm taking is that always keep the original file. And I'll tell you to generate a new file of the, the filter down, email. So like you can always go back and verify if AI messed it up somewhere, deletes things that you got it.

Isar Meitis

So, so you get a brand new CSV with just the approved leads.

David Yu

Yeah, exactly.

Isar Meitis

Got

David Yu

it. Okay. And because at this step we're not enriching any, uh, information, we're just making sure the email part is a hundred percent correct. Yeah. um, and then that leads us to the lead enrichment part where it does the, well first it does the categorization. well, for, for e-commerce, uh, there are different categories in e-commerce, right? And people sell different things. Uh, so it's, it, in e-commerce specifically, it helps with categorization, to put the different email groups into category. Maybe it's like, baby products or, uh, pets product you put in, in the, in the same category. so in the end, like when you're trying to, uh, research, tell or tell Claude code research, you would have a better direction where to go. Yeah. so after that, I, while I, you would have, information like the know, the revenue, like if they can get the revenue data, they will have the ve. And sometimes, you know, depending if they have like review data or not, like, sometimes it's just not enough review data. so I only have a really small, sample data here. but usually you will come up, come back with like a lot of, uh, more review data where you can use it for personalization.

Isar Meitis

quick question about the instructions for this. Yeah. Did you tell it which websites to go and look for the information or you just told it to do research and It does its own thing?

David Yu

So I, I told it to go on Trustpilot first, and then, you know, do another, just general, uh, web search in terms of like if there's any reviews on. on this company. And if, if they don't come up with anything, then it just doesn't come up with anything.

Isar Meitis

Got it. So, so, so it's a mix of both. Like you, you gave it a specific website but also told you to do its own research in trying to find information. Cool.

David Yu

yeah, exactly. yeah, and after you have all that information, you should have a brand new CSU file with all that information, and then it goes into personalization and follow up copies. so that's usually the, the step one and step two email. Uh, so, the co email this, the, I think the first email is the most important because if you, uh, don't get that right, then nobody cares. And, so yeah, like, so you want to have in the first line something that, Sounds like you, uh, but at the same time, it is very specific to, to their problem. let's say like, uh, for example, I, I scrape, uh, the e-commerce leads and, one of the common problem was, uh, delay delivery. Uh, a lot of customer, they'll complain about delayed deliveries. and, uh, one way, I can, personalize this is by saying that I can provide automation for, uh, inventory tracking and warehouse management. Um, and I just make it very specific to their, their brands. And, so yeah, and you can just go back and forth with Claude code at this step. if you find that the, the initial generation of personalization is not what you want, you can just tell it to change. And, uh, same thing with follow up one, follow up two. so in this whole process, like when I, first started out, I have no idea how to do co email, right? So I, I, what I did was actually. Grab a bunch of, uh, YouTube, uh, video about co email and then dump it into Notebook lm and then, give, extract the, the best tips from those information and then give it to Claude code and tell you to, you know, try to practice those best practices in the cold email. so, so that's why it has like some kind of base understanding of like what to do in cold email. but it, it, it's always changing like day by day. So like you kind of want to, uh, put in your own flavor at this stage. Yeah.

Isar Meitis

Yeah. One thing about that, something that I'm doing in, in most of my processes that are like this, you can build a built in verification step that AI will check itself. So you create a separate skill that is, will critique the emails. And if, if the email comes out, it's gonna read every single email that it generates every, called email that it generates. It said, oh, this is great. and it can come up with its own ideas on how to grade it, right? So it's gonna grade it for tone. It's gonna grade it for how salesy it is. It's gonna grade it for how much it touches the pain point. It's gonna grade it for all these things that again, it's going to learn on its own based on you giving it best practices from other people. yeah. But you can add a step that will grade the emails that it's generating and if it fails to pass to get to a passing grade, it will send it back to the previous step to rewrite the email with specific feedback. And so it is very, very easy to build a mechanism. Now the other thing that you can build to make it even more interesting is to have another file side file that. Gives it comments of what not to do that lower the grade and have the step that writes the emails, also read that. So over time it doesn't repeat the mistakes that it made. Mm-hmm. And every time it makes a mistake, the reviewer writes a comment on that other file that now becomes a ever growing evergreen, best practices kind of like document that that gets your output better and better. And if you really want to take it to the next step, you then connect it to the results in the end. So then you get a full verification of this email actually got a better score than this email, but it actually underperformed, which means my scoring system is not good. So it's gonna fix itself, uh, over time based on the actual results. It sounds really complicated. It's actually really easy to do. All you gotta do is explain in simple English to Claude code what you want to do and it will build this for you. And so these are the things where, where it gets you build a system that improves over time without knowing anything about machine learning or about any one of these things. It's literally a file with best practices, with here's what worked because these emails that had this thing had a better open rate and click-through rate and respond rate, and here's what didn't work. And so don't do this, but do more of that. and it's really that simple.

David Yu

Yeah, exactly. It is like a self-learning machine that you're training. That's pretty cool. Yeah. so yeah, I think, uh, we, we went through the, personalization and follow up copy. Um, and let's see if I go down.

Isar Meitis

So, so quick question, the follow-up copy Is created at the same time as the first email.

David Yu

yes,

Isar Meitis

that's correct. So, so it doesn't wait for the response and then generate the follow-up copy. It actually generates the follow-up copy. So is it a follow-up that only gets sent if they did not respond?

David Yu

Yes. The, so that's the call email sequence, like got, whenever, got, there's a reply, we stop the sequence, but if there's no Got it. Got, yeah. Then like, you know, we, we can, you know, wait three days or wait whatever days and then we have a follow up sequence.

Isar Meitis

Got it.

David Yu

Yeah. Yeah. So, yeah. And. it, it shows you the copy, the, um, and then you can kind of like see what it does in terms of like, what, the different segments. Oh, so this are coming from the vertical data that we have in the previous steps. so there was, in this sample data, there is home and garden, there's electronics, there's fashion and returns. so based on those different segments, it has different hook, different offers. but yeah, you can just tell it, you know, if you want to change, and, and before you want to go into instantly and because after you import it into it instantly, uh, it is, it is just a lot, a little bit harder to change. You can still tell, tell Claude code to change it, but, and you have to work with the A PN and all that stuff. Um, so like before. Phase five, you wanna make sure that you're happy with the copy. And then, then we go to phase five, it creates a campaign. and then, uh, uploads all the leads into instantly. I can show you in the dashboard what that looks like. Um, uh, I created this test, campaign. So in here you'll see that, there are the leads that are, that were loaded, and then there's the sequence that we talked about. Uh, so you'll see that the, it only has this thing called personalization, but you can click on preview or you was just passing the personalization that's generated, uh, by AI because it's a variable. It can be whatever you want. so it, you can see for the different leads what the copy will be. Just by clicking on it. Uh, so you see like the d different leads will have like different, personalization email.

Isar Meitis

yeah. And, and what you can see again, you can see, but it's not just the name, first name, last name. It's a completely different email because it's written from scratch based on what it knows about the company and the person and the pain points and what they've learned from their reviews. Like it's really a well written email, probably better than I can write on my own because it really knows how to find the nuances. On what are they struggling with right now based on the customer reviews that they're having. and so it's a, it's a completely different email every single time. They're not the same length. It's not, they're talking about the same, they're talking about different things. They're approaching it from different angles. So it's really, really personalized. It's not old school templates change the first name change, company name, uh, change, uh, problem. It's really written from scratch, which is very powerful.

David Yu

Yeah, for sure. Yeah, because I've been, like, one of the first internship I did in college was like trying to do the template, whole email, and then that didn't turn out well. Just'cause like no, I just get yelled at all the time by the replies. Yeah, I'm, I'm glad we don't live in that time anymore.

Isar Meitis

Me

David Yu

too. Um. So, yeah, this is the sequence and then the leads. It's all uploaded from, from Claude code. And then what you do here, it really just reviews and you wanna make sure that the emails are, you know, imported correctly and then the steps, uh, so you see that, you actually have zero, in the first step and the second step, and I think that's something that I noticed it, it did quite often, is that it doesn't actually put in the number for you. Uh, so you just wanna make sure, like the steps, you actually have some kind of breathing room. You don't wanna send two emails at the same time.

Isar Meitis

Yeah.

David Yu

Um, yeah. And then lastly, you just want to. Hook up some email accounts and then, schedule the, the launch day, or you can just launch it right away however you want. Yeah.

Isar Meitis

So again, when you say hook your email accounts, just to explain to people what that means.

David Yu

Um, so, in instantly, there's two ways to do this. Um, so you can hook up your own, Google e email account. Um, they have the integration with the Google workspace and that stuff. You just kind of walk through the step. Uh, it's kind of like you log in with Google, you can see everywhere. and then there's a second step where you can buy Prewarm email accounts. Uh, so what that means is that the email has been, sending emails to each other so that it's less likely go to go into spam when they sends, uh, a new email to. To a new, uh, recipient. Yeah. Uh, so yeah, so that's the method I usually go with just because it is, it saves a lot of time because if you, uh, set up with your own email account, you have to do this, warmup process where you have to send email, you know, maybe to your secondary email account or something back and forth a little bit. And so you can get that email account warm. Yeah.

Isar Meitis

Yeah. So what it basically does, one of the big, big red flags for spam filters is how long has this email been alive and actually emailing and getting emails. And so what these tools are doing, they're creating a gazillion fake emails that then email one each other one another on an automated cycle. So then they're okay in the. Universe of spamming emails because they've been a live email for a long time. And now you can just buy that email address and send emails from that email address. And then in the backend, when somebody responds to that, it can still flow to your own email, so you don't have to monitor 20 different email addresses when you're running campaigns. So that's how the process works. And, and it's roughly the same process in more or less any, called email tool.

David Yu

Yeah, for sure. And then I actually just remember, uh, it is not in part of this process, but I remember, make sure you, you want to have Claude code to check the different countries of your leads. Uh, because different countries have different anti-spam laws. Mm-hmm. And, uh, uh, one of my clients, they're like, oh, I, I really, I noticed that, you know, Australia has this anti-spam law that's more strict than others. Um, so you want to, uh, just kind of tell Claude code to scan through the countries and then say, oh, like look up the anti-spam law or any laws, regarding to call email, and then just if they're really strict, just take them out. It is not worth the risk because, um, yeah, if, if you run into on any of those legal issues, you won't be able to recover from those. yeah. Yeah. So

Isar Meitis

very good point. Uh, one, one thing on my end, again, just a small tip, when it comes to the, the reviews in the middle, so as the way David is doing it is the reviews. Are happening inside of Claude code, meaning, it just write, it shows you the, the text that it's gonna put out there. you read it, said, yes, this makes sense, and you literally just type back, yeah, this looks good, or go ahead and change this, or whatever you wanna do. Uh, just inside the chat, inside of Claude code, the way I do it usually is I created the Claude code integration with Clickup. Uh, you can do the same thing with Monday with Asana, with, uh, notion like, whichever, whichever tool you're using to manage tasks. And the cool thing over there is that it creates a command board or a list with different statuses. So when it gets to the point that I need to review things, it open tasks for me, just like my team would and says, oh, please review this and approve. And once you move it to the approved status, then the next thing happens, and then it may bring it back to another step. So my. Whenever I write, whenever I create any of these processes, I run all my approval steps through clickup, and then it's just like working with a regular team. I go to Clickup, I have tasks waiting for me. I look at them, I approve them, uh, I get responses back, we chat in the chat. Like it's, it's all working. Like there's an actual team of people there. Uh, there's really no big difference between the two. the user interface is just very, very different. and that to me, because I, I'm used to working in clickup, it's just a very helpful thing. And for those of you who, who are used to working like ban board or lists with statuses and so on, uh, you will find this very easy and intuitive as well. And so it's, it's just something to consider. And again, you don't need to know anything. You just need to tell Claude what you want to do and what's the name of the board and what are the statuses, and he will know how to work with you.

David Yu

Yeah, exactly. yeah, and then I, I do the same thing with, uh, with Notion. yeah, but it's a little bit different. but yeah, I, I like your point. It's everyone has their own preferred UI and like habits. Uh, but yeah, pick whatever is comfortable.

Isar Meitis

Yep.

David Yu

Yeah. so yeah, I think, um, that's, that was the whole process, right?

Isar Meitis

yeah. So, so let's do a quick recap.

David Yu

Yeah.

Isar Meitis

And then, and then I have maybe a very important follow up question, but, so what the process does is it does an initial setup then. Once it has the setup, it loads all the emails from a CSV file, which then you, again, like you said, you can get from multiple sources. it then verifies the emails to see that they're real emails and not fake, which is not gonna damage your email reputation. It then does the research to find more information about the person, about the company, about, customer reviews, about everything that it can find. Then it writes personal emails, right? That's the next step. Uh, so it writes personalized emails for each and every one of them. Then it loads them to the tool, like whichever tool you're using for the code lead. And then it actually creates the sequences and, and runs the sequences. Really well done. And again, the results you, you just said yourself are like really crazy high, uh, response rates because it's talking exactly to the pain point of the people. my follow up question is how did you create the skill? what was your thought process? What was the kind of like your practical steps that you took to develop the skill that now does this magical thing?

David Yu

Um, so whenever I work with Claude code, I usually just start a brand new folder, like a new project with nothing in it. And then I just tell exactly in plain English like what I want to do. And then there is like, testing, I go back and forth with Claude code and once, like I, I have the whole thing and then got the output I want. I, I just tell Claude code straight up, like, can package this whole process into a, a Claude skill and just does it for me. Uh, and then it condenses everything into the, the step. So it is kind of like a SOP you giving to an intern, like you just creating the SOP for ai. Um, so that's, that's what it does. Yeah.

Isar Meitis

and I love the way you explained it, right? It's a collaborative process, right? Yeah. To come in and say, okay, this is what I want to do. I want to create a, in this particular case, a cold email process, but this could be anything else that you want to create for your business. How do we do this? And if you have an idea, you can say, okay, I, I have a starting idea. But then you brainstorm back and forth what should be the steps, what should we do, what's important? And then you just go back and forth and you start testing it inside. Before you package it, you just start testing it. Okay, let's try this step. Let's try this step. Is it working? It's not working, and it's not gonna work out of the box usually. Two to five iterations until you get every step to actually work properly. And then once all the steps are working, you're like, okay, now let's run a test beginning to end. Let's here's a bunch of emails. Uh, let's see how this runs all the way through. And then when all of this runs through, you're like, okay, now package this as a skill. Now you can always update the skill. And I update my skills all the time. I actually learned last week that there's a limit to the number of skills you can have, which is not something I knew. Uh, so I, it's like 50 or 60 and it said, oh, you, you reached your maximum number of skills. I'm like, what? Why? why would that be? And there's, there's a solution for that as well. But nevermind right now. but you constantly update the skill, so you now learn, oh, let's say you just heard David talk about. Find a country and then make sure that you're not gonna get, sued in that country. Then you can go back and say, oh, I want you to add this to the skill, and it will go on research. And we'll add that as additional instructions. You don't need to know how the skill is written. You do need, you don't need to write the actual MD file of the, you don't need to do any of that. You're talking simple English with Claude code. You explain what you want and it will update and change whatever it needs to be, it needs to do in order for it to work properly. David, this was fantastic. If people want to work with you, learn from you, follow you, uh, what are the best ways to do that?

David Yu

Sure, you can find me on LinkedIn or on YouTube. I, I have a little channel just to find, search up David, you Future Men Labs, or you can go to future men labs.com. That's my company website. but yeah, it's, it is been really fun for me as well because, um, usually I work alone, so it is good to have some people to share this with and yeah. Thank you.

Isar Meitis

Awesome. Thank you so much. I really appreciate you taking the time and sharing all of this with you and building a really well explained step-by-step process.

David Yu

Thank you.

Head Shot

Isar Meitis

Host