Transcription
I booked over 400 sales calls for myself and my clients using automated AI systems over the past year. On this channel, you learn how to build agentic systems that generate real revenue.
If you don't know me, I'm Jose. I build automated lead genen systems for B2B companies and run an outbound AI agency. Let's get straight into it.
Okay. So in order to build this agentic LinkedIn lead scraping system, we need these four steps. We need LinkedIn sales navigator. We need a scraper from amplify. We need a script to clean up the data. And we need to export this data to a CSV. So let's go into Google Anti-gravity and build this out.
Okay, here I am in anti-gravity and I want to open the agents manager. Let's get this guy here. And we want to open a new workspace. Open new workspace. Let's create a new folder. Let's call it create and then open. Click yes.
What we need next is the initial agents.mmd file in order to create the environment for us to build the agent and we can get that here. Basically it's called the dose framework directive of what to do orchestration which is the decision- making layer. We have safeguards and we also have the execution layer of the Python code actually running and doing the task we wanted to. Let's go ahead and copy that.
Before we continue, if you want the framework I used to architect these agents, the agents.md file is in the description below for free. If you want an agentic AI system generating leads and scaling your revenue or you want a fractional AI partner handling the automations across your business, there's also a link in the description to book a call with me. Let's get back into it.
New file. Type in agents.md. Okay. And then we just copy and paste everything here. We make sure to save it. Command S. Now all we have to do is tell it to execute the agents.mmd. Execute the agents.md file and it's going to set up the environment for us.
So, it's asking us one of the questions on the safeguard of what's our budget constraint if the API is spending too much money within a certain time frame. Let's put in $10. And what this does is it prevents your credit card from getting overly charged if the API goes wild and you know it ends up scraping I don't know a million leads or gets stuck in some sort of loop. Uh this will prevent that from happening. Let's accept.
On the left here you can see it created the three folders. So it asks us if you want to create a NV file which we do because that's what we're going to store all of our API keys. Uh please create the env. Great.
So now what we're going to do is we're going to give it the prompt in order to set up the scraping system. So the more information and guidance you give it up front, the less back and forth you have to do in order to get the agent to do what you want. So, it makes sense to spend a bit of time up front uh figuring figuring out what the SOPs should be. So, let's give it the prompt. Create a directive on a LinkedIn scraping system. The way that the system is going to work is all I'm going to do is provide you with a LinkedIn sales URL. From that sales URL, you are then going to call the correct Appify scraper. Once the Appify scraper finishes running, you are then going to clean up the data to remove any weird fonts or weird company names and so on so it can be usable for an outbound campaign. Once this is done, please provide the data in a CSV file.
Now let's press enter. And basically it's going to give create a directive first instead of us just telling it to do the task as that allows us to review and fine-tune what we actually want. So so far this looks okay to me, but we want to give it a bit more details. Let's go to ampify. And again, it's much easier to just tell it what we want instead of it looking around for different scrapers. Uh, this is the scraper I want to use. Let's just read through the information to see if there's anything we and the agent needs to know. So, there's two flows in this specific scraper. it does a new search and then you need to fire it up again in order to fetch the results. So I'm just going to give this information to the agent.
Okay. So what I did is I just took pictures of the API document and gave it to it. I gave it the exact scraper we want to use. So now we can tell it to start building this integration for us. So, here are the API documents for the specific Apify scraper I want to use, as well as the scraper itself. This does a twopart scrape where there's two workflows, meaning we would need to wait about 5 minutes before calling the second workflow to see if the scraping has completed. If the scraping did not complete, then wait another five minutes, three more times until it does. Please build out the scraping function with Appify. So when I give you the LinkedIn sales URL, you're able to call that specific actor.
And now we send it the command. So you could see it's thinking it's doing the edits to the initial script. We will need to give it the API key for Ampify, but let's run it and then see if it's actually able to troubleshoot and then ask us for the specific API key we need. So I'm going to go ahead and get this LinkedIn sales URL that I already have preloaded. And while it works, you can pretty much watch it think or you can go ahead and grab a coffee, work on some other things. Um, it pretty much works in the background, allowing you to do whatever you want afterwards.
So, you can see it's asked us for the API key because it doesn't have it. So, let's go ahead and do that. So, on the left, we can go to settings. And you're basically going to click API and integrations. I'm not going to click this on video because it will expose my key.
So if we go back to ampify, we could see that the in the agent successfully ran the initial search. So what it's going to do next is in 5 minutes, it's going to send this request ID back to this Appify scraper in order to get the actual scraped list. So, this costs 50 cents each search and then a marginal amount of cents for every lead that you get and you get about $5 uh for free on a free plan on Apple. So, you're able to test this out without paying any money at all. The only thing you really have to pay for is LinkedIn Sales Navigator.
And here now it's just waiting the five minutes in order to recall um the scraper to download it. Cool. So after 5 minutes, it called the second part of the scraping workflow. It waited and it even fixed one of my filters where due to it being tied to my personal LinkedIn account of shared experiences, it actually can't be scraped through through a cloud-based scraper. So, it actually went and made it the targeting better on my behalf without me needing to tell it to do anything.
So, let's go ahead and click into the leads. So that's not going to give us anything. So what we want to do is doubleclick reveal in finder. And that brings us here. So we have the leads in JSON. And then we also have it in uh CSV. So let's take a look. And there we go. We have their name. We have their first name, last name, full name, job titles, company name. Um, we can see that some of it still has not been fixed. For example, we still have like LLC here. We have co, we have all caps here. So, all we want to do is just great job. Um, but the company names aren't clean. I see LLC's. I see all caps. Please format them correctly and then save that to the directive folder as well so you don't make that mistake again. Don't run the entire script. Just clean it without running the entire Ampify workflow. Great. Give me the link to the CSV file. Let's see.
And you can see that it actually did clean up the company names this time. And now it knows for future runs of the workflow to automatically clean the list without me needing to tell it again. So these systems get stronger and more powerful as you use them more and as more and more mistakes and edge cases are found, it then creates code to reinforce it.
So now that we have this information, this is great. But what would actually make it extremely powerful is if we added a email finder and verification system to this workflow, right? Because with this information, we can't really do an outbound campaign other than on LinkedIn. We would want their email addresses. So you can use a tool like prospel or any mail finder in order to find and enrich their email addresses and then automatically send that to a code email platform to automatically send. So I'll create another video on that full system at another date. But that's pretty much how you build a LinkedIn lead scraping agent using Google anti-gravity by only giving it a LinkedIn search URL. So that's the full breakdown.
The links are below for the agentic workflow strategy session as well as the agents.md file if you want the framework. And if you got value from this, uh, subscribe. I'm constantly dropping new agentic systems. So, I'll see you on the next.