I built my AI agent. What worked and what didn’t.
Since AI became a thing I immediately started studying it. At this point I am old enough to distinguish project which clearly feel like a hobby from projects are considered as a business ideas.
While I still like to play around with tools to explore how they work and test their limitations I am way more oriented on results. In the end projects that are not treated as a hobby should be measured somehow, whether, it is financial outcome, number of followers or anything else it should me measured in metrics.
This is exactly the approach I have taken when I build my first AI agent in application that is called n8n.
What is n8n?
n8n is an application that allows you to build workflows and connect different applications, so they can talk to each other. It is pretty similar to make.com of ifttt.com, where you also could build similar automations.
The difference here is that I have build it on my own server (Hostinger) and thanks to this approach I have way more control over my privacy.
The goal of the project
The objective of my project was pretty simple – I wanted to create a lead generator for myself, so I can just wake up and have list of potential leads I could contact right away. I do not think that it was an innovative approach in any sense, however definitely something that could improve my life significantly. Overall I would say that the project is working well, but it is not ideal and this is what I will want to break down in this post.
Limitations
As the starting point I wanted to start scraping all the major websites like LinkedIn, Reddit, Facebook, Instagram, X, etc.
Very quickly I encountered some issues – these websites protect themselves very well from exact behaviours I tried to perform.
For instance, Meta in general is extremely horrible platform to interact with (I believe that they do it on purpose), so I gave up with this connection very quickly.
LinkedIn is a bit easier to scrape, but I am a bit scared that they will ban my account and I will end up being in worst condition than before. I decided that risk-to-win ratio was not worth it.
I focused on Reddit which has always been quite friendly for scraping information. I used to scrape information from there very easily back in 2020 when I was building my applications with Robotic Process Automation tools like UiPath. To my surprise they implemented anti-bot features as well, so I was not able to connect easily. Which is okay as I also don’t believe you can get many leads from this website as everyone is hidden under a nickname.
Twitter… excuse me… X… (like if anyone uses this name?) also requires extra payments to access its API and you need to be approved by their developer team if I am not mistaken. I gave up on it too.
So what left?
I managed to successfully scrape YouTube comments which was a bit unexpected ,but works very well. The problem here is that you can’t really interact with users much apart from replying to their comments which after few months of testing it I can tell you was not a very successful strategy.
In the end I managed to connect to LinkedIn, but not in a way I initially wanted. Instead, I started using paid app called Apollo.io which has a strong database of users and contacts to them, so now my bot is looking for this information and then learns about the user as much as it can, which helps me to establish the connection with this person.
The outcome
Since I started using my workflow I am overall quite satisfied with the outcomes. Don’t get me wrong – there is still a lot of effort I need to put into building the relationships.
I am not convinced (at this point) to let AI initiating the conversation. Who knows, maybe I will get to this point, but not until I check the process personally and get some learnings out of it.
Frankly speaking, I do not believe that there are shortcuts when it comes to building relationships with people. This is something that AI will not replace anytime soon.
Artificial intelligence can for sure help you find relevant people, but will not meet them for you. I would look at it as Tinder app – you can get a match, but it is up to you whether you will end up in a relationship or not.