I Built an AI Agent That Scrapes Google Maps Leads Automatically (n8n Tutorial) transcript

I build this N8n workflow that acts as AI powered lead generation assistant. So the workflow takes normal conversation chat request like give me five HVAC leads in New York City and it automatically scrapes the live Google Maps data and stores the leads via AI and it also be preventing duplicate leads and also it will automatically save those leads in our Google Sheet and after that it will automatically replace back those leads in directly in our conversation.

So let me just show you the live demo and after that you know I will simply explain you how the data will flow from node to node from the beginning to the ending. So yep so let us just test the workflow. Give me five HVAC leads in New York City. Okay currently live scraping was happening. Okay currently the lead scoring was happening with AI obviously and the leads are automatically saved in our Google Sheet. Okay these are the leads and just have a look in the conversation right here.

All the leads were already in our conversation itself. So yeah this is one of the one of kind of you know cool automation out there. So let me just explain you how the data will flow from node to node from beginning to the ending but before that you just need to you know make a note of the headers of the Google Sheet so that you can simply replicate this particular Google Sheet and connect it back to the to this particular workflow so that it will automatically works.

So firstly whenever a user simply messages right here this particular workflow will get triggered and after that we'll be using the Gemini's AI model to simply you know convert the normal conversation request like you know give me five barbershop leads in New York City into a structured input like you know business type was for barbershop and location was New York City and right and I simply asked it for five you know leads right so the max is five and the output mode was full details so you know I mean it will automatically return structured output right there and after we simply normalize the data because you know here the whole thing was completely in a messy JSON right I mean in the text this is this text is the AI's output so nothing so serious about it but we will be simply you know make it as a structured individual item so that it will be so easy for us to move forward so we will be extracting business type location and the max results and we will automatically generate a API input like this by using these three you know dependent items because these are the three requirements that are needed for us to scrape the Google Maps via API so the API will be needing that particular you know JSON object from here so we will be normalizing and you know generating that particular structured JSON like this and just have a look search string array we will be seeing barbershop so yeah we will be automatically generating that particular JSON API's input because the actor will be needing the JSON's input so we will be passing that object into this particular API HTTP node and we will be grabbing the results as like this and there are a lot of you know noise right here so to clear out the noise we will be only grabbing the required seals with this particular set node or edit node and after that we will be removing the internal duplicated leads so in where here all the five leads are unique so there is no duplicate leads right here but just in case sometimes there is a possibility of duplication too so for that reason we will be simply ignoring the duplicate leads right here internally but you know here comes the data flow correction so just have a look whenever i mean there are five items that are coming out from the you know duplicate leads remover after that we will be aggregating all of those five you know leads into one particular JSON one particular JSON object okay so after why we actually do is we will be sending those aggregated you know leads to the gymna API so that in one API request the whole five you know leads will be ranked and given to us okay you got me what i'm saying so for that reason for you know making the whole workflow very simple i simply use the aggregate option okay so nothing serious about it but you know we can also use the official gem9 nodes and all but i'm just messing up and i i'm just you know playing with HTTP nodes so because i mean it became a kind of obsession for me so just ignore that and these are these all are you know gem9 nodes so nothing so serious just have a look this is the you know API request URL i'm using so completely i'm just playing with the agent out there so nothing serious so even after that we will be parsing the whole you know unstructured or combined output like this structured output that we get here okay but still just have a look even though even we simply parse the whole leads right here it was only one single thing it was combined one single you know JSON so to say so here we simply use it split out split out node and you know we simply split the combined single JSON into five items so that we can obviously check whether check with this particular google sheet okay whether in whether that particular individual you know place id was already saved in our google sheet so because we are already duplicating the leads that are coming uh for that for this particular session but you know there should be a master to deduplication right so for that reason we are simply using to get google i mean uh google sheet node and we will be giving the place id and currently there is very i mean it simply returns the empty output so because you know no none of these five leads were automatically saved at the time of execution not now now all of those files were actually there but at the time of this particular workflow execution none of those five leads were there in the google sheet so the the that is what it means when it simply gives us the empty output then we will simply combine the stored leads why we simply use this particular JSON node to actually check whether i mean to actually remove the already saved leads from the you know workflow so that we will be only saving the unique leads to our google sheets okay so just for the simplicity again we will be simply combining all the leads into one and the code node will be actually you know comparing the split out leads and the you know and the already saved leads from the google sheet so that you know only the unique leads will be saved back into the google sheet so the code node will be automatically written the unique leads so in this particular case all the five uh leads are not saved in the google sheet so so for that reason the JSON was automatically sent the five items and we automatically you know save those leads to our google sheet and after we simply aggregate them and after we will we simply format the whole combined leads as a chat message and simply reply back like this okay so this is how the whole system will actually work in life and this is how the data you know will flow from node from the beginning to the ending and there are no hidden secrets out here so yeah that's it for today