Building a Lead Gen Automation Prototype

Hey guys, Mitch from Aisle here. I wanted to run another experiment today. So, might be a long video. We'll see how we go. One of the questions we get a lot is around how Aisle as a platform compares to other AI automation platforms N8N for example, Gum Loop for example, Make, Zapier.

And we have a bunch of content on the blog around how we specifically compare to a visual canvas-based, node-based automation platform. I thought it might be interesting to go through the experience of building something together and see if that highlights any differences. And I thought it'd be cool to look at a vein of all that you see a lot in the N8N community for example. It has a great library of templates which are basically that you can just pull into your environment and configure to go. A number of them you can see here are to do with scraping Google Maps business information and basically creating leads.

So, you can see here this one scrape Google Maps business leads with Appify and GPT-4 email extraction. There's another one here using Bright Data and Google Gemini. You can see examples of the canvases over here and what they might do. Enrich Google Maps with Scrape Yer and Google Sheets. Various levels of complexity.

This one extremely complex. It's actually multiple loops here that are doing different things. All of these I'm sure do what they say on the tin. I know that Aisle is a platform that can do this kind of thing too. So, what I want to do is use this as a jumping off point to dive into the task builder and achieve essentially the same thing.

Let me close some tabs. We have a couple of great integrations that I already know that we're going to want to use. Manage tools. We're going to want Google Maps for sure. I think Apollo might be interesting.

Getting lead data, verifying emails. A lot of those and then automations just drop things in a Google Sheet. Let's close the loop and maybe put something in HubSpot. So, we have those three tools as a part of the package. I don't think we'll need any memory right now.

We're going to use AI to build everything. So, now we have those tools added. Let's think about what we are going to ask the builder to do. The overall purpose is to find local businesses matching a search um in a given city. Then we want to search one as a potential customer.

Then we want to find lead contact information and enrich. We want to push qualified leads into HubSpot and post a digest in a Slack channel. So, I think that's the overall steps, right? Let's talk about some inputs. All right.

So, the inputs, I think we want to query. So, for example, it could be like dentists and business to search for. Right? We want location, e.g. Austin, Texas, which is the location to search.

Our dental supply wholesale servicing all of North America. Okay, let's just try this, right? I'm more interested in the plumbing than anything else. So, if I copy this and let's go iterate with the AI. Post that in here.

Okay, let's look at the summary. So, task is fully built and ready to go. Here's the summary. It's a nine-step pipeline. Takes inputs, business type, for example, dentist and location.

Geocodes the location via Google Maps. Nearby search finds up to 20 matching businesses within 10 km. Gets place details. AI scores each lead. Right.

It's only using Google Maps, HubSpot, and Slack. Right? Let's accept those changes. Use Apollo for further lead enrichment and contact searching. Okay, so the enrichment pipeline is now set up.

We are doing contact search. We're doing enrichment. Everything's going there. I'm going to accept those changes. The only other thing I want to see if I can do is change the Slack channel that the digest will post to and then we'll go over the package and see what's there.

So, let's look at the overall package. We have Google Maps integration, the Apollo integration, HubSpot, and Slack. Two inputs. So, we have a query and a location. We have the main script here.

So, this is the overall pipeline. I think it was nine steps. So, number one is get inputs. These are just the inputs that we're sending here. So, it's getting both of those, saving them as query and location, and then logging them both.

It's geocoding the location. So, [sighs and gasps] it is doing a Google Maps geocode based on the location string. Right? And then a bunch of if else to find the nearest lat longs if needed. And then it's going to search for businesses.

Get the place details for each business. Right. Then it is going to run everything through this prompt that is in the package. Score enrich lead prompt, right? This is a Sonnet 4 6 prompt.

It's written all of this itself. Lead qualification expert scoring guidelines based on the input that we put at the start. We're a dental supplier. These are all the things that it is pulling from Google Maps. The output is going to be a JSON object.

That's fine. Go back here. Once it has that, then it is going to do some Apollo enrichment. So, it's going to find based off the bare domain contacts in Apollo. Right?

It's looking for decision-maker contacts. Then it is going to find LinkedIn, a bunch of other stuff. Right? Compile all of this down into a lead, pick the best contact for the lead, push it into HubSpot, and then send the Slack channel message with a digest. We're going to test this in a second to see if it works.

We think about the message that we started with, or the overall kind of input that we gave. This is generally what we want to do. You can see here that it's extrapolated a lot of complex pipeline logic to do this, and it has made a number of correct assumptions. We're going to say we're looking for dentists in Toronto, Ontario, Canada. Let's run it and watch the logs.

Geocoding, looking at dentist businesses, scoring the first 20 that it found using this scoring prompt, logging that it has 20 qualified leads all scored, running these all through Apollo to enrich them, find contacts where applicable. It's done. I have a chat ready here that in Aila is going to tell me what has happened. We can look at that. This is all of the information here.

At the same time, it has created contacts in the CRM with whatever details it can find. Contact insights has email, contact information. Pretty light so far. I think Did it create a companies? Tried to, but it didn't do a name.

Okay, that's something that you'd have to fix. Same time, it has created a digest here in Slack, right? So, you can see here it sent this message. Gave everything a score. General dentistry eight out of 10, that makes sense.

Nine out of 10, this one has office manager details and the owners and all that kind of stuff. So, it works if you're using prompts or external integrations, you want to tune the way those are used. But, after some debugging, we had a simple input that flushed out into a pipeline using one, two, three, four, five integrations taking the input query of business type and location, finding companies, enriching the details on those companies and the contacts within them, adding them to a CRM, and messaging a digest to Slack. All pretty neat to put together. So, yeah, fun experiment.

If you like these, let me know. Happy to do many more of them. I think it kind of shows the difference in workflow for building automations like this, right? Like, I am less focused on a canvas and the screen and clicking things around and making sure the plumbing is correctly, and I'm much more focused on what I want and the logic and things like that. Thanks.