A lead generation workflow I help businesses with - example

Hello, I thought I'd try to share some value here, so I'm sharing how i generally set up a lead generation workflow and how I walk through a business to build it for them.

Each lead generation workflow is similar from business to business. I've found it to be true across multiple niches, BioTech, AEC / MEP (Mechanical/electrical engineering), Life Coaching, and Drone surveying...

The workflow is typically 4 Phases - Gathering -> Filtering -> Researching -> Delivery

Gathering - The gathering phase is typically done through an API or web fetches. If an API is available, this is usually the most reliable. Any common model can do a web fetch, but it's less deterministic.

You will also need sources for the web fetch and API. You would determine this based on the clients typical sources (websites) and/or doing quick market research to determine where the ideal leads reside.

I typically don't use filters in this phase unless really broad filters can be applied (Gather anyone who has a particular title CEO, P.E. Mechanical Engineer, etc.), otherwise agent web fetches can have large bottlenecks. API filters in the phase can be useful, but make sure your agent knows the correct parameter values for the particular API. It's useful to find the page where it describes the API parameters and give it to your agent when it's building.

Web fetches gather the whole page data, so you'll want your agent to know "I'm looking for people on this website and their contact info, email, phone, title, etc."

Have it output into an MD file or something simple for the agent to read. This trims down so the filtering agent can work with less noise. (Tell the agent to include the URL/Link to the particular person)

Filtering - This phase is pretty self explanatory, it filters the leads you've gather based on the companies parameters. When talking with the client, you'll want to ask a series of questions to figure out what truly makes a good lead for them / a good client. Often times, they don't know, so you'll have to do some questioning (not interrogation hah).

Once you know what makes a quality lead, you can either put that into the prompt, MD file (lead qualifications), or a python script (more deterministic).

Researching - Once you've filtered down your list, you can then do some extra research on the remaining leads to truly determine which leads are the best quality for the company. I usually use playwrite MCP and let me agent do some extra research on the companies or individuals. This is where you'll search for key indicators of the leads that are ready to be prospects for the client you're working with. You'll need to get as much info from the client about this step because lots of people share similar profiles / snapshots of where they are at in terms of lead quality, however there extra indicators/context a typical filter wouldn't be able to parse, so that's why this step is extra useful. I'll share more about this below.

Delivery - Finally, when you've got your leads that passed all the phases, these are your qualitfied leads. Now you need to deliver it them. Depending on what your client wants, you'll deliver the leads in a particular way. The most common I've seen are telling your agent to make an excel file, an email of the list of leads, an HTML format, or potentially they'll want you to connect the leads directly into their CRM.

Just make it look nice!

That's it! I know this is a bit rough around the edges, but that's essentially the birds eye view of a lead generation service done for a business!

I'll share a real run for a BioTech company.

In my gather phase, it gathered 1,000 raw potential leads (took roughly 30 seconds. I usually do 5,000 - 10,000 at a time)

After the 1,000 were gathered and output, I sent the output through the filtering phase, which did a series of filtering (something like 10 specifics that I was looking for and was baked into the workflow) and it gave me an output of roughly 197.

After that, I wanted to do some extra filtering and had recently gotten an email from the client indicating more specifics for a quality lead, so I simply had my AI mid workflow check that output and run the extra filter on the 197, which brought the total down to 18.

Once I had the 18, I ran the researcher (Agent prompt + Playwrite MCP) and it gathered all the information about these companies that my client wanted.

I ended up with a list of brand new leads for my client in less than 30 minutes. On average my client recieves 30 - 60 leads per month without me. So I just increased their lead count by potentially over 25% - 50% in an hour when it typically takes 30 days... Unreal.

Summary - 1,000 Raw leads -> 197 filtered leads -> 18 high quality leads -> delivered directly to the client.

Total time once built, less than 1 hr. (typical time is 1 month for 30 - 60 leads)

Total time to build the system, 4-8 hours of dedicated work and troubleshooting.

Total time to do this the traditional method:

Traditionally: 1 month of time for 30 - 60 leads (from an agency that doesn't really care)

AI Automated: ~4 hours per 60 high quality leads.

7 comments