
I'm experimenting with a different way to do outbound marketing for my SaaS.
Instead of manually finding prospects, cleaning lead lists, writing emails, and making calls, I'm trying to turn the whole workflow into something an AI agent can execute.
The interesting part is that I don't have one giant app doing everything.
I give Grok Bot access to the tools it needs:
Lead Atlas for finding businesses
My email account for sending outreach
Call to AI for making AI phone calls
Then I give the agent the goal.
For example, my current experiment is basically:
"Use Lead Atlas to find 500 dealer shops in Dallas. Filter out the low-quality results. Send personalized emails to 20 businesses per day from my email account, and make 20 calls per day through Call to AI to introduce the product."
And then I let the agent work through the process.
Step 1: Find the prospects
The first job is finding the right businesses.
Grok Bot uses Lead Atlas to search by location and industry.
In this experiment:
Location: Dallas, TX
Industry: Dealer shops
Target: ~500 businesses
Lead Atlas returns business information such as names, websites, phone numbers, locations, and available email addresses.
But 500 leads doesn't mean 500 good prospects.
Which leads to the next step.
Step 2: Clean the list
I don't want the agent blindly contacting everything it finds.
Grok Bot reviews the results and filters out businesses that don't look useful for the campaign.
This is an important distinction.
I'm not trying to automate:
"Send this message to 500 email addresses."
I'm trying to automate:
"Find businesses that fit this target, decide which ones are worth contacting, and then work through them."
That's much closer to how I would approach the task manually.
Step 3: Send the emails
Grok Bot also has access to my email account.
It can use the business context to prepare the outreach and send emails in small daily batches.
For this experiment:
20 personalized emails per day.
I deliberately don't want to blast hundreds of businesses at once.
I'd rather start small, see what gets responses, and improve the outreach.
Step 4: Make the phone calls
This is the part I'm most interested in.
Grok Bot uses Call to AI to make outbound phone calls to the businesses.
Call to AI is the AI calling product I'm building.
I can provide the phone agent with:
who it's calling
why it's calling
what product it's introducing
what the goal of the conversation is
what it should and shouldn't say
what to do when it reaches voicemail
what to do when it reaches an automated phone system
Then the AI makes the call.
I'm currently working with roughly:
20 emails/day + 20 AI calls/day.
The real test isn't voicemail
Getting an AI to dial a number isn't particularly interesting.
Detecting voicemail isn't particularly interesting either.
The interesting moment is when somebody actually answers:
"Hello?"
Now the agent has to have a real conversation.
Can it explain why it's calling without sounding ridiculous?
Can it explain the product?
Can it understand a question?
Can it respond when the person isn't interested?
Can it recognize when someone is interested?
That's the part I'm testing right now.
And every real conversation exposes something I need to improve.
So my outbound workflow now looks like this
One prompt
↓
Grok Bot
↓
Lead Atlas -> find prospects
↓
Grok Bot -> filter + personalize
↓
Email -> 20/day
Call to AI -> 20 calls/day
↓
Real conversations
↓
Follow-up
↓
Customers
The important part for me isn't the "500 leads" number.
It's that I'm moving from doing individual outbound tasks myself to managing an agent that uses other tools to do those tasks.
Instead of:
"Find another 20 companies."
"Write another 20 emails."
"Call another 20 businesses."
I'm moving toward:
"Here's the market I want to reach. Here's the product. Go work through the campaign."
I think that's a much more interesting use of AI agents than simply generating another sales email.
And yes, I'm using the agent to market the agent
There's a funny loop here.
One of the products I'm trying to sell through this workflow is Call to AI.
So I'm using AI phone calls to find customers for an AI phone-calling product.
That's intentional.
If I can't use my own product to generate real conversations for my own business, that's something I need to know.
And if it works, the same workflow could be useful for small businesses that don't have people spending all day doing outbound.
I'm still early in the experiment.
I'm particularly interested in what happens when more real people answer the calls, what objections they have, and how much of the first-touch conversation an AI agent can realistically handle.
I'll share the actual results as I collect more data.
The two products I'm building and using in this experiment:
Lead Atlas: targeted business lead generation
https://www.leadatlasapp.com/
Call to AI: AI-powered outbound phone calls
https://www.calltoai.net/
I'm curious how other founders are approaching this:
Would you let an AI agent handle the first outbound call to your prospects?
Thanks for writing this up. Bookmarking it for later.