The Pool Installer That Never Sleeps
Someone built a business that runs while they sleep. Here's how.
An AI agent (a program that runs tasks on its own) scans property records, finds homes worth between $500K and $1.2M that don't have a pool, creates a picture of a pool in the backyard using AI image tools, and mails a physical before-and-after postcard to the homeowner. No salesperson. No ad budget. Just an automated program, public property data, and the postal service.
The homeowner gets a personalised mailer showing exactly what their backyard could look like -- with a quote. It works because the person sees their own home, not a generic ad. That kind of personalisation, sent to hundreds of people automatically, just wasn't possible before AI.
This isn't a concept. It's running right now. And it's one of dozens of businesses using the same framework.
The 3-Step Framework
Every AI sales agent follows the same pattern. Find the data, build the pitch before they ask, show the numbers.
- 1 Find the data. Every industry has public information that tells you who needs your service. Property records, business registrations, permit filings, job postings, social media activity. The data is free. What you do with it is where the value is.
- 2 Build before they ask. Don't wait for the lead to come to you. Show up with the solution already done. The pool installer doesn't ask "would you like a pool?" -- they show you what your pool looks like. The proposal is the marketing.
- 3 Show up with the numbers. "Here's what you're missing out on" closes faster than "here's what we could do." When you can put a dollar amount or time saving on it, the sale is half made before the conversation starts.
The AI does the prospecting, the analysis, and the pitch. A human only shows up for the close.
Five Industries This Works In Right Now
| Industry | Data Source | What the Agent Does | The Pitch |
|---|---|---|---|
| Pool / landscaping | Property records, satellite imagery | Renders improvement on actual property | "Here's what your backyard could look like" |
| Solar installation | Property records, roof imagery, energy data | Calculates savings for specific home | "You're overpaying by $X/year on energy" |
| Real estate | Market data, ownership records | Identifies undervalued or off-market properties | "Your property is worth $X more than you think" |
| Home renovation | Permit data, property age, comparable sales | Renders renovation on actual property | "This renovation would add $X to your home value" |
| Commercial cleaning | Business registrations, lease data | Identifies new businesses opening | "Welcome to the neighbourhood -- here's a first-month offer" |
The pattern is the same: public data + AI analysis + personalised outreach. The industry doesn't matter.
Building Your First AI Sales Agent
- 1 Choose your data source -- what public information tells you someone needs what you sell?
- 2 Set up the AI processing -- use AI to go through the data and find the right people to contact. You can use tools like Claude, ChatGPT, or a no-code tool like n8n (no programming needed)
- 3 Create something personalised -- a picture, a savings estimate, a custom report. Whatever makes the pitch feel real and specific to that person
- 4 Set up the outreach channel -- physical mail, email, SMS, or a landing page with the personalised result
- 5 Add the human handoff -- when a prospect responds, route them to a real person for the close
- 6 Track what works -- see how many people respond, figure out what's getting results, and adjust from there
Tools You Need
- 1 Data access -- ways to pull public records. Tools like Apify or Bright Data can grab info from websites automatically. Google Maps API can pull location data
- 2 AI processing -- Claude or ChatGPT to analyse the data and write personalised messages
- 3 Image creation -- tools like Midjourney, DALL-E, or Flux that create realistic pictures (like showing a pool in someone's backyard)
- 4 Automation -- n8n, Make, or Zapier to connect everything together so it runs without you
- 5 Outreach -- Lob (sends physical mail automatically), SendGrid (sends emails), or Twilio (sends texts)
- 6 CRM (a system to track your contacts, like HubSpot or Pipedrive) -- to keep track of who you've contacted and when a real person needs to step in
You don't need all of these on day one. Start with the data source, the AI layer, and one outreach channel. Add complexity once it's working.
What to Automate vs. Keep Human
| Automate | Keep Human |
|---|---|
| Collecting data and finding the right people to contact | Final pricing decisions |
| Creating personalised images and estimates | Relationship building |
| Initial outreach and follow-ups | Handling objections |
| Ranking which contacts are most likely to buy | Closing the deal |
| Scheduling and calendar management | Checking outreach quality before it goes out |
The goal isn't to remove humans from sales. It's to remove humans from the parts of sales that don't need them -- so they can focus on the parts that do.
Follow @reallyusefulai on Instagram for more practical AI guides.