Building Spotlight Contractor: a real AI app, built the way this book teaches

Spotlight Contractor Edition is a marketing app for home improvement pros. A contractor answers a few questions about the business. The AI drafts the ads from those answers.

This post is the case study of that build. The process is the scope-first method from The AI Advantage for Home Service Businesses, applied to one real app.

The problem: marketing that lives in the gaps

Every contractor knows this story by heart:

• Marketing happens between jobs, a Facebook post here, a Nextdoor mention there, a flyer that never quite gets made.
• Referrals are the best customers, but referrals only happen when the phone rings, and the phone only rings when people remember you.
• Ads fail not because contractors don't know their business, but because the ad platforms are built for marketing agencies, not for a crew lead who just finished a 12-hour day.

That's the Four Losses from Chapter 3, concentrated into one industry: time lost fighting ad tools, knowledge lost (the good stories stay in the owner's head), margin lost on ads that don't sound like the business, and opportunity lost on leads nobody followed up.

The one-page map

Chapter 2's exercise: draw the whole business on one piece of paper. For a contractor, that map looks like this:

1. Homeowner sees an ad or a review
2. Homeowner calls or fills a form
3. Phone tag. Estimate scheduled.
4. Estimate given, and here's where it stalls
5. Follow-up (usually none, or one awkward call)
6. Job booked, job done, invoice sent
7. If they're happy, maybe a review, if they remember

The binding constraint jumps off that map: everything between step 1 and step 4 is manual, inconsistent, and mostly skipped. The before/afters and the "we finished Tuesday, here's the reveal" moments never reach the people deciding between three bids.

The entire build exists to lift that one constraint.

Design what good looks like

Before a single line of code, "fixed" had to be defined. The acceptance criteria:

• The goal written into the scope was that a contractor who can describe the business in plain English could set up a campaign in under 10 minutes.
• The AI writes the ad copy, imagery, and targeting from their answers, no copywriting skill required.
• The owner sees the whole picture in one place: leads in, cost per lead, what's working.
• It speaks the trade's language. A roofer, a kitchen remodeler, and a basement finisher get different campaigns because they are different campaigns.

The first line is a goal you could check with a stopwatch. That is what makes a scope buildable. Putting real ads in front of homeowners also needs a connected ad account, ad spend paid in advance, and the ad platform's own review, so the done line stops at the draft.

The extraction interview

Chapter 8's interview pulls the knowledge out of the owner's head and into the system. Five questions, each mapped to something the AI needs:

• "What do you do?" → Business Category
• "Where do you work?" → Service area
• "What are you offering?" → The campaign's offer
• "Why should someone pick you?" → Proof and trust
• "How should you sound?" → Brand voice

Five questions. No repeats, no waste. Every input has a purpose; every output traces back to an input.

The build

These are the main pieces this app used. The AI Advantage for Home Service Businesses covers how to think about a development stack in its stack chapter. Pick per job. This list is one app's choice.

• Frontend: Next.js on Vercel, the wizard, dashboard, and landing pages
• Backend: FastAPI on Railway, campaign logic, lead tracking, integrations
• AI images: FAL.ai, ad creatives from the contractor's project description
• Ads: a connection to Meta Ads (Facebook and Instagram)
• Data: PostgreSQL behind per-user accounts

The app also used outside services for payments, email, and AI writing. Plain notes on tools a first build often needs are in the free tool guides.

The advantage is knowing what each piece is for and how to wire them together.

The 7-step wizard

The heart of the app. It walks a contractor from "what do you do" to a draft campaign:

1. Business Category, Remodeling, Restoration, Hybrid, General Contractor, or Specialty Trade
2. Business Basics, name, contact, service area, hours
3. Business Details, the specifics that make a campaign real
4. Services Checklist, which services they actually offer
5. Proof & Trust, reviews, credentials, guarantees
6. Brand Voice & Goal, how they sound, what the campaign should achieve
7. Review & Complete, everything in one place for the owner to approve

Each step is a narrow question with clear options. The owner never faces a blank page, they face choices they already know the answers to. The software asks, the AI writes, the contractor approves.

The dashboard

After the draft, the owner has three views: the lead funnel (who came in), active campaigns (what's running, and the cost per lead), and recent activity.

The dashboard's job is to answer "are my ads working?" before the owner has to ask.

The results

Written against the binding constraint:

• Before: marketing lived in the owner's head. Ads were sporadic, follow-up was luck, and the photos that win bids were buried in a phone.
• After: the scope asked for a campaign draft from the wizard's answers: AI-written copy, generated imagery, and targeting, without a blank page.

The app does not do the marketing for you. It makes marketing possible for a person who has a business to run.

Your turn

The book gives you the map. The work is the build: the extraction interview, the scope, and wiring the services into one system.

The AI Advantage for Home Service Businesses teaches the method: the Four Losses, build-ready specs, standing up your agent team, the development stack. The free course, Your First AI Build, Scope First, teaches the scope-first process for planning your own first build, with this app as one of two worked examples.

The book is at /book.

Spotlight Contractor Edition shows the process in The AI Advantage for Home Service Businesses. Campaign data in this post is sample data, not customer results.