Video summary
I Built a Client Website in 51 Minutes With AI (The 2026 Agency Workflow)
Main summary
Key takeaways
Tech/Workflow Summary (AI Website Build in ~51 Minutes)
The speaker demonstrates an end-to-end “agency workflow” for creating a client website quickly using multiple AI tools—covering:
- Research
- Design
- Code generation
- Publishing
- Connecting a CRM/contact backend
- Generating missing assets (photos/icons)
- Basic SEO verification
Tools & How They’re Used
-
Stitch (Google AI design tool)
- Used first to generate a premium website design from a prompt (initially home/about/contact pages).
- Supports:
- Prompt enhancement (improving the design prompt)
- Alignment to a branding guide (logo/colors/typography)
-
Antigravity (AI coding tool)
- Used to:
- Run a Claude agent for business research extraction
- Convert Stitch output into an actual Next.js + Tailwind codebase
- Create shared components (header/footer), build pages, generate config/CSS, and wire everything for deployment
- Audit/fix broken UI elements (icons/images) and update missing assets
- Used to:
-
Claude (inside Antigravity / Claude agent + code skills)
- Used to:
- Fetch business context from a Nextdoor profile, extracting photos, testimonials, and business details
- Save research into a structured business research markdown file for reuse
- Create/expand prompts for Stitch (via “skills”), and generate additional files/folders like a branding guide
- Used to:
-
GoHighLevel (CRM / contact form backend)
- Used to:
- Create an inbound webhook trigger
- Map webhook fields into CRM contact records
- Test and confirm lead submission flow
- Add an automation step: send a follow-up email to the lead after form submission
- Used to:
-
Vercel + GitHub Deployment
- Antigravity pushes the generated project to GitHub
- Then deploys/publishes via Vercel to make the site live
-
Image/asset generation via “MCP” + Nano Banana Pro / Key marketplace
- When external images/icons were missing or broken, the system generates replacements locally.
- Generates multiple gallery photos, an owner/hero image, and downloads them into the site’s public assets.
- Also generates/fixes certain missing map imagery and updates visuals on pages
Key Steps / Tutorial Flow Highlighted
- Start a timed test claiming “<30 minutes” and complete in 51:20.
- Pick a local business (example used: Schupp Construction and Landscaping / “Shoop Construction” appears later).
- Extract business info automatically:
- Claude fetches the Nextdoor profile and gathers: phone/email/address, years in business, services, and review metrics.
- Create a design in Stitch:
- Prompt includes required pages (homepage, contact, about).
- Later, use an “enhance prompt” skill for a more premium/chic design prompt.
- Build a branding guide:
- Antigravity asks Claude to create a folder with logo + colors + typography rules.
- Stitch designs are iteratively corrected to match branding specs.
- Generate additional pages (services, about, contact) within Stitch and re-prompt to match branding accurately.
- Export Stitch design → generate a real site:
- Stitch output is exported and uploaded back into Antigravity.
- Antigravity generates a full Next.js app with shared components and separate pages.
- Deploy to production:
- GitHub repo created/updated
- Vercel deployment executed and verified as live
- Connect contact form to GoHighLevel:
- Use webhook from GoHighLevel into the site.
- Validate webhook submission by sending a test lead into the CRM and saving a note.
- Generate and replace missing assets:
- Audit external image URLs and replace broken ones with AI-generated images via MCP/Nano Banana Pro.
- Fix logo placement in header/footer and verify final visuals.
- Add an email automation in GoHighLevel:
- On inbound contact, send a thank-you email referencing the inquiry.
Product/Quality Findings & Issues Mentioned
- Design quality: Stitch quickly produced “premium” visual styling, including UI effects (e.g., glass morphism and hover interactions).
- Branding adherence: Iterations were needed—color/typography sometimes initially didn’t match the brand guide, so the speaker re-ran prompt enhancement/editing in Stitch.
-
Broken form submission at end: The final test submission initially failed (“Something went wrong. Please call us directly.”). The speaker notes this is a quick fix in principle, and that the demo’s workflow was otherwise largely successful.
-
SEO claim: The deployed page was stated to be “completely SEO optimized” with title/description/keywords, validated using a “Google crawler simulator.”
What’s Specifically Emphasized as “Fast Wins”
- Automating research → design → code → deploy using chained AI tooling
- Reusing extracted business context (Nextdoor info + reviews) to avoid generic “AI slop”
- Using a branding guide to keep consistent colors/typography/logo across pages
- Automated asset recovery (generating missing photos/images/icons) rather than manual searching
- Wiring the site to CRM + webhook without building custom backend logic
Main Speakers/Sources
- Main speaker: Braden Carter (marketing agency owner; ~3 years experience building websites/SEO for local service businesses)
- System/tools/sources referenced:
- Google Stitch
- Antigravity
- Claude (Claude agent/code inside Antigravity)
- GoHighLevel
- GitHub
- Vercel
- Nextdoor (business research source)
- Nano Banana Pro / Key marketplace via MCP (image generation)