Teardown No. 59 · Video

90 to 95 percent of his GCI came from content: how William Zhang uses Claude Code to prep his YouTube videos

William Zhang is a solo real estate agent in Austin, Texas. YouTube is his primary lead generation channel. He says 90 to 95 percent of his $180,000 in GCI last year came from content creation, but he was skipping most of the prep work behind it. For seven straight days he ran his business through an assistant built on Claude Code. He describes a video, and the agent returns competitor analysis, a research report with angles and hooks, talking points, a slide deck and four or five thumbnail options. A "good morning" command returns a dashboard with his CRM activity, his YouTube performance and data on other Austin real estate creators. He calls it the most productive week he has had in years. He does not report any change in leads or revenue.

William ZhangVerified Oct 26, 2026Markdown copy

The problem

Last year he closed 16 deals as a solo agent with “No team, no admin, no virtual assistant,” only a video editor handling some of his YouTube content. Most of that income came from content, yet he was not doing the preparation he believes makes videos perform. He names that preparation as competitor research, keyword analysis, scripting, slide decks and thumbnails. He writes: “Content prep should be 50% of the creation process. I was doing maybe 10% because it felt like a chore.”

His mornings also went to switching between five tabs:

  • Follow Up Boss
  • YouTube Studio
  • Facebook Ads Manager
  • Google Sheets
  • a notes app

That routine took 30 to 60 minutes. He considered a virtual assistant at $800 to $1,500 a month, but hiring and training felt daunting and he kept putting it off.

The blueprint (as he describes it)

  1. Start the day with a dashboard. He types “good morning” into Claude Code and gets a dashboard in 60 seconds. It shows:
    • CRM activity, such as which leads browsed or saved listings and who needs follow-up.
    • His YouTube performance alongside competitor data. For other Austin real estate content creators, this covers subscriber counts, upload frequency and video performance.
    • Insights on content gaps he should fill.
    • His top priorities for the day.
  2. Let competitor data pick the topic. He gives an example of what the dashboard “might flag.” In that example, negative framing plus listicle-style videos outperform his generic market updates by 2x. He says these insights “directly inform what content I create next.”
  3. Describe the video and let the agent research it. The agent runs competitor analysis on what is ranking, what is getting views and what gaps exist in his market.
  4. Get a research report, talking points and a deck. The agent writes a research report with specific angles and hooks. It scripts talking points and builds the slide deck.
  5. Pick a thumbnail from options. The agent produces four or five options in under a minute, each built on a different concept: before/after comparisons, bold text overlays or portrait compositions. He picks one and moves on.
  6. Record it himself. He says the deck for the video behind this post was “Created entirely by AI” and the opening hook was “Written by AI.” He adds that “I’m the one talking and delivering the content.”
  7. Feed the CRM side of the funnel. He gave Claude Code his business context, and one prompt rewrote his 50 to 100 Follow Up Boss email and text templates with localized copy. The agent also reads his CRM activity daily and flags leads to contact on the morning dashboard.
  8. Non-distribution work. He also taught it to log into his MLS through the browser and pull showing details for buyer tours.

Stated results

  • “This was the most productive week I’ve had in years.” The trial ran seven days.
  • Morning dashboard: 60 seconds, compared with 30 to 60 minutes across five tabs before.
  • Thumbnails: four or five options in under a minute, compared with 35 to 45 minutes of design work before.
  • The slide deck and the hook for the video behind this post were made by AI.
  • CRM templates: all rewritten in one prompt. No change in reply or conversion rates is stated.
  • MLS tour prep: about 5 minutes per client, down from 30.
  • The 16 deals, $180,000 GCI and 90 to 95 percent content share are from last year, before this trial.
  • No change in leads, views, subscribers or revenue is stated after the week.

Caveats

  • This is a post on his own site, Real Estate AI Society, which he founded. The site sells a newsletter, a free setup file and a paid four-week Real Estate AI Bootcamp. The post repeatedly points readers to the newsletter for the free setup file and his prompts.
  • It is a one-week self-report. The insight examples are framed as what the dashboard “might flag,” not as measured findings.
  • He does not say how the agent gets YouTube or competitor data, how it connects to Follow Up Boss, or what tool generates the thumbnails.
  • The planned items are not built yet:
    • AI video editing for basic cuts, captions and assembly
    • automated lead outreach tied to Facebook ad campaigns
    • AI phone calls, which he describes as “mixed but improving fast”
    • contract automation
  • The post references a companion video of about 19 minutes. It was not reviewed for this teardown.
  • Handles: YouTube @aiforrealtors (verified 2026-10-26: linked from his own site). LinkedIn linkedin.com/in/william-zhang-369335104 (verified 2026-10-26: linked from his own site). An Instagram account in search results is not linked from his site, so not tagged. No X, Bluesky, Threads, Mastodon or TikTok verified.

Sources

Want agents like these working your distribution? Get It Built. Get new teardowns weekly: subscribe (double opt-in).

Get one teardown a week.

Same format, a new operator every issue. Double opt-in.