How Do You Turn One Video a Week Into a Month of Content With AI, Without Sounding Like AI?

An AI content engine starts with one recorded video a week, pulls the transcript, and turns it into an email, three short posts, a written article and a follow-up sequence, all drafted by AI and all reviewed by you before anything ships. The engine works because the source is your own voice answering a real question; the AI is only cutting, never inventing.

If you want this built with someone who has run it at scale, coaching is where we do it, and the free AI scorecard is the four minute place to start.

Why does the engine start with a video and not a prompt?

Because a prompt produces content that sounds like everyone else's prompt. A video of you answering "how do I stop being the bottleneck in my business" produces your examples, your phrasing and your opinions, which is the only thing a prospect cannot get from a chatbot. I run three channels with a combined 282,000 subscribers and coach as a Tom Ferry Mastery Coach; every piece of written content I publish traces back to something I said on camera or on a call. AI makes the cutting fast. It does not make the thinking.

What are the five outputs from one video?

1. The email. 150 to 250 words: the answer, one example from the video, a link to the video, the lead magnet. Sent to the list the day the video goes live.
2. Three short posts. One quotable line each, pulled from the transcript, with the video link. Spread across the week.
3. The article. The transcript restructured under question headings with the answer in the first paragraph, published on your site with the video embedded. This is the piece search engines and AI assistants read.
4. The follow-up sequence. Three short emails for anyone who completed the lead magnet that week, each built from one point in the video.
5. The question list. AI reads the transcript and the comments and proposes next week's question. You pick.

What does the weekly workflow look like, step by step?

• Record (30 minutes): one question, answer in the first 30 seconds, show the work. Phone, tripod, window.
• Transcribe (automatic): the video goes to a transcription tool the moment it is uploaded.
• Generate (5 minutes): the transcript runs through five saved prompts, one per output, each written once with your voice rules in it.
• Review (30 minutes): you read all five drafts. Cut anything you would not say. Fix any number. This step is not optional and it is where the engine earns trust.
• Schedule (15 minutes): email on publish day, posts across the week, article the next morning, sequence attached to the lead magnet.

Ninety minutes, most weeks. The first month takes longer because the prompts are being tuned; by week five they rarely need editing.

How do you keep it from sounding like AI?

Write the voice rules once and put them in every prompt. Mine include: short sentences, no exclamation points, no "in today's fast paced world," no bullet lists of adjectives, one example per point, and no em dashes. Then the second rule, which matters more: the model only works from the transcript. If a fact is not in what you said, it does not appear in the draft. AI invents confidently; the transcript is the fence.

The test before anything ships: read the draft aloud. If a sentence is not something you would say to a client across a table, cut it.

What should AI never write in this engine?

The source video, the numbers, and anything a client will treat as advice. The video has to be you, because it is the thing that builds trust. The numbers have to come from your records. And a reply to a client who asks a real question gets a human, not a sequence. I keep the same rule in AI automation for coaches: AI runs the work between the calls, never the call.

What is the engine actually for?

Staying in front of the list between videos. Clients come from the video being found and the lead magnet being completed; the system in how a business turns YouTube views into leads covers that. The engine makes sure the person who completed the lead magnet hears from you five times that month in your own voice, without you writing five things. That is the difference between a list and a pipeline.

What do the five saved prompts actually say?

Each prompt has the same three parts: the voice rules, the fence, and the shape of the output. Here is the structure, which you adapt once and reuse every week.

• Voice rules: "Write as Aaron speaks to a client across a table. Short sentences. No exclamation points. No em dashes. No phrases like 'in today's world' or 'game changer.' One example per point, taken from the transcript."
• The fence: "Use only what is in this transcript. If a fact, number or name is not in the transcript, leave it out. Do not add statistics."
• The shape, email: "150 to 250 words. First line is the answer. One example. One sentence linking to the video. One sentence offering the lead magnet by name."
• The shape, posts: "Three posts under 60 words each. Each opens with a sentence the speaker said. No hashtags."
• The shape, article: "Restructure under four to six H2 headings phrased as questions. First paragraph under each heading answers it in under 40 words. Keep every example. End with a short FAQ drawn only from questions the transcript answers."

The fence line is the one most people skip, and it is the one that keeps the engine honest.

What does one video become, concretely?

A 14 minute video answering "how many videos before YouTube sends a lead" became: an email with the 12 to 15 upload threshold as the first line and Scott's 90 day story as the example; three posts, one of them the sentence "most people quit at video six"; a 1,400 word article under six question headings with the video embedded; a three email sequence for people who took the scorecard that week, each unpacking one reason channels stall; and next week's question, pulled from a comment asking whether Shorts count toward the threshold. One recording, 90 minutes of review and scheduling, five weeks of staying in front of the list.

How do you know the engine is working?

Not by output volume. Three numbers, monthly: email replies and clicks to the video, lead magnet completions from people already on the list, and conversations booked by people who were on the list more than 30 days. The engine's job is the third number. If it is flat while the list grows, the sequence is not making an offer, or the offer is unclear. Fix the last email in the sequence first; it should say plainly what to do next and why now.

What tools do you need, and in what order?

A phone, a transcription tool, a current AI model with saved prompts, your email platform and your website. Build in this order: the email first, because it pays fastest; then the article, because it compounds in search; then the posts; then the sequence; then the question list. Do not buy a platform to run this until you have done it by hand for a month and know which steps you actually keep. The operating idea behind all of it is in Systems Over Hustle: build the thing once, then run it weekly.