One Source, Ten Pieces: The Guide to AI Marketing Automation

Contents

1. The problem isn't writing, it's volume
2. The method: one source, ten pieces
3. What to give the AI, and what never
4. Brand voice: one page that saves a hundred corrections
5. AI images and video: labels on social networks
6. The human review step: seven questions per piece
7. Example: a dental clinic, one week
8. From content to enquiries: what happens after someone reads
9. When to do it yourself and when to hand it over
10. Frequently Asked Questions

The problem isn't writing, it's volume

A small business that wants to be present today needs an Instagram post, a Facebook version, a short video, a Google Business Profile update, a newsletter and every so often a website article. Nobody has time to write all of that from scratch every week, so most businesses publish something once every few weeks and give up.

Artificial intelligence doesn't fix this by "writing for you". It fixes it when you work with one method: one source with real substance, which the AI breaks into ten pieces, and a person checks each one before it goes out. In this guide we go through the method, what you may and may not hand to the AI, and the review step without which none of this is worth much.

What is AI-based content automation?

Content automation is not a machine that invents posts. It is a process in which you bring the knowledge and the latest language models (Claude, Gemini, ChatGPT and the like) do the mechanical work: shortening, adapting to the format, writing several headline versions, translating and arranging everything on a calendar. Your part is the source and the approval. The AI's part is the multiplication.

The method: one source, ten pieces

Step 1: pick one source a week

A good source is something that holds knowledge only you have. For example:

• A three- or four-minute video in which you answer a question customers ask all the time.
• A voice note on the way home: "what a customer taught me this week".
• A page of notes about a new service.
• An existing website article that can be refreshed.

A weak source (a generic topic like "five marketing tips") produces ten weak pieces. The AI multiplies what you give it, for better and for worse.

Step 2: write a five-line brief

Before asking for anything, write down: who it is for, what the one message is, what action you want the reader to take, which facts may be mentioned, and what must not be said. Those five lines save three rounds of corrections.

Step 3: break it into ten pieces

One four-minute video usually yields:

1. A website article that answers the question in depth.
2. A six-to-eight-slide carousel.
3. Three short video scripts, each with a different hook for the first two seconds.
4. An Instagram post.
5. A longer Facebook version, or a more professional LinkedIn one.
6. A Google Business Profile update.
7. A paragraph for the email newsletter.
8. A short message for the WhatsApp broadcast list, only for people who opted in.
9. Three versions of ad copy, to test which message brings enquiries.
10. A ready answer to this question, to use when a customer asks it in a message.

Step 4: a person checks every piece

More on this below. This is the step you must never skip.

Step 5: spread it over two weeks and measure

Ten pieces don't go out on one day. They spread over a week or two, and each carries a tagged link or a WhatsApp button with a pre-filled message, so you know which format started a conversation. Want to understand how to measure whether all this pays off? Read about ROI measurement in AI solutions.

What to give the AI, and what never

What to give it:

• The source itself: the transcript of the video, the recording or the notes.
• A fact sheet you approved: services, hours, address, what is included and what isn't.
• A short brand voice guide (see the next section).
• Five earlier posts you liked, as examples.
• The list of questions customers ask, in their own words.

What the AI must never invent:

• Facts and numbers. Statistics, success rates, "studies show". A language model writes with confidence even when it is wrong.
• Prices, terms and promises. A price is written by a person or links to your pricing page. An AI that writes "first visit free" when there is no such offer can cost you money.
• Customer stories, quotes and reviews. A customer story is published only if it is real and only with written consent. An invented review is deception.
• AI-made "before and after" or "customer" photos presented as real.
• Medical, legal or financial claims not verified by a professional.

And one more rule: don't paste customers' personal details (names, phone numbers, medical information) into a public AI tool. What goes in there leaves your control.

Brand voice: one page that saves a hundred corrections

The reason AI content sounds "like AI" is that nobody told it how you sound. A one-page brand voice guide includes:

• Three words that describe you, for example: direct, warm, no jargon.
• How you address the customer: second person, plain "you", no "dear valued customers".
• Words you use and words you never use, such as "revolutionary", "game-changing", "in today's world".
• Sentence length and emoji policy.
• One good and one bad example of the same post.

Familiar signs of unedited text: a generic opening you could paste onto any business, three adjectives in a row, an exclamation mark on every line, and a "so what are you waiting for?" ending. When you see these, the piece goes back for editing.

AI images and video: labels on social networks

The big networks now require transparency about AI-made content, and detect some of it themselves:

• Meta (Instagram, Facebook, Threads) adds an "AI info" label to content it detects as made or edited with AI, or that you marked yourself.
• TikTok requires you to label realistic AI content, and labels content that arrives with provenance data (Content Credentials) automatically.
• YouTube requires disclosure when a video contains realistic content that was synthetically created or altered.

The practical rule: an AI illustration or background that explains an idea is fine, and you label it when the platform asks. People, results and places shown as real must be real. A simple phone photo of your own team beats a "happy customer" who doesn't exist.

The human review step: seven questions per piece

Before a piece goes out, someone who knows the business answers seven questions:

1. Is every fact correct, and does every number come from a source we can show?
2. Is there a price, promise or term we haven't approved?
3. Does it sound like us, or like any other business?
4. Is there a customer quote, story or photo, and do we have consent for it?
5. Is any AI-made image or video labelled the way the platform requires?
6. Is the link correct and tagged, and is the call to action clear?
7. For mailings: does it go only to people who opted in, with a way to unsubscribe?

It takes a minute or two per piece, and it is the difference between content that builds trust and content that embarrasses.

Example: a dental clinic, one week

The hygienist films four minutes on her phone about "how often you really need a scale and clean". Out of the video come a website article, a "five signs it's time" carousel, three short videos with voiceover and captions, a Facebook post, a Google Business Profile update, a paragraph in the monthly newsletter and a ready answer to the question that keeps coming up on WhatsApp. The AI prepares the drafts, a person edits, the dentist approves the professional side and the clinic owner approves the week. In practice the owner spent four minutes filming and twenty minutes approving.

From content to enquiries: what happens after someone reads

Good content makes people ask. The question is what happens the moment they ask, including evenings and weekends:

1. A potential customer reads an article on the website or sees a video on social media.
2. At the end there is a button to send a WhatsApp message or open the website chat.
3. The conversation starts immediately with an AI-based sales assistant, which answers in any language (Hebrew, Russian, English and others) based on what the customer writes.
4. The assistant asks the right questions, logs the enquiry with its source in the CRM, and hands over to a person when needed or books a meeting.

That way content and response work together: what brought the customer is recorded, and you can see which piece actually brought enquiries.

When to do it yourself and when to hand it over

Doing it yourself works when you have a good source every week, someone who enjoys editing, and an hour or two for review and publishing. Everything described here can be done with widely available AI tools and a simple calendar.

It's worth handing over when you want a steady pace on several networks, videos every week, a newsletter that goes out on time and measurement tied to enquiries. That is how content creation at WhaleBiz works: one topic becomes posts, carousels, short videos, an article and a newsletter by email and WhatsApp, social publishing runs to a calendar, and you approve the week before anything goes live. Want to talk about it? Contact the experts at WhaleBiz.

Frequently Asked Questions

What is AI-based content automation?

It is a process in which you supply a source with real knowledge, such as a short video, a recording or notes, and the latest language models break it into pieces for each channel: posts, a carousel, video scripts, a newsletter, an article and ad copy. A person checks and edits every piece, and the business owner approves before publishing.

What must the AI never write on its own?

Unverified facts and numbers, prices, terms and promises, customer stories, quotes and reviews, images presented as real, and medical, legal or financial claims. A person writes or approves all of these. In addition, never put customers' personal details into a public AI tool.

Do you need to label AI-made content on social networks?

Yes, in most cases. Meta adds an "AI info" label to content detected or marked as AI-made, TikTok requires labelling realistic AI content, and YouTube requires disclosure of realistic synthetic content. The simple rule: an illustration is fine, but people and results shown as real must be real.

Does content written with AI hurt your Google rankings?

Google judges content by how helpful it is to the reader, not by how it was produced, but mass-producing thin pages just to catch searches violates its policies, so an article built from your real knowledge and edited by a person works well, while a generic article the AI wrote alone and published unchecked can do harm.