Online Learning · Build it

Build the Content Repurposing Engine — in an afternoon.

Workflow #1 from our 40-workflows menu. The highest-leverage AI workflow for anyone who publishes anything. One input becomes a week of content. Built once, runs forever.

From the AI Future Lab team 1,400 words 6 min read Updated May 2026

Why this one first

If you create anything at all — a blog, a newsletter, a podcast, a long internal memo, a recorded talk — you are leaving an enormous amount of value on the floor. You publish a piece once, on one platform, and stop. Meanwhile the same insight, rewritten for the platform it's living on next, could carry your week.

The Content Repurposing Engine fixes that. You publish your long-form piece as normal. Then a configured agent takes that input and produces a thread, a set of standalone posts, a newsletter section, and a short video script. Each one rewritten — not copy-pasted — for its destination. You review for an hour. You schedule them. Your week's content is done.

We recommend this workflow first to almost everyone who comes through the Work Lab, for one reason: it automates work you're already doing. That makes the gains visible inside a week. Most people who build it never go back to publishing once-and-stopping.

The shape of the workflow

One input · Four outputs · One critic pass

Input One long-form piece (blog, podcast transcript, memo, talk)
Repurpose Configured agent, with your context bundle attached
Out 01A thread (10–14 posts)
Out 023 standalone posts
Out 03Newsletter section (300 words)
Out 04Short video script (90 sec)

Three principles to hold while you build:

Build this in an afternoon

Five steps. Work through them in order. Don't skip step two — it's the one that determines whether the whole thing works.

1

Pick your long-form input

Choose the kind of thing you'll feed in week after week. A blog post, a podcast transcript, a long LinkedIn essay, the recording of a talk you give regularly. Pick what you actually produce — not what you wish you produced. The workflow's value compounds when it runs on a real cadence.

2

Write your context bundle

This is the part everyone wants to skip. Don't. The context bundle is a single document that contains: your voice rules (5–10 specific guidelines), your audience per platform (who you're talking to, what they care about, what bores them), 3–5 example outputs of each type you've personally written and approved, and your no-go list (words you don't use, claims you don't make). Without this, the workflow produces AI slop dressed as your work. With this, it produces a credible first draft.

3

Configure the agent

Set up a configured agent in your AI tool of choice. Attach the context bundle. Write the system instruction: "You take one long-form input and produce four outputs — a thread, 3 standalone posts, a newsletter section, and a short script. Each output rewritten for its platform, not copy-pasted. Use the context bundle for voice, audience, and constraints." Test it on a real piece. The first run will be rough. That's expected.

4

Add the critic step

Set up a second agent — or a second pass on the same agent — that reads the first agent's output and flags problems. Its instruction: "Read each output against the context bundle. Flag anything that sounds AI-generated, contradicts the voice rules, makes claims not in the source, or is generic." Run the critic over every output. Use its notes to refine. This is what stops your audience noticing.

5

Schedule it & ship

Only do this once the manual version is producing outputs you'd actually post. Then put the whole thing on a schedule: every time you publish your long-form piece, the engine runs and drops the four outputs in a draft folder. You review for an hour. You post — yourself, by hand, with the human button-press intact. The "while you sleep" part is the labour, not the posting.

What's actually hard about this

Three things break this workflow more than anything else, in our experience.

One. The context bundle gets skipped. Everybody wants to start with the agent. You can't. The output is exactly as good as the context. We've watched people spend a weekend on the agent setup and twenty minutes on the context bundle and then wonder why the workflow doesn't work. It's the context, not the model.

Two. The critic step gets skipped. "It looks fine, I'll just check the outputs myself." You won't. Or you will for two weeks and then stop. The critic agent runs every time without getting tired or distracted. Build it.

Three. It gets automated too early. Schedule the workflow only after the manual version has produced outputs you'd publish without hesitation. Automating a half-working process just produces half-working output faster, on a schedule, while you sleep — which is the worst version of all.

Get those three right and the workflow runs cleanly for as long as you keep feeding it long-form. Most people who build it properly say it pays for itself in the first week and stops feeling like AI somewhere around month two — it just feels like the way they publish now.

Two rules to lock in before you ship

Nothing posts without you pressing a button. The engine drafts. You publish. Always. The minute you let an agent post on its own behalf is the minute it embarrasses you on a Tuesday morning when you're not looking.

The output is your responsibility. The model drafts. Your name goes on the post. Treat every output like a first draft from a sharp junior — useful, fast, requires reading. People who treat AI outputs as finished work are the people who eventually get caught making things up.

Read next

The full 40-workflow menu — what else to build, in order.

Want help building ten of these?

Six weeks. Ten workflows. One cohort.

The Work Lab is the structured, paid cohort version of everything we publish online. You walk out with running automations, not screenshots. $999 + GST. Capped at twelve.