MMarketing Against The Grain
← All frameworks
Innovation

AI Ideation, Human Creation

Let AI discover unfamiliar angles, then use human experience to create the work.

Difficulty
Easy
Time to result
~days to results
Steps
6
Confidence
98%

AI Ideation, Human Creation divides content production according to comparative advantage. AI performs the high-breadth work: searching many sources, clustering discussion, identifying overlooked ideas, assembling evidence, suggesting hooks, and preparing a structured brief. The human performs the high-authenticity work: deciding what is true and worth saying, connecting the idea to lived experience, selecting stories, exercising editorial judgment, and writing or recording in a recognizable voice. The handoff is explicit—the model should not be asked to turn its own brief directly into publishable content. This separation reduces convergence because the creator begins with wider collective information but finishes with experiences and interpretations that other users cannot reproduce. The method also adds a verification checkpoint between machine research and public communication, helping prevent unsupported claims from flowing automatically into polished copy.

Origin

Extracted from Marketing Against The Grain, where the hosts argue that AI's greater content advantage lies in ideation and research, while the creator should retain final authorship.

Core principles

  • 01Use AI where breadth of research and synthesis create leverage.
  • 02Keep final authorship where voice, judgment, and lived experience matter.
  • 03Separate idea discovery from prose generation.
  • 04Treat an AI brief as scaffolding rather than publishable content.
  • 05Enrich researched ideas with distinctive human stories.

How to run it

  1. 1

    Give AI the discovery task

    Ask the system to research the topic, map current discussion, and identify useful angles the creator may not already know.

    Pro tip Specify the audience, source types, and desired evidence.

    Watch out Do not begin by requesting a finished post.

  2. 2

    Request a decision-ready brief

    Have AI organize the selected idea into hooks, arguments, counterarguments, statistics, stories, and possible conclusions.

    Pro tip Ask it to distinguish verified evidence from hypotheses or illustrative material.

    Watch out A polished brief can still contain fabricated or weakly sourced claims.

  3. 3

    Verify and choose

    Check the sources and decide whether the proposed position matches the creator's beliefs, expertise, and audience obligations.

    Pro tip Discard an exciting angle if you cannot defend it without relying on the model's authority.

    Watch out Do not confuse machine confidence with factual confidence.

  4. 4

    Add proprietary experience

    Connect the brief to firsthand events, observations, customer conversations, or original examples that only the creator can credibly provide.

    Pro tip Use the AI research to trigger memory rather than asking AI to fabricate personal stories.

    Watch out Never present an invented anecdote as lived experience.

  5. 5

    Create the final work

    Write, speak, or design the content independently from the brief, using personal language and editorial judgment.

    Pro tip Keep the brief nearby for factual structure but avoid copying its phrasing mechanically.

    Watch out Turning the same model loose on final drafting can reintroduce generic style and consensus language.

  6. 6

    Audit the handoff

    Review the finished piece for accurate claims, genuine voice, useful differentiation, and traces of generic AI prose.

    Pro tip Ask whether the content contains at least one insight or story that another user could not obtain from the same prompt.

    Watch out Do not publish until evidence and attribution have been checked.

In the wild

Research brief to podcast episode

The demonstrated agent researches GPT 5.5 commentary, identifies a differentiated CMO-governance angle, and supplies hooks, statistics, counterarguments, a story concept, and closing lines. The hosts do not treat that output as the finished episode. They plan to use the research and angle as the basis for their own subsequent discussion.

AI accelerates discovery while the hosts retain the interpretation, delivery, and relationship with the audience.

Founder newsletter workflow

A founder asks AI to map competing explanations for declining trial conversions and produce a sourced brief. After verifying the evidence, she connects the strongest idea to customer calls and an experiment her team ran. She writes the newsletter herself, including the decision she made and what she learned.

The newsletter gains research depth without losing the founder's credibility or voice.

Common mistakes

Automating the entire pipeline

Moving directly from machine research to machine-written publication removes the human judgment and experience that make the result distinctive.

Treating the brief as verified

AI can package weak or fabricated evidence persuasively. Every important source, statistic, and causal claim still requires checking.

Adding generic human polish

Minor edits to an AI draft do not equal human creation. The creator should shape the argument, examples, stories, and language from the brief upward.

Is it for you?

Best for

It is best for experts and creators who have a real voice or experience but need broader research and stronger starting ideas.

Not ideal for

It is not ideal for creators seeking fully automated publishing or for topics where they cannot independently evaluate the AI's claims.

From the transcript

It's not writing the content for you, it's ideation. Yeah. That's what you're missing. That's where the power is. It's the ideation. You should write…

Kieran Flanagan and Kip Bodnar · 10:00

I think you're making the key point that it's about the ideation versus the creation. And AI is actually better at ideation than even it…

Kip Bodnar · 11:00

all of our ideas come from like our own experiences, and AI has all of our experiences together. So it's gonna find things that we…

Kip Bodnar · 10:30

From the episode

Use AI for Ideas, Not Content (Here’s How)