Source-First Deck Generation
Give AI the raw source and instructions, then edit its complete first draft
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 98%
Instead of asking one model to write a slide outline and handing that outline to a presentation tool, provide the underlying source directly to the deck system. The source might be a survey, Slack channel, drive file, transcript, or automation payload. Add specific instructions describing the audience, objective, and desired output, then allow the system to perform synthesis, outlining, writing, image selection, and layout as one integrated first pass. A person subsequently edits the complete artifact. This preserves more source context and removes an artificial intermediate deliverable. The accompanying decision rule is to reconsider text by default: whenever a large block of text is about to be sent, ask whether a visual or mixed-media presentation would make it easier to consume.
Origin
The hosts contrasted their outline-first habit with Kristen Frackett’s direct use of raw data sources and connectors to generate complete first drafts. Extracted from Marketing Against The Grain.
Core principles
- 01Start from evidence rather than a manually written outline
- 02Let the system perform synthesis, outlining, and visualization together
- 03Use instructions to define the output’s purpose
- 04Edit the generated artifact instead of assembling it slide by slide
- 05Prefer visual form when it improves comprehension
How to run it
- 1
Choose the source
Identify the dataset, document, channel, transcript, or connected system that contains the authoritative raw material.
Pro tip Use the closest available source rather than a summary of a summary.
Watch out Unfiltered sources may contain confidential, irrelevant, or low-quality material.
- 2
Define the transformation
State who the deck is for, what it should help them understand or decide, and any required structure.
Pro tip Describe the communication outcome rather than dictating every slide title.
Watch out Raw data without an objective produces unfocused output.
- 3
Generate directly
Give the source and instructions to the presentation system and let it create the first outline, copy, visuals, and layout.
Pro tip Preserve source references so claims can be checked later.
Watch out Do not insert an outline-only step by habit when it adds no control or value.
- 4
Review the narrative
Check whether the generated sequence supports the audience’s decision and accurately represents the source.
Pro tip Review omissions as carefully as incorrect statements.
Watch out An attractive structure can still emphasize the wrong evidence.
- 5
Edit the artifact
Revise claims, hierarchy, visuals, and pacing inside an interface where every element remains editable.
Pro tip Spend human effort on judgment-heavy changes rather than mechanical formatting.
Watch out Pure-image outputs can be harder to correct if the generation system lacks granular editing.
- 6
Apply the visual-format test
Before sending a long text output, ask whether charts, diagrams, cards, or a mixed-media deck would improve comprehension.
Pro tip Use text when precision or searchability matters more than visual storytelling.
Watch out Slides are not automatically superior; choose them for the audience and task.
In the wild
A community manager supplies the introduction channel and asks for a deck summarizing the week’s new members. The system determines the initial categories, writes the explanations, selects visuals, and builds the pages before the manager reviews them.
→ The source moves directly to a complete editable artifact without a manually prepared outline.
A marketing leader provides the current strategy document and project-management export, asks for a deck explaining what the team is doing and why it should work, and then edits the generated story for executive review.
→ Operational source material becomes a stakeholder-ready narrative with less manual assembly.
Common mistakes
Writing the outline by reflex
A separate outline step can discard source detail and duplicate work when the presentation system can reason over the original material.
Using raw data without instructions
Direct source access still requires a clear audience, purpose, and requested transformation.
Choosing slides for everything
The visual-format test is a decision rule, not a mandate; some outputs remain better as precise searchable text.
Is it for you?
Best for
Knowledge workers turning transcripts, surveys, channels, documents, or connected records into an editable visual first draft.
Not ideal for
Tasks where the source is unreliable, the output format is tightly regulated, or direct access to raw data creates unacceptable risk.
From the transcript
“Pick a data source, whether it be survey, whether it be a Slack channel, whether it be a G drive file, a zap, what have…”
“And then you can go back and edit it from there.”
“anytime you have an instinct to just send like giant blocks of text to anybody, you should just put it in a visual format now…”
From the episode
This AI Workflow Turns Every Sales Call Into a Custom Deck