Audience-Specific AI Slide Deck Workflow
Convert a source corpus into a tailored teaching deck for one team or task
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 6
- Confidence
- 97%
Start with a NotebookLM corpus containing the material the audience needs to understand, then define the exact team, customer, role, or task the presentation should serve. Prompt NotebookLM to build a slide deck that contextualizes the source material for that audience instead of merely summarizing it. The system can perform several stages at once: review the sources, identify themes, create an outline, write slide copy, and generate supporting illustrations or screenshot-like visuals. Review each slide for factual grounding, instructional usefulness, layout, text rendering, and visual fidelity. The generated inventory of ideas and images may be stronger than the typography or sequencing, so assess those dimensions separately. Iterate until a viewer can learn something relevant and apply it to their work. The core advantage is repeatable personalization: the same corpus can support different decks for individual teams or tasks.
Origin
Extracted from Marketing Against The Grain, where the host generates and critiques a Gemini 3 deck tailored to the HubSpot marketing team.
Core principles
- 01Tailor every deck to a specific team, audience, or task
- 02Ground slides in source material the audience can trust
- 03Use AI to compress research, outlining, and initial design
- 04Review generated visuals, claims, structure, and instructional value
- 05Optimize for useful understanding rather than instant presentation readiness
How to run it
- 1
Assemble the knowledge base
Add the podcasts, videos, documents, websites, or other sources that should ground the presentation.
Pro tip Use material that contains concrete examples and use cases, not only general descriptions.
Watch out The deck can amplify inaccuracies or gaps present in the corpus.
- 2
Specify the audience and outcome
Name the exact team, customer, role, or task and state what the audience should understand or be able to do afterward.
Pro tip Use the audience's actual priorities and vocabulary in the prompt.
Watch out A generic audience request tends to produce a generic presentation.
- 3
Generate the tailored deck
Ask NotebookLM to synthesize the sources into a structured slide deck with appropriate examples and visuals.
Pro tip Request contextualization for the audience rather than a simple source summary.
Watch out Generated screenshots or depictions may look plausible without matching the originals exactly.
- 4
Audit every slide
Check the deck's agenda, claims, quotations, text rendering, examples, layouts, and closing. Confirm that each slide contributes to the intended learning outcome.
Pro tip Separate the review of idea inventory, copy, layout, and imagery so one strong dimension does not mask another weak one.
Watch out A single rendering error or fabricated-looking visual can reduce trust in the entire deck.
- 5
Convert insights into actions
Strengthen the presentation so the audience leaves with practical ideas they can apply to their role or current task.
Pro tip Replace abstract capability statements with audience-specific opportunities and examples.
Watch out A visually impressive deck can still fail if it does not change what the audience understands or does.
- 6
Reuse the corpus for other audiences
Generate separate variants for different teams, customers, or tasks instead of forcing one universal deck to serve everyone.
Pro tip Keep the sources stable while changing the audience context to produce efficient personalization.
Watch out Do not assume that terminology, examples, and priorities transfer unchanged between audiences.
In the wild
The host asked NotebookLM to create a deck specifically for the HubSpot marketing team. It synthesized podcast material into marketing superpowers, HubSpot opportunities, video-analysis examples, generated visuals, and ideas such as interactive lead magnets, experiential marketing, rapid prototyping, HubSpot Academy improvements, and changes to inbound content. The host found the idea and image inventory strong but identified layout, text, closing, and rendering problems.
→ A 15-slide audience-specific draft was generated in minutes and appeared capable of becoming educationally useful after roughly ten more minutes of iteration.
A company could maintain a notebook containing competitor websites, product announcements, customer commentary, and internal notes. It could then generate a sales-specific deck that explains competitive differences, likely objections, and account-relevant talking points rather than presenting a generic market summary.
→ Salespeople receive a presentation framed around their decisions and conversations without repeating the underlying research process.
Common mistakes
Creating one deck for everyone
A universal presentation sacrifices relevance. Generate separate versions around the priorities and tasks of each audience.
Trusting synthetic screenshots
Generated screenshots and reconstructed scenes may not faithfully reproduce the source. Verify or replace visuals that could be mistaken for documentary evidence.
Confusing completeness with readiness
A full 15-slide deck can still have a poor agenda, weak closing, layout errors, or little practical value. Review the complete communication experience.
Is it for you?
Best for
It is best for internal enablement, customer education, sales presentations, competitive intelligence, and role-specific training.
Not ideal for
It is not ideal for high-stakes external presentations that require exact branding, editable layouts, or guaranteed fidelity in generated screenshots and illustrations.
From the transcript
“So I asked Gemini to build a slide deck. I give it a prompt to build the following slide deck specifically for the HubSpot marketing…”
“If I said, hey, go listen to my podcast, that's like 30 minutes right there.”
“And what's really cool is I can actually have it build these decks specific for individual teams or individual tasks people are working on, which…”
From the episode
This New Google AI Feature Replaces 10 Hours of Work