Audience Guide Construction Pipeline
Turn scattered customer evidence into an actionable AI audience guide
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 99%
The pipeline begins by retrieving customer evidence from existing repositories instead of drafting a persona from memory. An AI retrieval tool synthesizes that material into a concise customer profile with traceable sources. A second prompt expands the profile into an operational audience guide covering reader characteristics, tone, language, content structure, preferred formats, attention hooks, trust signals, power words, and practical expectations. The guide is then tested against real content to expose missing or weak instructions. Rather than rebuilding the whole artifact, the team gathers targeted evidence for each unclear section and iterates line by line. The final guide becomes reusable context for generation, critique, rewriting, and team alignment.
Origin
Extracted from Marketing Against The Grain, where the hosts used Gemini to synthesize HubSpot research and Claude to create a detailed guide for the Growth Gabby audience.
Core principles
- 01Start with evidence already collected about customers
- 02Use retrieval tools to synthesize scattered source material
- 03Translate a customer profile into operational content rules
- 04Make audience guidance detailed enough to generate and critique work
- 05Iterate weak sections with targeted additional research
How to run it
- 1
Retrieve the evidence
Search internal research, persona decks, customer documents, calls, emails, and relevant collateral for material that describes the target audience.
Pro tip Use a retrieval system that can cite or link the underlying sources.
Watch out Do not let one easy-to-find document dominate the profile when broader evidence exists.
- 2
Synthesize the customer profile
Create a concise profile covering roles, demographics where relevant, goals, pain points, behaviors, characteristics, and buying context.
Pro tip Separate well-supported findings from tentative assumptions.
Watch out A fluent summary can conceal thin or conflicting source evidence.
- 3
Expand into content rules
Translate the profile into guidance on voice, language, structure, length, formats, hooks, content angles, trust signals, and practical takeaways.
Pro tip Use specific requirements, such as a two-to-three-sentence executive summary, where evidence supports them.
Watch out Avoid generic rules that could describe almost any professional audience.
- 4
Test against real content
Apply the guide to an existing asset and ask the AI to grade, explain, and rewrite passages according to the guide.
Pro tip Choose an asset with measurable claims and a clear intended audience.
Watch out Do not judge the guide solely by whether the rewritten prose sounds smoother.
- 5
Close evidence gaps
Review the guide line by line, identify unclear sections, collect focused data for each one, and revise the relevant section.
Pro tip Work in several evidence-and-revision cycles rather than one giant prompt.
Watch out Do not add specificity merely to make the guide appear comprehensive.
- 6
Package for reuse
Store the approved guide as a shared PDF, document, knowledge file, or application context that every contributor can access.
Pro tip Version the guide and record the evidence date.
Watch out An audience guide becomes stale as the product, market, and customer mix change.
In the wild
The team asks Gemini to profile HubSpot's ideal customer using research across Google Docs. It produces a sourced one-page customer profile, including material from an existing persona deck. Claude then converts that profile into a guide covering reader roles, voice, language, structure, formats, hooks, trust signals, and key phrases.
→ Scattered research becomes detailed context that can generate and critique audience-specific content.
Common mistakes
Inventing the audience from memory
Internal intuition can be useful, but it should not replace retrieval of customer research and direct evidence.
Stopping at a persona summary
A demographic or role description is not yet an operational guide for hooks, structure, proof, language, and formats.
Accepting the first version
The hosts explicitly propose reviewing each section and collecting targeted evidence wherever the profile remains unclear.
Is it for you?
Best for
Organizations with customer research spread across documents, calls, decks, emails, and other repositories.
Not ideal for
Teams with no reliable customer evidence and no access to representative users for validation.
From the transcript
“can you build a profile of HubSpot's ideal customer based on all of the research you can find about customer in my docs?”
“And so here we build a customer profile using Gemini and Google Docs.”
“let's go get a bunch of data around this topic and tell it to just apply everything to this particular part of the audience outline”
From the episode
We Built a Custom AI Content Grader in 5 Minutes (No Code)