Collaboration Council
Recruit contrasting AI personas to deliver candid creative feedback
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 99%
The Collaboration Council creates a small panel of AI reviewers whose tastes, priorities, and objections differ substantially. Rather than asking one generic assistant whether a draft is good, define several personas with strong points of view—for example, a demanding tastemaker, an artist, and a skeptic. Give every council member the same artifact and ask each to assess it through their own lens without softening the criticism. The value comes from variance: areas of agreement suggest clear weaknesses, while disagreement reveals genuine trade-offs rather than a single supposedly correct answer. The creator then decides which feedback supports the intended audience and objective. This is primarily an initial-feedback mechanism, not proof of customer demand, and its output should inform revision or determine what needs testing with actual users.
Origin
Jason Keith designed the prompt for website feedback and built its council around personas inspired by Steve Jobs, Maya Angelou, and Anthony Bourdain. He later applied it to ad creative and other use cases.
Core principles
- 01Broad perspective improves creative judgment
- 02Useful reviewers need distinct points of view
- 03Strong opinions reveal issues that polite feedback hides
- 04AI can supply a provisional council when human reviewers are unavailable
How to run it
- 1
Define the Review Target
Provide the council with the artifact, intended audience, objective, and any important constraints.
Pro tip Ask a specific question such as what damages trust or what prevents the page from converting.
Watch out A vague request for general feedback will produce less actionable criticism.
- 2
Assemble Contrasting Personas
Select multiple reviewers whose standards and instincts differ meaningfully. Define what each one feels deeply about and reacts against.
Pro tip Choose differences in worldview, not merely different job titles.
Watch out Personas with nearly identical priorities create the illusion of diversity.
- 3
Demand Strong Independent Critiques
Have every persona review the work independently and express a clear point of view.
Pro tip Explicitly ask reviewers not to equivocate or protect the creator's feelings.
Watch out Harshness is useful only when tied to concrete observations and the stated objective.
- 4
Triangulate the Feedback
Separate repeated concerns from persona-specific preferences and identify the trade-offs behind conflicting advice.
Pro tip Prioritize issues identified by several reviewers before reacting to isolated taste judgments.
Watch out Do not treat a fictional persona's authority as factual evidence.
- 5
Revise and Validate
Revise the artifact using the most relevant feedback, then seek real-user evidence where the decision warrants it.
Pro tip Run the council again after a major revision to see whether the same concerns remain.
Watch out Do not let the council replace feedback from the actual target audience.
In the wild
An agency owner submits a homepage to three AI personas: a product tastemaker focused on clarity, an artist focused on emotional resonance, and a skeptic focused on unsupported claims. All three object to the vague headline, while only the artist dislikes the restrained visual style. The owner rewrites the headline immediately but validates the design question with customers before changing it.
→ The owner finds both an obvious shared weakness and a subjective issue that requires audience evidence.
Common mistakes
Choosing Look-Alike Reviewers
A council composed of similar personas reproduces the narrow feedback range the framework is intended to correct.
Allowing Polite Equivocation
Generic encouragement and hedged criticism obscure the distinctive judgment each persona is supposed to provide.
Mistaking Simulation for Research
AI personas can expand perspective, but they cannot establish how actual customers will behave.
Is it for you?
Best for
It is best for early website, campaign, content, and ad drafts that need a broader range of reactions.
Not ideal for
It is not ideal as a substitute for real customer research, legal review, accessibility testing, or specialist validation.
From the transcript
“So the point of this collaboration council prompt is to give you multiple personas to give you feedback.”
“But like when I first used it, I was shocked at how well the feedback matched kind of what I knew needed to be fixed…”
“It's a great when you need that initial feedback from a broad audience.”
From the episode
I Tested 5 AI Prompts That Replace Weeks of Creative Work