Rally

Ever wished you could run your ideas past a diverse focus group before sharing them with the world? What if I told you that you could get feedback from dozens (or even hundreds) of different perspectives in minutes rather than weeks—and for pennies rather than thousands of dollars?

Welcome to the world of Personas of Thought—a prompt engineering technique I developed that transforms how we use AI for creative problem-solving and decision-making.

The Problem: One AI, One Average Opinion

Standard AI responses often feel like corporate press releases—safe, balanced, and utterly devoid of strong opinions. When you ask ChatGPT or Claude a question like "What do you think of this headline?" you're essentially getting the average opinion of the internet, filtered through layers of helpfulness training that prioritize politeness over honesty.

This happens because large language models are:

Trained on massive datasets representing diverse viewpoints
Fine-tuned to give "helpful" (read: inoffensive) responses
Designed to synthesize information rather than take strong positions

But real humans don't work this way. We have strong opinions. We disagree. We see things through the unique lens of our experiences and biases. And those distinct perspectives are precisely what makes human feedback so valuable.

The Solution: From Individual Thought to Crowd Wisdom

What if, instead of asking an AI for a single response, we asked it to simulate multiple people with different backgrounds, experiences, and perspectives—and then synthesized their collective wisdom?

That's exactly what Personas of Thought does.

Personas of Thought prompt

As I explained on the Marketing Against The Grain podcast:

"Rather than asking ChatGPT or Claude to give you an answer, which gives you the stock answer—kind of the average of the internet—what I found is if you ask it to think of a bunch of personas to roleplay as first, then you get the individual responses from those personas, and then you combine that together into a final answer."

The result? Richer, more nuanced, and often more accurate insights than a single AI response could ever provide.

The Basic Template: Get Started in Minutes

Here's the simplest version of the Personas of Thought prompt template that you can use right away:

First, give me [NUMBER] demographic personas (regular people) who [RELEVANT CRITERIA]. Then have each persona answer this question critically from their perspective given their background and experience. [YOUR QUESTION] Finally, combine these responses into a single paragraph response as if these people had collaborated in writing a joint anonymous answer. Do not name any of the people in the combined response.

Let's look at an example:

First, give me 10 demographic personas (regular people) who would be potential buyers of HubSpot's CRM software. Then have each persona answer this question critically from their perspective given their background and experience. Which headline do you prefer: "Grow Better with HubSpot" or "Grow Without the Guesswork"? Finally, combine these responses into a single paragraph response as if these people had collaborated in writing a joint anonymous answer. Do not name any of the people in the combined response.

Real-World Applications: Where Personas of Thought Shines

1. Market Research Without the Cost

Traditional market research is expensive and slow. Focus groups typically cost $5,000-10,000 and take weeks to organize. With Personas of Thought, you can run dozens of simulated focus groups in an afternoon—testing messaging, product ideas, or content before investing real money.

As I wrote in How I Made AI Think Like a Focus Group:

"Being able to ask the AI versions of my customers allows me to move faster to gut-check ideas and iterate on them without bothering my real ones."

2. Content Testing and Optimization

One of the most powerful applications is headline testing. In my article I Created a Hacker News Simulator to Reverse-engineer What Goes Viral, I describe how I created AI personas based on real Hacker News users to predict which headlines would perform best:

"Most people don't realize that when they leave comments on social media or online message boards like Hacker News, they talk a lot about themselves... All of this information can be extracted by an LLM and turned into a user profile."

The results can be surprising—often contradicting what we personally prefer:

"I don't even really like it, but my favorite one lost. I was OK with it. It doesn't matter what I like—I'm not the audience. Marketers and founders often get that wrong."

3. Product Feedback and Development

When developing new products or features, getting diverse perspectives early can save months of development time. Personas of Thought lets you test concepts with simulated users representing your target market:

First, give me 8 demographic personas who are frequent online shoppers aged 25-40. Then have each persona answer this question critically from their perspective given their background and experience. What features would make you try a new shopping app instead of using Amazon? Finally, combine these responses into a single paragraph response as if these people had collaborated in writing a joint anonymous answer. Do not name any of the people in the combined response.

Advanced Techniques: Taking It Further

Enhancing Persona Quality

For more realistic personas, provide specific criteria:

First, give me 5 detailed personas of marketing professionals with at least 5 years of experience working at SaaS companies with 50-200 employees. Include their job title, company size, key challenges, and career aspirations.

Using External Data

You can also ground your personas in real data:

Based on the following customer interview transcripts, create 6 personas that represent different segments of our user base. Then have each persona answer...

Making the AI Less Polite (More Useful)

As I discovered in my testing, sometimes you want the AI to be brutally honest:

"AI models are trained to be polite, but polite answers aren't useful when you're trying to build a business. Running your idea past ChatGPT is a waste of time for the same reason you shouldn't get startup feedback from your mom: They're hardwired to protect your feelings."

The solution? Provide examples of harsh but constructive criticism in your prompt:

I want brutally honest feedback. Don't hold back to spare my feelings. Examples of the tone I'm looking for: - "This landing page is confusing. I have no idea what you're selling." - "Your pricing structure makes zero sense to me. Why would I pay more for fewer features?" - "The headline is generic marketing jargon that tells me nothing about the actual value."

Specialized Templates for Different Use Cases

For Copywriting Testing:

First, give me 10 demographic personas who match our target audience of [AUDIENCE DESCRIPTION]. Then have each persona rate the following copy options on a scale of 1-10 and explain their rating: Option A: [COPY VERSION A] Option B: [COPY VERSION B] Finally, combine these responses into a recommendation, including which version performed better and why. Include any suggestions for improvement.

For Product Idea Validation:

First, give me 12 demographic personas who have the problem that [PROBLEM DESCRIPTION]. Then have each persona evaluate this product idea and answer these questions: 1. Would you pay for this solution? Why or why not? 2. What's the most valuable aspect of this idea to you? 3. What's missing or concerning about this idea? [PRODUCT IDEA DESCRIPTION] Finally, summarize the overall sentiment toward the idea, key strengths, and areas for improvement.

For Identifying Personas:

Sometimes you don't know who your audience is. Personas of Thought can help with that too:

First, identify 8 demographic personas who would be most interested in [PRODUCT/CONTENT TYPE]. For each persona, explain: 1. Why they would be interested 2. What specific pain points this addresses for them 3. How they would typically discover this type of [PRODUCT/CONTENT] Finally, prioritize these personas from most valuable to least valuable for our business, with a brief explanation.

Taking It to the Next Level: Rally

These techniques work with any AI, but as I started using them more, I realized we needed a dedicated tool. That's why I built Rally — a platform that makes creating and managing AI personas effortless.

With Rally, you can:

Build persistent personas for repeated use
Run queries against hundreds of personas simultaneously
Import real audience data to ground your personas
Analyze aggregate results across different segments

As I wrote in my latest article:

"When I'm testing headlines, I can see which one my audience picked most often, learn why, and decide whether I need to further refine what I've written."

Why This Matters: Beyond the Hype

AI agents aren't perfect predictors of human behavior, but they're surprisingly good—and getting better. Research suggests that AI agents can predict human survey responses with up to 85% accuracy. For context, if you ask a human the same question two weeks apart, they only give the same answer 81% of the time.

This isn't about replacing human research—it's about augmenting it. Use Personas of Thought to:

Quickly eliminate bad ideas without wasting resources
Refine concepts before testing with real humans
Discover blind spots in your thinking
Generate more diverse perspectives than you might get from your immediate network

Getting Started Today

You don't need any special tools to start using Personas of Thought—just copy the templates from this guide and paste them into your favorite AI assistant.

For those wanting to go deeper, I cover this technique extensively in my Prompt Engineering for Generative AI course, and I'm continuously developing new applications at Rally.

The future of decision-making isn't a single AI giving you the "right" answer—it's leveraging the collective intelligence of many AI personas, each with their own perspective, to help you see possibilities you might have missed.

Because sometimes the most valuable insight isn't the average of all opinions—it's the unexpected viewpoint that makes you think differently.

Ready to try Personas of Thought with your own projects? Sign up for early access to Rally and join our community of forward-thinking AI experimenters.