Jacob Zangel recently shared something that made me stop scrolling: "Here’s a set of prompts that do McKinsey-level audience research in under 10 minutes" and "Good marketing and sales systems start with deep audience insight." As someone who has spent far too many hours in slide decks and spreadsheets trying to extract those same insights, that line hit home.

In the rest of his viral LinkedIn post, Zangel explains that he "tested 7 different prompting approaches to break an audience down fast and go deep." Most didn’t do that well, but one did shockingly well. His promise is bold: a few minutes of structured input can now produce useful and deep insights that used to take teams weeks or months to collect.

In this article, I want to respond to and expand on what Jacob is really pointing at: a practical, repeatable way to use AI for McKinsey-level audience research in under 10 minutes.

Why audience research is ripe for an upgrade

Traditional audience research is slow, expensive, and often out of date by the time it lands in your inbox. You run surveys, pay for panels, interview customers, scrape forums, and then synthesize it all into decks that few people actually read.

Jacob’s post cuts straight through that:

"Use this to get to know your audience at a level previously not possible."

If you work in marketing, sales, or product, that line should make you sit up. Because almost everything you do depends on how well you understand three things about your audience:

What they care about right now
Where they spend their time and attention
How they talk about their problems, desires, and constraints

Most teams are guessing on at least one of those. Jacob’s workflow uses AI to systematically reduce that guesswork.

Inside Jacob Zangel’s 5-step AI research flow

Here’s how Jacob summarizes the process:

Step 1️⃣ Give a one-liner about your audience (high level).
Step 2️⃣ AI maps the "watering holes" where they actually hang out online.
Step 3️⃣ It scans those places and pulls the topics they talk about.
Step 4️⃣ Builds a psychographic snapshot of your ICP from that material.
Step 5️⃣ Runs a second pass to deepen and structure the analysis.

Let’s break down what each of those steps really means in practice.

Step 1: Define your audience in one sharp sentence

Most audience research projects fail at the first step because the target is fuzzy. Jacob starts with a single, clear line about who you care about. For example:

"US-based SaaS founders doing $1M–$10M ARR"
"Enterprise sales leaders at B2B cybersecurity companies"
"Freelance designers who want to move into productized services"

The point is not to be perfect; it’s to give the AI enough of a starting point to find real people and real conversations, not generic personas.

Step 2: Map the real "watering holes"

Next, AI identifies where these people actually spend their time online. Not where your brand wishes they hung out—but where they already are:

Subreddits and niche forums
X (Twitter) conversations and lists
Slack and Discord communities
Comment sections on specific blogs, newsletters, or YouTube channels

This is one of the most powerful shifts Jacob is highlighting. Instead of brainstorming channels in a meeting room, you ask AI to map them based on live, public data.

Step 3: Mine those spaces for real conversations

Once those watering holes are mapped, AI can scan the content to surface what people are actually talking about:

Recurring problems and complaints
Goals, aspirations, and fears
Tools and products they mention, love, or hate
Words and phrases they naturally use

This is where you stop thinking in abstract "pain points" and start seeing the messy, emotional language real humans use when they are frustrated, hopeful, or stuck.

Step 4: Build a psychographic snapshot of your ICP

From those raw conversations, Jacob’s workflow asks AI to build a structured psychographic profile of your ideal customer profile (ICP). That means going beyond demographics into questions like:

What do they believe about their work and their industry?
What do they feel they have tried already that "didn’t work"?
What risks feel unacceptable to them?
What tradeoffs are they secretly willing to make?

This is exactly the kind of thing a consulting firm might spend weeks extracting from stakeholder interviews and customer calls. Now it can be modeled in minutes—as long as you remember it is a model, not the final truth.

Step 5: Run a second pass to deepen and structure the analysis

Jacob mentions that the process does not stop with a first draft. A second AI pass is used to improve the structure, depth, and usefulness of the insights. For example, you can ask the model to:

Cluster themes into clear categories (problems, desires, objections)
Prioritize issues by urgency and frequency
Translate findings into messaging angles, offer ideas, or content pillars

This is where the output becomes something your team can actually use, not just an interesting wall of text.

From months to minutes: what changes

Jacob summarizes the result simply: "Output: insights you’d normally spend months collecting." That sounds like marketing copy, but it’s not far off if you think in terms of directionally correct insight.

With this kind of workflow, you can:

Prototype audience insights before you spend money on formal research
Brief your marketing or creative team with rich psychographic context
Train new sales reps with realistic scenarios, objections, and language
Generate better prompts for downstream tasks like ad copy, landing pages, or outbound sequences

As Jacob puts it:

"Use it to power other prompts, brief your team, or train your sales reps. Access to this quality of data in minutes, not hours, changes the game."

How to try a version of this yourself

You may not have Jacob’s exact prompts, but you can recreate the logic behind them. Here’s a simple way to start:

Write a one-sentence description of your audience.
Ask your AI tool: "List the top 15 online places where [audience] spend time, and explain why each matters."
Pick 3–5 of those watering holes and have the AI summarize the main themes, frustrations, and desires discussed there.
Ask for a psychographic profile based on those summaries, including beliefs, fears, motivations, and typical objections.
Run a second pass where you ask: "Turn this into a concise briefing for my marketing and sales teams, with practical recommendations."

You can run this entire loop in under 10 minutes once you’ve done it a couple of times.

Stay grounded: AI as hypothesis, not gospel

There is a risk here, and Jacob hints at it when he says: "The analysis you’ll see here will break your own beliefs about your audience." That’s powerful—but it’s also a reminder to treat AI’s output as a set of hypotheses to validate, not a replacement for real conversations.

The smartest way to use this workflow is:

Let AI generate structured, surprising insight.
Highlight what contradicts your current assumptions.
Take those contradictions into customer interviews, sales calls, and surveys.
Update your prompts and profiles based on what reality confirms or rejects.

In other words, AI compresses the exploration phase. It does not remove your responsibility to test what you learn.

Final thoughts

What I appreciate most about Jacob Zangel’s post is that it is not just another "AI will change everything" claim. It is a concrete example of how to plug AI into a critical but painful business process—audience research—and get 80% of the value in a fraction of the time.

If you adopt even a lightweight version of this approach, you’ll likely find what Jacob promises:

"This will give you more levers to pull when you try to help them. It might even fundamentally change how you approach marketing and sales."

And you can get there in under 10 minutes.

This blog post expands on a viral LinkedIn post by Jacob Zangel. View the original LinkedIn post →