How to Train AI to Think Like You

Want to use AI to scale your natural voice and professional expertise? Wondering what you need to make AI sound and reason like you instead of a generic robot?

In this article, you’ll discover a step-by-step framework to turn your everyday meeting transcripts into a personalized AI profile that replicates how you think, write, and operate.

This article was co-created by Max Bernstein and Michael Stelzner. For more about Max, scroll to the end of this article.

Why Use the Cognitive Fingerprint Interview to Train AI 

The most common advice for personalizing AI output is to have an AI interview you: answer questions about your communication style, describe your approach, and feed those responses back as context.

Max Bernstein says this method works reasonably well as a starting point, but it has a fundamental limitation. It only captures what a person can consciously articulate in a structured interview setting, and that’s a fraction of how an expert actually thinks.

The dominant narrative around AI right now frames it as a threat to knowledge workers: AI will commoditize jobs, expertise, and professional value. Max’s argument is that the way to counter this is not to work harder at describing yourself to AI, but to train it on how you actually think.

The result is what he calls a cognitive fingerprint, a rich context document built from real conversations that captures the decision logic, mental models, and reasoning patterns that make your expertise genuinely unique.

That work is grounded in an idea from cognitive science that predates AI by decades. Philosopher Michael Polanyi described it as tacit knowledge: the principle that experts know more than they can say:

A seasoned copywriter can look at a headline and know immediately that something is wrong, without being able to articulate why. A consultant walks into a meeting room and senses the dynamic is off before anyone speaks.

This kind of expertise operates largely below conscious awareness. It surfaces in unscripted moments, such as solving a real problem, coaching a client, thinking through a challenge out loud, but it doesn’t show up when someone sits down to describe themselves in an interview format.

The cognitive fingerprint methodology is built to capture exactly that type of unique expertise.

#1: Understand The Four Layers of Knowledge Inside Every Transcript

Max developed a four-layer framework for categorizing the types of thinking that appear in transcript data. Each layer represents a different depth of insight, and extracting all four is what separates a superficial AI profile from a genuine cognitive fingerprint.

Declarative Knowledge: This is the surface layer and covers what someone would say if asked what they do. It’s the job description version of expertise, the kind of answer that appears on LinkedIn profiles or in introductory bios. Most AI personalization attempts stop here.

Procedural Knowledge: This layer covers how someone actually executes what they do. It is the step-by-step sequences and processes they follow. Where declarative knowledge names the outcome, procedural knowledge captures the method. It’s the layer that surfaces when people extract SOPs from transcript data.

Conditional Knowledge: This is where the training begins to get personal. Conditional knowledge is the decision logic behind actions: the if-then reasoning that governs how someone responds in different situations. To clarify: conditional knowledge guides you on when and why to use certain processes, describing the rules or conditions for actions.

When a certain type of client shows up, a specific response pattern kicks in. Certain project characteristics trigger particular work sequences. These patterns feel automatic from the inside but result from accumulated judgment that no one else assembles in the same way. Max calls this decision DNA.

Metacognitive Knowledge: The deepest layer, and the most valuable for AI training, is how someone thinks about thinking — their mental models. It’s also the hardest to self-diagnose.

When Max has asked clients to describe their own mental models before running this process, and then shown them what their transcripts actually reveal, the two versions rarely match. The mental models people describe in a self-assessment aren’t the ones driving their actual behavior.

This gap is the reason an outside perspective matters. A coach who has spent years in weekly sessions with someone will eventually develop a working map of that person’s mental models through accumulated observation. AI can accomplish a version of that same work much faster, given sufficient transcript data.

#2: Collect the Right Transcripts

Questionnaire responses and stylistic preferences won’t work.

Transcripts from real conversations where you are actively problem-solving, advising, or reasoning through challenges in unscripted conditions are the raw material for this process.

The most useful transcripts come from interactions where expertise operates without active curation.

Client calls and coaching sessions are high-yield because the back-and-forth creates natural conditions for tacit knowledge to surface. Sales conversations capture how a person reads a room and adjusts in real time. Brainstorming sessions and team meetings reveal how ideas get generated and evaluated. Even solo voice recordings — captured while walking, driving, or processing ideas between tasks — contribute useful material.

What doesn’t work as well is anything heavily scripted. A prepared presentation or a talk delivered many times captures packaged knowledge, not live thinking.

More transcripts produce richer results, but a minimum of 3 to 5 transcripts from genuinely different situations is enough to get started. In fact, variety matters as much as volume so that your unique thought patterns emerge across contexts, not just within one type of conversation.

Pro Tip: Max recommends adding a brief context note at the top of each transcript file before processing. Something like “client coaching session” or “team brainstorm” helps the AI understand which mode you were in when it’s extracting patterns, and makes aggregate analysis across multiple files more accurate.

Suggested Transcription Tools

For virtual meetings, tools like Google Meet, Zoom, and Fathom automatically generate transcripts. Max recommends Granola for its customization depth: it runs as an audio-only background process rather than a visible meeting bot, offers preset and custom output templates, and integrates with tools like Notion and various AI platforms, as well as shared team workspaces and prompt libraries.

For in-person conversations, wearable recording devices like Plaud clip to a shirt or attach to the back of a phone.

For individual thinking sessions, a voice-to-text tool like Wispr Flow lets someone speak context and ideas directly into an AI session rather than type them, and Max notes that spoken prompts consistently produce better AI output than typed ones.

#3: Prompt AI to Extract and Build Your Cognitive Fingerprint

When a set of labeled transcripts is uploaded to a ChatGPT project, a Claude project, or Gemini Gem, the extraction process begins with a prompt that asks the AI to analyze each transcript across all four knowledge layers simultaneously.

He asks the AI to identify specific knowledge instances across the transcript that correspond to each layer: declarative statements, procedural sequences, conditional decision logic, and metacognitive patterns. He also instructs AI to flag potential blind spots, such as consistent gaps or unstated assumptions in the reasoning.

Getting started doesn’t require a sophisticated setup. Describe the four layers, upload a transcript, and ask the AI to identify where each layer appears is enough to produce useful initial output.

Wispr Flow makes it easy to speak the goal and four-layer framework out loud before uploading a transcript, rather than typing the prompt. Spoken context tends to be richer and more natural than typed context, leading to more thorough extraction.

As AI processes each additional transcript, it builds on what it has already found rather than starting fresh. It actively looks for the patterns it identified previously, either confirming them or finding exceptions. The cumulative result is the fingerprint file: a compiled document capturing all four layers as they appear across the full transcript set.

#4: How to Use Your Cognitive Fingerprint File

The fingerprint file is designed to function as a portable context layer that can be loaded into any AI tool at any time.

The portability matters because AI models keep changing. New models launch regularly, and people switch between them. With a fingerprint file, whenever a new tool launches or a different model becomes preferred, the same rich context loads immediately. The AI landscape shifts; the fingerprint travels unchanged.

When the fingerprint file is loaded into a project, the AI isn’t operating from generic defaults. It has access to the conditional logic, the mental models, and the characteristic ways you frame and approach problems. Outputs will reflect how you would actually handle something, not how a general model approximates an expert.

When built from a sufficient transcript set, the fingerprint document typically runs 20 to 30 pages. That full document is worth reading; seeing your expertise reflected in organized, specific language is consistently one of the most clarifying experiences for Max's clients. But the working version is a condensed version trimmed to its most essential content.

What the Cognitive Fingerprint File Makes Possible

The most direct use is to improve AI output day to day. But the second benefit Max’s clients consistently report is confidence: once the logic of how someone operates is written down and clearly articulated, it becomes easier to explain and communicate. Every marketer faces the challenge of differentiation, and describing what makes their approach distinct is difficult when the expertise has always felt intuitive. A fingerprint file makes that implicit logic explicit.

From there, it becomes intellectual property. Coaching programs, sales frameworks, facilitation guides, and courses can be built on a methodology derived from the fingerprint, rather than reconstructed from scratch. The content is already there; the fingerprint file makes it visible.

For teams, the process scales outward. When multiple people have fingerprint files, patterns become visible at the group level — who reasons analytically, who responds to narrative, who generates ideas most effectively in unstructured brainstorms versus focused work sessions. That information changes how projects get assigned, how new ideas get pitched internally, and how the team compensates for each other’s blind spots.

Other Notes From This Episode

Connect with Michael Stelzner @Stelzner on Facebook and @Mike_Stelzner on X. Watch this interview and other exclusive content from Social Media Examiner on YouTube.

Where to subscribe: Apple Podcasts | Spotify | YouTube Music | YouTube | Amazon Music | RSS

✋🏽 If you enjoyed this episode of the AI Explored podcast, please head over to Apple Podcasts, leave a rating, write a review, and subscribe.

---

Stay Up-to-Date: Get New Marketing Articles Delivered to You!

Don't miss out on upcoming social media marketing insights and strategies! Sign up to receive notifications when we publish new articles on Social Media Examiner. Our expertly crafted content will help you stay ahead of the curve and drive results for your business. Click the link below to sign up now and receive our annual report!