The Voice Capture Toolkit

01 Mar, 2026

This is the companion piece to The Voice Problem. That post covers why most content written on behalf of leaders sounds generic. This one covers exactly how I'm building the system to fix it — tools, folder structure, all of it.

The system has two sides: their side (dead simple) and your side (where the AI infrastructure lives). The person giving you their voice does almost nothing. You do the building.

All 60 questions are at the end of this document.

Their Side

You send them the 60 questions at the bottom of this page and tell them to paste them into Claude. Claude runs the session — asks the questions one at a time, pushes back on safe answers, asks for examples, goes deeper when something interesting surfaces. The instructions baked into the questions tell Claude exactly how to conduct the conversation.

No meeting. No comms person watching. Just them and Claude, at their own pace.

When the questions are done, Claude compiles everything into a structured voice profile — their beliefs, their patterns, their hard lines, all of it. The person reviews it, then sends you the finished voice-profile.md. That's the only thing they need to do.

Some people will knock it out in one sitting. Some will chip away at it over a week. Both are fine. The only thing that kills the process is safe, sanitized answers — and Claude is instructed to push past those.

For someone who's reserved or short on time: even 20-25 questions focused on Beliefs and Hard Lines will give you more than you currently have, which is probably a handful of recycled Slack messages.

Your Side: The Stack

Everything runs on:

• Claude (Cowork) — the writing engine. Cowork reads and writes directly to your local folders. No copy-pasting between apps.
• Obsidian — the vault where everything lives. Portable, searchable, version-controllable. This is optional — any folder on your computer works.
• Markdown files — the format for everything

That's the whole stack. Cowork pointed at a folder in your Obsidian vault. Everything else flows from that.

Step 1: Set Up the Folder

Create this in your Obsidian vault:

/Voice Profiles/
  /[Person Name]/
    voice-profile.md
    voice-interview-answers.md
    /weekly-input/

When they send you the compiled voice profile, drop it in as voice-profile.md. Then — and this is critical — read through the entire thing and edit it yourself. Claude will get 80% right. The other 20% is where it smoothed out something important, missed a nuance, or organized something in a way that loses the feel. You know this person. Trust your ear over Claude's structure.

Two more things to put in the folder:

Save their raw Q&A answers. Ask them to send the full Claude conversation alongside the compiled profile. Save it as voice-interview-answers.md. The profile is the compressed version — the answers have texture the profile can't fully capture. Specific stories, the way they phrased something in the moment, tangents that didn't fit neatly into a section. When the profile feels thin on a topic, the answers will bail you out.

Reference their published writing. If this person has a blog, a newsletter, a LinkedIn history they're proud of — anything where they've already expressed themselves in their own words — note it in the Instructions for Claude section at the top of the voice profile. Published writing is the most refined expression of someone's voice. It's the standard Claude should be writing toward.

Because the instructions are baked into the top of voice-profile.md, every new Cowork session is the same: point it at the folder, tell it to read the voice profile. Claude loads the voice, the rules, and the context. No separate system prompt, no project setup, no uploading files. The folder is the system.

Step 2: The Weekly Feed

This is the part that keeps the system alive. Without it, you're back to mining Slack within three months.

The person needs to give you something every week. The bar has to be absurdly low or they won't do it. Here's what works:

Option A: Voice memos into Whispr. They record 3-5 minutes on their commute or between meetings. "Here's what I was thinking about this week." Whispr transcribes it. They send you the transcript.

Option B: A shared note. A Google Doc, a Slack DM, an email — whatever has the least friction for them. They drop a few sentences in whenever something hits them. A conversation that surprised them. A take they had in a meeting. Something they read that pissed them off. No structure. No pressure to make it good.

Option C: A quick Claude session. If they won't write or record, have them open Claude for 10 minutes once a week. Claude asks: "What pissed you off this week? What excited you? What did you change your mind about?" They copy the conversation and send it to you.

Option D: Enterprise AI (Glean, etc.). If your company runs something like Glean, the person's thinking is already being captured — meeting notes, Slack threads, email threads, internal docs. Set up a recurring export or saved search that compiles their recent contributions and sends a weekly digest. They don't even have to remember to do anything. You just pull what's already there.

Option B is the lightest lift. Option D is even lighter if you have the tools. Option C is the most reliable for depth. Option A works great for people who think out loud.

Whatever they give you, drop it as a new dated file in /weekly-input/ — something like 2026-03-01.md. Raw, unedited. That's it. Next time you open Cowork and ask it to write something, it reads the voice profile and the weekly input files. Fresh synthesis every time, from the raw material.

What This Looks Like in Practice

Week 1: You send them the questions. They paste them into Claude and have a conversation. Claude compiles the profile. They send you voice-profile.md. You review, edit, and drop it in the folder. Total active time for you: 1-2 hours. Total time for them: 60-90 minutes with Claude, no meetings.

Week 2 onward: They send you 5 minutes of input — a Whispr transcript, a few sentences in a shared doc, a pasted Claude conversation, or a Glean digest you pulled yourself. You save it as a dated file in /weekly-input/. Total time: 2 minutes.

When you need to write something: Open Cowork, point it at the folder, "read voice-profile.md and the recent files in /weekly-input/, then write me a LinkedIn post about [topic]." Claude reads it all fresh — the profile, the raw input — and drafts. The output sounds like a person — not because you got lucky, but because you're writing from real material.

One folder. One file. One AI. Just a folder that gets smarter over time.

One More Thing

I built this because I was tired of writing from scraps. Tired of guessing. Tired of producing content that was technically fine but sounded like it could have come from anyone.

If you're in comms or content and you're doing the Slack-mining, all-hands-scraping, transcript-recycling thing... you don't have to. Build the system. It takes an afternoon to set up and minutes a week to maintain.

The work gets better. And you stop feeling like a forger.

Voice Profile — 60 Questions

Paste everything below into Claude. Claude will walk you through all 60 questions conversationally, then compile your answers into a voice profile you can send to your comms person.

Credit: This approach was inspired by Ruben Hassid.

Instructions for Claude

You are conducting a voice capture session. Your job is to ask the 60 questions below, one at a time, conversationally. Don't rush. Don't dump multiple questions at once.

Your role:

• Ask one question at a time. Wait for their answer before moving on.
• Push back on safe, generic answers. If their answer could apply to any executive in the industry, say so and ask them to go deeper.
• Ask for specific examples. "Can you give me a time that actually happened?" is your best follow-up.
• Call out contradictions. If something they say doesn't match an earlier answer, surface it.
• Let them go on tangents. Some of the best material comes from the thing they weren't asked.
• Don't summarize or rewrite their words. Preserve exactly how they say things.
• Keep it conversational. This should feel like talking to a sharp friend, not filling out a form.

When all 60 questions are done, compile everything into a voice profile document structured like this:

1. Instructions for Claude — a short section at the top that tells any AI reading this file how to use it: how to draft content in this person's voice, what rules to follow, and the final check: "Does this sound like something they would actually say, or does it sound like an AI trying very hard to imitate them?"
2. Core Identity — 2-3 sentences capturing the essence of who this person is
3. Beliefs & Hot Takes — preserve full answers
4. Voice & Personality — preserve full answers
5. Aesthetic Crimes — preserve full answers
6. Hard Lines — preserve full answers
7. Communication Patterns — preserve full answers
8. Beyond the Role — preserve full answers
9. Quick Reference Card:
   • Always: specific patterns to follow, labeled as HARD RULE, STRONG TENDENCY, or LIGHT PREFERENCE
   • Never: specific things to avoid, same labels
   • Signature Phrases: actual language from their answers
   • Voice Calibration: 3-5 direct quotes that capture their tone
10. Anti-Overfitting Guide — include this section at the end:
    • Spirit over letter. The goal is to internalize their sensibility, not mechanically apply every pattern. A piece that uses 3 of their tendencies naturally will always beat one that forces in 10 of them awkwardly.
    • Frequency labels. For each tendency, note whether it's a HARD RULE (never violate), STRONG TENDENCY (70-80% of the time), or LIGHT PREFERENCE (context-dependent). When unlabeled, assume light preference.
    • Context matters. A tweet is not a newsletter is not an op-ed. Use judgment about which patterns fit which format.
    • Natural variation. Real writers aren't perfectly consistent. Don't start every piece the same way just because they have a signature open. Don't avoid a word forever just because they said they dislike it — sometimes it's the right word.
    • The litmus test. Before finalizing anything: "Does this sound like something they would actually write — or does it sound like an AI trying very hard to imitate them?" If it feels forced, pull back. Less imitation, more inhabitation.

Preserve their actual words. Don't summarize — keep the full depth. Clean up for readability but don't rewrite in your voice. Save the result as a markdown file called voice-profile.md and tell them: "Send this file and the full conversation to your comms person. They'll take it from here."

The Questions

Section 1: Beliefs & Hot Takes

1. What do you believe about your industry that most people in it would disagree with?
2. What's a piece of conventional wisdom in your space that you think is flat-out wrong?
3. When you see people in your space — competitors, peers, thought leaders — talking publicly, what bothers you most about how they do it?
4. What's something you've built, done, or been part of that you're genuinely proud of — not the version you'd put on a stage, the real version?
5. What do you think people in your world get completely wrong about the people they're trying to serve?
6. If you could change one thing about how leaders in your space communicate, what would it be?
7. What's a trend everyone's excited about that you're skeptical of?
8. What hill would you die on professionally?
9. What do you think matters most about what you do — and is that different from what you say matters most?
10. What's something you've changed your mind about in the last few years?

Section 2: Voice & Personality

11. When you disagree with someone in a meeting, what does that sound like? Are you direct? Diplomatic? Do you get heated?
12. When you're genuinely excited about something, how do people around you know? What does that look like?
13. How would your direct reports describe your communication style?
14. Do you use humor at work? If so, what kind — dry, self-deprecating, sarcastic, something else?
15. When you're explaining something complex to someone, how do you approach it? Metaphors? Data? Stories?
16. How do you write on Slack or email when you're moving fast? Short and blunt? Longer and detailed?
17. Do you tend to think out loud or process internally before sharing?
18. When you give a presentation, what are you like? Do you stick to the slides or go off-script?
19. Tell me about a conversation you had recently where you felt like you really nailed your point. What did you say?
20. Is there a leader, writer, or public figure whose communication style you admire? What about them resonates?

Section 3: Aesthetic Crimes

21. When you read a piece of content with your name on it and it doesn't sound like you, what's usually wrong?
22. What kind of marketing or corporate language makes you cringe?
23. Are there specific words or phrases you never want associated with you or the company?
24. When you see other executives posting on LinkedIn, what turns you off?
25. What kind of content do you think is a waste of time — either to produce or to read?
26. When someone sends you a draft to review, what's the thing that makes you immediately want to rewrite it?
27. Is there a type of email or internal message you get that drives you crazy? What about it?
28. What does "trying too hard" look like in professional communication?
29. Do you have opinions about formatting — bullet points, headers, long paragraphs, emojis?
30. What's the difference between content that feels authentic and content that feels corporate to you?

Section 4: Hard Lines

31. Are there topics you don't want to publicly comment on? (Political, social, competitive, personal?)
32. Are there claims — about your company, your expertise, your track record — you'd never make publicly, even if someone pushed you to?
33. How do you feel about being positioned as a "thought leader"? Does that term resonate or make you uncomfortable?
34. Is there a way people frame you or your work that feels dishonest, even if it's technically true?
35. How personal are you willing to get in professional content? Family? Failures? Past experiences?
36. How do you feel about taking a strong public stance on something? Where's the line?
37. Are there competitors you don't want to mention by name, even positively?
38. What's something a well-meaning comms person might write for you that would make you say "I would never say that"?
39. When should your name be on something versus the company's?
40. Is there anything about your background or experience you prefer to keep private professionally?

Section 5: How You Actually Communicate

41. Can you show me a few Slack messages or emails you've sent recently that you think represent how you naturally communicate?
42. When you write something important — a board update, a team email, an investor note — what's your process?
43. Do you outline or just start writing?
44. How long are your emails and messages typically? Do you go long or keep it tight?
45. Do you use exclamation points? Emojis? ALL CAPS? How do you add emphasis?
46. When you're writing something and it's not coming out right, what do you do?
47. Do you read things out loud before sending?
48. What does your internal monologue sound like when you're working through a problem?
49. If you had to describe your communication style in one sentence, what would you say?
50. What's something about how you communicate that people often misread or misunderstand?

Section 6: Beyond the Role

51. What would you write about if it had nothing to do with your job?
52. What are you reading, watching, or listening to right now that's shaping how you think?
53. What life experience outside of work most influences how you see things professionally?
54. What opinions do you hold that have nothing to do with your industry but everything to do with who you are?
55. If someone followed you for a week, what would surprise them about how you spend your time or what you care about?
56. What's something you're figuring out right now — not at work, just in life?
57. Who outside your professional world do you admire, and what about them resonates?
58. What do you want to be known for beyond your job title?
59. Is there a personal story or experience that keeps coming back to you — one that shows up in how you think even when you're not talking about it directly?
60. What would your closest friends say you care about most — and is that different from what your colleagues would say?

#building