The quality of output you get from any AI model is a direct function of how well it understands you and your context.
Not the model. Not the prompt template. The context.
A 2025 study of 243 professionals across industries found that users who gave AI structured, context-rich inputs reported significantly higher task efficiency and better output quality than those who used generic prompts. The gap was not about the AI. It was about what they fed it. ¹
The most practical version of this for leaders: two files, always present, always current. One about you. One about your business. Built once through an AI interview. This issue shows you exactly how to create them.
📚What These Files Do
About-me file. Captures your role, your working style, your standards, what good work looks like in your field, and what you never want AI to do when it writes for you. This is the file that makes outputs sound like you instead of like a press release.
Company/team context file. Captures your current goals, what you're focused on this quarter, what you're actively saying no to, and where your time is going. This is the file that stops AI from giving you generic recommendations that ignore your actual situation.
Both files should be short. Dense. High-signal. Under 2,000 tokens for the about-me. Under 1,000 for the company file. The goal is not a comprehensive profile. It is a precise briefing.
🛠️ This Week’s Build
Platform: Claude, ChatGPT, Gemini, or Copilot. Any AI with a persistent context or system prompt feature works. The interview prompts below run on all of them.
Step 1: Build your About-Me file
Open a fresh chat session. Paste the prompt below. Let AI interview you one question at a time. Answer as you would in a voice note. You don't need to be precise. AI will distill your answers into a structured file.
You are building my about-me context file. This file will be referenced at the start of every AI session so the model understands who I am, how I work, and how to produce outputs that match my standards. It needs to be concise and high-signal, under 2,000 tokens total.
Your job: interview me using 15–20 questions, one at a time. Push back on vague answers. If I say "I like clarity," ask me what clarity actually looks like in my best work.
Cover these areas:
WHO I AM (3 questions)
- My role, firm, and industry
- Who I work with (clients, team, stakeholders)
- What a good week of work looks like for me
HOW I WORK (4 questions)
- Tools I use every day
- How I approach a task from zero to done
- What my review and quality-check process looks like
- What "done" means when I hand something to a client
WHAT GOOD LOOKS LIKE (3 questions)
- Best work I've produced recently and what made it good?
- What separates excellent from average in my field?
- What should I be looking for when evaluating my own output?
WHAT I HATE (3 questions)
- What does bad work look like in my field?
- What AI writing habits make me cringe most?
- When AI gets my work wrong, what is usually off?
MY RULES (2 questions)
- Hard lines I never cross in my work
- Two or three non-negotiables every piece of work must have
After the interview, compile everything into a single structured file. Do not save raw Q&A transcripts. Extract the patterns and write them as condensed prose and bullet points. Use this structure:
## Who I am
## How I work
## What good looks like
## What I hate
## My rules
## Instructions [10 numbered rules for how to work with me - what to do and not do]
Target: under 2,000 tokens. Every sentence must carry signal. If a sentence can be cut without losing information, cut it.
Step 2: Build your Company Context file
In the same session (or a new one right after), paste this. This file covers goals and strategy, not identity. The about-me file covers identity. No overlap.
Now build my company-context file. This file tells AI what I am working toward so it can make better decisions on every task. It is separate from the about-me file and covers only goals, focus areas, and strategic decisions.
Your job: interview me using 6–8 questions, one at a time.
GOALS (3–4 questions)
- Top 2–3 goals for this year, specific numbers or milestones
- Platforms, channels, or markets that matter most right now
- The one metric that would tell me this year was a success
- Revenue, client, or product milestones I'm targeting
DECISIONS (3–4 questions)
- What am I actively saying no to right now?
- What did I recently stop doing and why?
- Where am I spending most of my time and energy this quarter?
- Anything I'm betting on that most people in my field aren't?
After the interview, compile into a single file. Short sections. Bullet points. No filler. No identity information (that's in the about-me file).
Structure:
## Goals
## Focus right now
## Saying no to
Target: under 1,000 tokens. Update this file when priorities change, not on a schedule.
Step 3: Load your files into your AI platform
These two files are platform-agnostic. Build them once. Use them everywhere.
Every major AI platform today has a way to give a persistent briefing to your AI before it starts work. Your context files slot into all of them.
Projects and Notebooks: Claude Projects, ChatGPT Projects, and the newly launched Gemini Notebooks all let you attach files and custom instructions that load automatically at the start of every conversation. Upload both files and every session starts informed.
Custom Assistants and Agents: Any Custom GPT you build in ChatGPT, any Gem you configure in Gemini, or any agent you set up in Microsoft Copilot can carry these files in its knowledge base. Every assistant you build from here forward should have them loaded by default.
Claude Cowork: Save both files inside an ABOUT ME folder in your Cowork directory. Cowork reads that folder before every session. Any skill you fire inside Cowork will have your full context available from the first message.
The asset is the file. Not the platform. If you move tools, your files move with you. You don’t have to rebuild.
Step 4: Add one instruction line
At the top of your system prompt or project instructions, add:
Before every task, read the about-me file and the company-context file. Use them to calibrate tone, standards, and relevance. Do not re-explain who I am when I ask a question, you already know.
Step 5: Test with a real task
Pick something you would normally do this week. A client update. A short memo. An email to a stakeholder. Run it in your newly configured project. Compare the output to what you'd normally get.
🧭 Try This Week
Block 30 minutes. Open a Claude or ChatGPT session and run the about-me interview prompt.
Stay in the same session and run the company context prompt.
Save both outputs as text files (about-me.md and company-context.md).
Load them into whichever platform you use daily using the setup guide above.
Run one real work task and note what is different about the output.
📊 Outcomes
Time saved: Eliminates 3–5 minutes of context-setting at the start of every AI session. For a team running 10 sessions per day, that compounds to hours per week.
Workflow steps eliminated: No more mid-conversation corrections for tone, style, or audience. No more re-explaining your standards every time.
Cost and efficiency impact: Shorter input prompts mean fewer tokens consumed per session. For firms managing multiple users on AI platforms, this translates to real cost reduction at scale.
Strategic value: Your AI now operates from a standing briefing. Every output starts calibrated to your standards, your language, and your current priorities, not a generic professional default.
How valuable was today’s issue for you?
Found this useful? Your team and friends will too. It takes 30 seconds to forward and could save them hours or days.
Was this forwarded to you? Subscribe to Clarity Prompts
↗︎ Think Better. Execute Faster.
See you next week — Purti
Clarity Prompts The weekly AI operating system for sharper leaders.
¹ Anam, R.K. et al. (2025). Prompt Engineering and the Effectiveness of Large Language Models in Enhancing Human Productivity. arXiv:2507.18638. Available at arxiv.org/abs/2507.18638
