There's a version of AI that most people have and a version of AI that actually works well.
The difference is usually context.
When you load an AI agent with a real understanding of who you are — your voice, your decision-making patterns, your values, how you like outputs delivered — the quality of what it produces changes substantially. It stops sounding generic. It stops giving you answers that require major editing. It starts working the way a well-briefed assistant would.
The problem is that creating a good context profile sounds like a lot of work. You'd need to write something comprehensive about yourself — your communication style, your priorities, your preferences. Most people look at that task and push it off indefinitely.
Here's how I skip that entirely: I use AI to write the context profile.
The Problem With Writing It Yourself
Trying to write a context profile from scratch puts you in an awkward position. You know yourself, but articulating your own patterns is surprisingly hard.
Most people don't have a clear sense of what their “decision-making style” looks like in writing. They know how they make decisions — they've been doing it for years — but describing that as a document takes a kind of structured self-reflection that doesn't come naturally.
Same with voice. You know what sounds like you. But writing a guide for how AI should sound like you requires you to reverse-engineer your own instincts, and that's genuinely difficult.
The result is that most context profiles, when people actually write them, are too short and too vague to be useful. “Conversational and direct.” “Values relationships.” These phrases don't tell an AI anything it can act on.
The Interview Method
The approach that works better: have AI ask you the right questions.
Give Claude (or ChatGPT) a set of category headers — the structure of what you want to capture. Good headers for a personal context profile include:
- Voice and communication style: How do you like to communicate? What's your tone in writing? What phrases or patterns would others recognize as “you”?
- Decision-making principles: When you're choosing between options, what matters most? What do you optimize for?
- Values: What drives your work? What won't you compromise on?
- Operational preferences: How do you want AI outputs delivered? Long or short? Structured or flowing? What should they avoid?
- Context about your work: What are you working on? Who are your clients or stakeholders? What's the current priority?
Give those headers to Claude and say: “I want you to interview me on each of these categories. Ask me one question at a time. After I've answered all the questions, write a structured context profile from my answers.”
Then answer the questions naturally — the same way you'd talk to someone. Don't write carefully crafted responses. Just answer.
What You End Up With
The synthesized profile is richer than anything most people would write themselves. Because the interview process surfaces things you know but wouldn't think to include.
I did this with Evan Baehr, who runs Arena Hall, a co-working and event space in Austin. He had multiple AI workflows being built — meeting prep agents, weekly briefings, communication drafts — and they all needed to understand how Evan thinks and operates.
We didn't sit down and “write” his context profile. Claude asked him questions about how he makes decisions when two priorities conflict, how he communicates with members versus with investors, what he cares about when setting the week's agenda.
Evan answered. Claude wrote.
What came out was a document that captured how Evan actually operates — including patterns he hadn't consciously articulated before. The agents loaded with that profile produce outputs that require far less editing because they start from an accurate understanding of who Evan is and what he needs.
Context Files as Reusable Assets
Once you have the profile, it becomes an asset you can use everywhere.
Load it into Lindy's system settings for your automation workflows. Add it to Claude Projects for your writing and thinking work. Paste the relevant sections into any agent you're building.
The model I use: one central Google Doc that contains the master context profile. Every agent I build for a client gets that document. When something in the profile needs updating — a priority shifts, a communication preference changes — I update the doc once. Every agent that reads it gets smarter at the same time.
This is how context files become leverage. You build the asset once. It pays off every time you use AI, because your agents stop needing to ask the same clarifying questions and stop producing outputs that miss the mark.
One of my clients ended up with a 33-page context profile built over months of annual reviews, weekly reflections, and coaching sessions. Every AI he uses reads that document before doing any work. Generic AI becomes personalized AI. That's where the real performance difference is.
How to Build Yours This Week
Step 1. Open Claude or ChatGPT and give it a prompt like this: “I want to create a context profile for my AI agents. Please interview me on the following categories, one question at a time: [paste your headers]. After I've answered all the questions, synthesize my answers into a structured profile document I can reuse.”
Step 2. Answer the questions. Don't overthink it. Answer the way you'd talk to a person, not the way you'd write a document.
Step 3. When the interview is done, ask it to write the profile. Review the output and add anything it missed or adjust anything that feels off.
Step 4. Save it somewhere central — a Google Doc works well. This is your context asset.
Step 5. Load it into your agents. In Lindy, paste the relevant sections into system prompt settings. In Claude Projects, attach the document. In ChatGPT, add it to custom instructions.
Step 6. Test it. Ask the agent something it would have gotten wrong before. Notice the difference.
The whole process takes 30-40 minutes the first time. Every AI interaction after that benefits from the investment.
The Meta-Lesson
There's a pattern worth noticing here. The thing that makes AI useful is context. The thing that makes creating context feel like work is having to write it yourself. The solution is to use AI to help you create the context that makes AI better.
This is the kind of recursive improvement that becomes available when you start thinking of AI as a tool for building better AI systems — not just a tool for getting individual tasks done.
You don't have to write a context profile. You just have to answer questions.
I help founders and operators build personalized AI systems that actually know who they're working for. If you want to set up proper context profiles for your agents, reach out or check out my AI consulting and workshop programs.
