Every morning, my AI reads a 20-page document about me before it does anything.
My working style. My active projects and priorities. Who I am trying to stay in touch with. My preferences for tone, follow-up, and communication. When I am traveling. What I am trying to avoid having on my plate.
This is not a journal. It is a context profile — a document I have built to give my agents the background information they need to actually make decisions on my behalf. And it is probably the highest-leverage thing I have done to improve how my AI agents perform.
The Stateless Problem
Most AI interactions start from zero.
You open a chat. The agent has no idea who you are, what you care about, what you are working on, or what a good outcome looks like for you specifically. Every request requires you to re-establish the context that should already be known. Every session, you are explaining things the agent should already have internalized.
This is not a flaw — it is the default state. AI models do not retain memory across sessions unless you give them something to remember. And most people never do.
The result is AI that is generically capable but never quite calibrated to you. It is the difference between a new hire who technically knows the job and a trusted assistant who knows your preferences, your patterns, and your priorities without having to ask every time.
A context profile closes that gap.
What a Context Profile Is
A context profile is a readable document — mine runs 20 pages — that captures the information your agents need to work effectively on your behalf.
At minimum, it should cover:
- Who you are and what you do — your role, your business, the kinds of decisions you make
- Current priorities and active projects — what is urgent, what is on the back burner, what is coming up
- Working preferences — how you like to communicate, your tone, your schedule and energy patterns
- Key relationships — who you need to stay in touch with and why, context on your most important contacts
- Upcoming travel or schedule changes — anything that would change how the agent should behave
- What to avoid — decisions you do not want made without you, topics that require your direct attention
I learned this approach from Flow, the CEO of Lindy, who has an even longer version than mine. His take: it is the single highest-leverage thing he has done to improve how his agents perform. Once an agent has this context, it stops being a generic tool and starts being something that actually knows how you work.
What Changes When You Load It
The difference shows up in every interaction.
Here is a concrete example. I set up a personal daily podcast — an AI workflow that pulls articles from my reading backlog each morning, converts them to audio using voice synthesis, and delivers them to me as a podcast episode. It is a useful setup, but what made it powerful was loading my context profile into the workflow.
Instead of just summarizing the articles, the AI now discusses them through my lens. It connects the ideas to projects I am actively working on. It surfaces action items that make sense given my current priorities. Same AI, same articles — completely different output, because the agent knows who it is talking to.
This pattern generalizes. An agent handling email suggestions performs differently when it knows your communication style and your relationship with each sender. An agent prepping you for a meeting performs differently when it knows your history with that person. An agent managing your backlog performs differently when it knows which types of tasks you are trying to reduce.
The context profile does not make agents smarter. It makes them more accurate — more calibrated to you specifically.
The Failure Mode: AI That Keeps Forgetting You
The pattern I see most often is what I call the one-night stand approach to AI. You have a great interaction. The agent does something useful. Then you close the tab and the next time you open it, the agent has forgotten everything. You start from scratch.
No continuity. No memory. The same clarifying questions every session. The same generic outputs that do not quite fit your situation.
This is fixable. But it requires treating your agents the way you would treat a new team member who needs proper onboarding — not just throwing them tasks and hoping they figure out what you need.
The fix is a durable, readable context profile that you load into your most important agents. Not every agent. Just the two or three that touch the highest-stakes parts of your work day.
How to Build Yours
The fastest way to create a context profile is not to write one from scratch. That produces something thin and generic.
The better approach: let AI interview you.
Open ChatGPT. Tell it you want to create a context profile for your AI agents — something that will help them make decisions on your behalf. Ask it to interview you. It will ask questions. Answer them honestly and in detail. After 20-30 minutes, ask it to compile your answers into a structured document.
Then review and edit. You will probably end up with something 5-15 pages long. Save it somewhere accessible — Google Docs works well because many agents can read directly from it.
Then load it into your most important agents. For me, that is my email summarizer, my daily briefing agent, and my meeting prep agent. Those three touch enough of my day that having them properly calibrated makes a noticeable difference.
The Payoff
When it is working well, you notice it in small ways at first.
The agent stops asking you to re-explain your role. It makes reasonable assumptions about tone. It surfaces the right things without being asked. It knows when you are in a different context — traveling, in back-to-back meetings — and adjusts accordingly.
The failure mode is an agent that keeps guessing who you are. The success state is an agent that already knows.
Your context profile is the bridge between those two. Build it once. Update it occasionally. Load it into your best agents.
Want to build smarter AI systems that actually work for you? The 4-Day AI Sprint covers practical frameworks for designing agents, workflows, and context systems that fit how you actually work.
