Building a custom working environment that actually works for you
Here’s the thing about AI systems—they’re incredibly powerful, but they’re also like that enthusiastic new employee who means well but has no idea how you like things done. Without guidance, they’ll default to their training patterns, which usually means verbose academic explanations when you need quick answers, or surface-level overviews when you need deep analysis. The solution? Create a session protocol that turns your AI into the perfect collaborator for your specific work style.
Think of your protocol as the difference between working in a generic office versus your perfectly customized home workspace. You wouldn’t tolerate a desk chair that doesn’t fit or lighting that strains your eyes, so why tolerate an AI that doesn’t match your communication preferences?
The Core Components of a Solid Protocol
Glossary and Terminology Start with your canonical definitions. This isn’t just about avoiding semantic drift—it’s about creating a shared vocabulary that makes your conversations more efficient. I typically include 10-15 key terms that come up repeatedly in my work. For instance, when I say “optimization,” I mean computational performance, not cost reduction or workflow improvement. When I reference “security,” I’m talking about data protection, not physical or operational security.
Your glossary should also flag potentially ambiguous terms and specify which meaning you want used. This prevents those frustrating moments when your AI starts talking about “user experience design” while you’re trying to discuss “user experience metrics.”
Interaction Style Preferences This is where you define the personality of your AI collaborator. Do you want Socratic questioning that challenges your assumptions? Straightforward answers without the hedging? A collaborative tone that treats you as an equal rather than a student?
I personally prefer what I call “senior consultant mode”—someone who’s knowledgeable but not condescending, who pushes back when I’m wrong but doesn’t waste time with excessive politeness. You might prefer a more supportive approach, or something more direct and businesslike. The key is being explicit about what works for you.
Output Formatting Standards Nothing kills productivity like having to repeatedly ask for information in a usable format. Specify exactly how you want responses structured. Do you want numbered lists for action items? Bulleted lists for options? Headers for different sections? Code blocks for technical examples?
I always request that complex responses use hierarchical numbering (1.1, 1.2, 2.1, etc.) because it makes follow-up questions easier—I can say “expand on point 2.3” instead of re-describing what I’m referring to. For technical discussions, I want brief explanations with detailed examples rather than lengthy theoretical treatments.
Session Hygiene Policies These are the rules that keep your conversations on track. Specify how you want the AI to handle ambiguous questions (ask for clarification vs. make reasonable assumptions), how to deal with off-topic tangents (redirect vs. briefly acknowledge), and what to do when reaching the limits of knowledge (admit uncertainty vs. speculate).
I use what I call “thread isolation”—when we’re discussing multiple related topics, I want clear boundaries between them so insights from one area don’t contaminate analysis in another. I also establish “fact vs. opinion” protocols so I know when the AI is stating something definitive versus offering interpretation.
Domain-Specific Behaviors Your protocol should adapt to the type of work you’re doing. Technical analysis requires different behaviors than creative brainstorming or strategic planning. For technical work, I want concise explanations with working examples. For strategic discussions, I prefer exploring multiple perspectives and challenging assumptions. For creative projects, I want expansive thinking and unusual connections.
Error Handling and Correction Define how you want mistakes handled. Should the AI acknowledge errors immediately and provide corrections, or wait for you to point them out? How should it handle situations where your information contradicts its training? I prefer systems that flag potential conflicts rather than silently choosing one interpretation over another.
Platform-Specific Considerations
Claude responds well to detailed role definitions and explicit behavioral boundaries. I typically frame my protocols as “You are a senior technical consultant with expertise in X. Your communication style should be…” Claude also handles complex, multi-part protocols better than other systems.
ChatGPT needs more specific formatting instructions and benefits from explicit examples of desired behavior. It tends to over-explain, so emphasizing conciseness helps. I often include sample exchanges showing the exact style I want. If you keep your protocol document as a canvas document, you can have it auto-load when you start up a session, using a system instruction.
Gemini works best with clear scope boundaries and specific requests for depth vs. breadth. It defaults to broad overviews, so if you want detailed analysis, you need to be explicit about that preference.
Building Your Custom Environment
The goal isn’t to create a perfect protocol on day one—it’s to build a working system that evolves with your needs. Start with basic preferences and add refinements as you discover what works and what doesn’t. Keep multiple protocol templates for different types of work.
Most importantly, remember that your protocol is an investment in efficiency. Yes, it takes time to craft and load at the beginning of each session, but it pays dividends throughout your conversation. Instead of fighting your AI’s default behaviors, you’re working with a collaborator that’s been configured to match your preferences and work style.
The difference between using AI with and without a proper protocol is like the difference between borrowing someone else’s car and driving your own perfectly adjusted vehicle. Both will get you where you’re going, but one makes the journey infinitely more pleasant and productive.
