Stop micromanaging your AI
Re-explaining your situation at the start of every chat is the same as briefing a new contractor every morning. The fix is boring: write the context down once, somewhere the model can always see it.
Most people using AI haven’t noticed that they’ve become managers. You’ve been handed a team of capable, eager interns, and how much you get out of them now depends on how well you brief them. Whether you’re prompting a chatbot or fine-tuning your own model, you’re setting objectives, handing over resources and judging the output. Same as with people, the quality of the management shows up in the quality of the work.
Context is the job
You wouldn’t delegate to a person by saying “write a report”. You’d tell them the purpose, the audience and the points that matter. Models are the same. A vague prompt gets you a vague response. Treat each one as a briefing and explain the why behind the what.
| Briefing a person | Briefing a model |
|---|---|
| ”Draft a Q3 sales report, focusing on retention." | "Analyse Q3 sales data. Highlight churn and retention rates. Prioritise actionable insights." |
| "Design a landing page that converts." | "Design a landing page for [product]. Objective: drive sign-ups. Audience: [demographic]. Include a clear CTA and social proof.” |
The clearer the context, the better everyone performs.
Micro versus macro
The difference between micromanaging a model and managing it well comes down to one question. Are you writing a prompt per task, or have you built something that supplies the context every time?
- Micromanaging. You hand-craft every prompt, tweak every parameter, and fix up each output yourself. It’s the equivalent of writing every line of code rather than leading a team, and it doesn’t scale.
- Macro-managing. You put the context in place up front. The model can already see your values, your data and examples of good work, so it needs much less hand-holding to get somewhere useful.
Building the context system
Three things that consistently work:
- Personal context. Spend an hour letting a model interview you about your work, your style and your sense of humour. What you get out of it is a profile you can drop into every prompt, and it’s the difference between generic output and output that sounds like you.
- Business context. Tools like Claude Projects let you upload the relevant documents once and build a knowledge base the model draws from. Coding assistants do the same thing with your codebase. No more re-explaining the basics every session.
- Show, don’t tell. Example outputs are the highest-value thing you can hand a model. A short collection of what good looks like moves it towards your preferences faster than any amount of describing, and it cuts down the revision loop.
There’s real up-front work in this, and it pays for itself quickly, both in output quality and in how much less irritating the whole thing is. The move you’re making is from tinkering with prompts to building the system the prompts run inside.