This is called personalization
In reality, using AI often comes with friction:
- answers are too long and difficult to scan
- you have to repeatedly ask to simplify or shorten things
- relevant information is buried and takes effort to extract
- the tone feels too complex or “overly polished”
- responses are not structured in a way that works for you
Instead of fixing this every time manually, you can:
- reduce output complexity
- define structure that works for you
- establish consistent response patterns
- remove unnecessary cognitive load
Copilot offers two simple ways to do this:
- Copilot Memory → AI learns and adapts to your working patterns over time
- Copilot Custom instructions → You define how AI should behave.
- Limitations
- Mitigations
A simple rule of thumb:
- use Memory when something describes how you generally work
- use Custom Instructions when you want AI to consistently respond in a specific way
- the pattern repeats
- it reflects a personal preference or recurring challenge
- you do not want to repeat it every time
Examples:
- “I prefer short, structured answers”
- “I work best with checklist-style outputs”
- “I need tasks broken down into small steps”
- “I struggle more with finishing tasks than starting them”
- you want a stable default format
- you want less variation in responses
- you already know what works well for you
Examples:
- “Always use bullet points”
- “Keep answers concise”
- “Use simple, direct language”
- “Give me no more than 3 next steps at a time”
In many cases, the best setup is a combination of both:
- use Memory for your recurring personal patterns
- use Custom Instructions for the response style and structure you want by default
That way, AI becomes more useful without creating more work.
You might recognise this:
- You get a long answer and don’t know where to start
- You spend more time refining the output than using it
- You constantly rephrase the same instructions
- You need a specific format, but don’t get it by default
- You feel the answer is “good” — but not usable
The problem is not the AI.
It is that the AI is not yet adapted to how you work.
Start simple.
You do not need a perfect setup.
Instead:
- identify one recurring challenge
- adjust AI slightly
- see if it helps
- keep what works
- change what doesn’t
Personalization is not about control.
It is about:
making AI work in a way that reduces friction for you