Skip to content
Join 7,000+ leaders following Alastair's work on LinkedIn.

Your Core Four

Stop re-briefing AI every time

Four short files tell any AI tool who you are, what you're working on, how you work, and what matters when a decision is close - so it starts with the context that changes the answer.

See the book Already have it? Jump to resources

The missing piece

You've had this experience. You open ChatGPT or Claude, spend ten minutes explaining your role, your preferences and the tone you want, and get back something generic and robotic. So you rewrite it. You fix the tone. You take out the words you'd never use.

Then you open a new chat and do it again.

Your AI may know plenty about you by now. It still doesn't have the briefing this job needs.

In Hello, I'm New Here! I set out a system that fixes this. Instead of hunting for better prompts, you build four short files that tell AI who you are, what you're focused on, how you work, and why you make decisions the way you do. They take about 30 to 45 minutes to create.

The difference shows up straight away. You stop re-explaining yourself every session. The output starts sounding like you. And AI shifts from a tool you wrestle with into something closer to a well-briefed colleague who already knows the background.

In this book, you'll learn how to:

  • Eliminate the repetitive re-explaining that eats hours every week
  • Get AI output that matches your voice and your priorities
  • Build portable context files that work across any AI tool - now and in the future
  • Prepare for the next wave of AI agents without scrambling to catch up
See the book

The system

Four files. About 30 to 45 minutes.

Each file covers a different dimension of your working life. Together, they give any AI tool enough context to skip the guessing and get straight to useful output.

// update: quarterly

Who.md

Your identity - your role, your team, your organisation's size and culture. The "read this first" file.
// update: every few weeks

What.md

Your 3-5 active projects, their current status, and what "done" looks like for each one.
// update: quarterly

How.md

Your voice, your banned words, your formatting preferences, your tools. The file that stops AI sounding like AI.
// update: rarely

Why.md

Your values, your current trade-offs, and how you make difficult decisions. The AI's tie-breaker.

Get started

Starter templates

Plain Markdown files with the structure already laid out. Open them in any text editor, follow the prompts from the book, and fill them in with your AI tool.

CoreFour-StarterKit.zip Who.md, What.md, How.md, Why.md - with prompts

Plain text files. No special software needed. Works with every AI platform.

Setup guides

Wire it into your AI tool

AI platforms move buttons around constantly. These guides are kept current so the book doesn't have to be.

Claude (Anthropic)

Last verified: April 2026

1

Global style settings: Go to Settings > Profile > Custom Instructions. Paste your How.md file here. This ensures Claude always respects your voice and banned words in every conversation.

2

Create a Project: Click "Projects" in the sidebar, then "New Project." Name it something like "Daily Work" or "Q2 Operations."

3

Upload your Core Four: Inside the Project, find "Project Knowledge" and upload all four .md files. Every chat you start within this Project will read them automatically.

Claude's Projects feature makes it the strongest platform for managing context files right now.

ChatGPT (OpenAI)

Last verified: April 2026

1

Custom Instructions: Click your profile picture > Settings > Personalisation > Custom Instructions. Paste Who and What in the top box. Paste How and Why in the bottom box.

2

Custom GPTs (paid tier): Go to Explore GPTs > Create. Paste all four files into the Instructions box, or upload them as Knowledge files for a dedicated work assistant.

Google Gemini

Last verified: April 2026

1

Create a Gem: Open the Gem Manager and click "New Gem." Name it "My Advisor" or similar.

2

Paste your Core Four: Put all four files directly into the Gem's Instructions box. Select this Gem whenever you want to work with your full context.

Any other AI tool

Universal guidance

1

Find the System Prompt: Look for "Custom Instructions," "System Prompt," "Persona," or "Settings." Most tools have some version of this.

2

Paste your files: Copy the contents of all four .md files into that field. If there's a file upload option, use it.

3

Fallback: If you can't find any settings area, paste your files at the top of your first message in a new chat. Less elegant, but it works.

Common mistakes

Things that trip people up

These come up in nearly every workshop we run. Save yourself the debugging.

Writing "professional tone" in your How file

"Professional tone" means nothing to the machine. It's too vague to act on. Instead, describe how you actually write: "Direct, concise, first-person. Short paragraphs. No corporate jargon." The more specific you are, the less correcting you'll do later.

Pasting your LinkedIn bio as your Who file

Your LinkedIn profile is a marketing document. Your Who file is an operating manual. The AI doesn't need to know you're "a passionate leader driving innovation." It needs to know you manage three people, report to a CFO who hates long emails, and work in a highly regulated industry. Actual weeks, not aspirational bios.

Saying "Use UK English" but forgetting currency

AI models default to American norms. If you just say "UK English," you'll get British spelling but American dollars. Separate the two explicitly: "UK English spelling. All currency in Euros (or GBP)." Two different instructions, not one.

Only saying what NOT to do

Telling the AI "Don't use the word 'utilise'" is weaker than telling it what to do instead. Always pair bans with replacements: "Never use 'utilise' - use 'use' instead. Never use 'leverage' - use 'use' instead." Positive instructions are stronger than negative ones.

Putting ten projects in your What file

Your What file isn't your to-do list. It's the 3-5 major workstreams you'd mention if a colleague asked "What are you focused on this quarter?" If you list everything, the AI can't tell what actually matters right now. Be ruthless. If a project doesn't need AI help, leave it out.

Skipping the Why file entirely

Most people skip this one and plan to come back later. They never do. Then they wonder why the AI gives them "on the one hand, on the other hand" answers to every question. A rough Why file today - even just three bullet points about what you're optimising for right now - is worth far more than a perfect one you never write.

Letting the files go stale

A stale Core Four is worse than no Core Four. If the AI thinks you're still working on a project that was cancelled two months ago, its advice will be actively unhelpful. Set a 15-minute recurring calendar appointment every third Friday. Call it "Update Core Four." Your What file needs it most; the others decay more slowly.

Go deeper

The part a book can't do

The Core Four system gets you 80% of the way. The last 20% is the blind spots you can't see because you're too close to your own work. That's the bit we help with.

In a free 25-minute Focus Call, we'll review your context files, spot what's missing, and help you figure out where AI can make the biggest difference in your specific situation. No pitch, no pressure - just a conversation with someone who's done this with hundreds of professionals.

Book a Focus Call

25 mins or less. Video call. No commitment.