The free guide
How I actually use AI
How to get real value from AI, whichever tool you use. How it works in plain words, then the setups that make an agent worth trusting, each one a block you paste in. It's a starter version of how I actually work with AI every day.
Before you start
All you need is a free account with an AI chat tool.
Your first five minutesAlready use AI? Skip to Set up your agent.
Start
Your first steps
Open a tool, brief it properly, pick up a few techniques and keep private things private. Four short pages you can read on a phone.
Open Start- 01Your first five minutes4 minOpen a free AI tool, send your first message and switch on two privacy settings.
- 02Brief it properly3 minLet it ask you questions first, then write briefs that get the answer you need.
- 03Prompting techniques3 minLet it write the prompt, reply to one part of an answer, and other short techniques that work in any tool.
- 04What not to paste4 minWhat's fine, what needs care, what stays out, and who decides.
How it works
What's actually going on
The few things worth knowing about what's inside an AI tool, because nearly every odd thing it does makes sense once you know them.
Open How it works- 01What's inside an AI model4 minTokens, a network of numbers, and why it predicts instead of looking things up.
- 02Why it makes things up3 minWhy a tool that's usually right can be confidently wrong, and what helps.
- 03Sessions, context and memory3 minWhy long chats forget, what carries over, and how tools split work into sessions.
- 04Models and reasoning levels3 minFast, default and thinking, and when each is worth it.
Set up your agent
Copy-paste setups for any tool
Each setup is one page, with a block to paste into your agent and a way to tell it's working. They work in any AI tool.
Open Set up your agent- 01Set up your agent3 minWhere setups go, and which four to do first.
- 02Ground rules2 minWhat it may do on its own, and what it must ask first.
- 03A memory that works3 minA one-page note it reads at the start and updates at the end.
- 04Straight answers2 minSo it tells you when you're wrong instead of agreeing with you.
- 05Check the facts4 minSources, a second pass, and a check that runs the same way every time.
- 06Fresh sessions2 minWhen to start a new chat, and the handover note that carries things across.
- 07Write like you2 minA voice file built from your own writing, and better with every correction.
- 08Save what works3 minTurn the job you repeat most into a skill.
- 09Connect your tools4 minRead-only first, reply means draft, and a connection log.
- 10Routines3 minA morning brief or weekly review that turns up on its own.
- 11Test your setup3 minTen real jobs, rerun after every change.
- 12More than one agent4 minWhen a second agent earns its place, and how to keep them from tripping over each other.
At work
For teams and anyone who owns the risk
The rules to set before anything's connected, sensitive data and incidents, and the setups a team leans on most.
Open At work- 01Set the rules first8 minData buckets, draft-confirm-execute, and the difference between a request and a control.
- 02Health, client and member data5 minA small practice, a regulated business, incident clocks and questions for the vendor.
- 03Connect your tools4 minRead-only first, admin approval, and a log of what's connected and who owns it.
- 04Test your setup3 minTen real jobs, and why your test can be wrong too.
- 05Routines3 minRoutines with an owner, an off switch and a quiet-day line.
- 06More than one agent4 minWhat extra agents cost, and the three signs one has earned its place.
How I run it
Every setup in this guide next to the bigger version I run every day, with the real files.
See how I run itThis guide works with any AI tool. For one tool in detail, there's Make Copilot yours and the Grok Bot guide.
Appendices
- Worked examplesThe everyday jobs, across different kinds of work.
- Prompt libraryPrompts for email, meetings, writing and research. Copy one, fill in the brackets and paste it in.
- The nine starter skillsNine skills you can paste into any chat or install as folders.
- Connector cheat sheetsEmail, calendar, files, work chat and notes, with the permissions to expect and the first safe thing to try.
- Safety checklistThe short version of Set the rules first, with checks for before you connect and before you automate.
- When it plays upQuick fixes, getting your files in, and which plan or model.
- GlossaryThe words, in plain English.
- Update logA dated list of what changed and why.
Every setup is also on GitHub, ready to hand to your agent: github.com/joelbrilliant/how-i-use-ai.