Free Guide, Part Two
You've sent your first messages, corrected a few answers, learned to check the ones that matter. This is where an AI chat stops being a box you type into and starts being a system that already knows who you are, so you stop re-explaining yourself every time.
The Toolkit
None of these are required. They're what turns "I use AI sometimes" into "AI already knows my business before I've typed a word."
Right now, every new chat starts from zero. The AI does not know your business, your role, your customers, or how you like things written, unless you type it out again. A master prompt fixes that: one document you keep and paste in (or upload) that answers the basics once.
Interview me so you can write a one-page master prompt about my business and how I like things written. Ask one question at a time.
A master prompt says who you are. A system prompt says how to do one particular job, every time, the same way. If you catch yourself writing a similar request over and over, weekly emails, meeting notes, social captions, that task is worth turning into a saved instruction set.
Based on everything we just went back and forth on, write me a reusable instruction set that would get me straight to this result next time, without the corrections.
In part one you learned Role, Context, Command, and Format. These seven go further, drop any of them into a prompt and the answer usually gets noticeably more useful.
Most AI chat tools now let you group related conversations into a folder that holds shared files and background, sometimes called a project or a workspace. Instead of pasting the same brand guide, pricing sheet, or report into every new chat, you upload it once and every conversation inside that folder can see it.
Many AI tools have a settings page where you can set standing preferences that apply to every conversation, not just the current one. Look for something called "custom instructions" or "personalisation" in your account settings. Worth setting:
Set this once and every future chat starts already tuned to how you like things.
Once a system prompt reliably gets you the same good result, some AI tools let you save it as a standalone assistant with its own name and icon, sometimes called a custom GPT or a custom assistant. Anyone with access can use it without knowing how to prompt at all, they just answer its questions.
Good candidates: a client-intake assistant that always asks the same questions in the same order, a first-draft-email writer trained on your voice, a support assistant that only answers from your own documentation.
Trying to build all six at once is how this whole page ends up ignored. Do them in this order instead, each one makes the next one easier.
Just for you. Takes twenty minutes, pays off in every single chat after.
Pick the task you repeat most often and turn it into a saved instruction set.
Somewhere ongoing, a client, a business area, that keeps needing the same background.
Only once you have a system prompt good enough that someone else could run it.
Reference, Not A Step
Part one covered prompt, model, hallucination, context, and token. These pick up where that left off.
Once The Toolkit Is Built
A deeper interview about your goals and skills, finding your blind spots, a real roadmap, and a profile brief that carries it all into every chat. Still free.
Rather someone just built this for you?
If you'd rather skip building this yourself, submit a brief and we'll tell you what fits your business and what it costs.