Prompting as a Professional Skill

You picked your assistant in the last lesson. Now you type in "write me a marketing email," and out comes something flat and generic, an email you would never send. Same tool. The gap is not the tool; it is what you asked it.

Why This Matters

The quality of what AI gives back is set by the quality of what you put in. This is Specification Thinking, the P5 skill of briefing an AI tool as precisely as you would brief a capable colleague. By the end of this lesson you will be able to:

• Construct a complete prompt using the five parts of the CRISP framework
• Transform a vague one-line request into a structured prompt that returns usable work
• Diagnose which CRISP part is missing when an output disappoints

The CRISP framework

CRISP is five parts you add to a request. Include them and you are specifying the work the way you would brief a colleague. Leave one out, and the AI quietly fills the gap with an assumption, usually a generic one.

The order is not arbitrary. Context and Role set who the AI is and the world it is working in; Instructions and Specifics say what to do and how tightly; Parameters fix the shape of what comes back. You will not need all five every time, but when an answer comes back wrong, the fault is almost always a missing part. Without a Role the tone is off, and without Specifics it invents details. No Parameters, and you get five paragraphs when you wanted five bullet points.

One caution. A well-built prompt narrows the room for the AI to invent, but it does not close it. The model answers with the same confidence whether your specifics were complete or not, so a tighter prompt raises the odds of a good draft. It never lifts your duty to check it.

Build your AIRI Mentor

Now apply CRISP to something you will reuse for the rest of the course: a coaching agent of your own. The slow-changing parts (Context, Role, Parameters) get baked in once, and you load it with the course's own knowledge so it works from the real framework instead of guessing. This is your AIRI Mentor, and from here it coaches you through every Part 3 task and helps you capture each one.

You give your Mentor context in two different places, and the difference is worth learning. The prompt goes in the instructions field: it sets who the Mentor is and how it coaches, and the tool reads it on every message. The two reference files go in as knowledge (Claude calls it Project knowledge. Gemini puts it in a Gem's files. A Custom GPT calls it knowledge), material the Mentor looks things up in when it needs them. Keeping the framework and the course map as knowledge files, rather than pasting them into the instructions, keeps the instructions short and lets the Mentor quote the real source instead of guessing. That split, short instructions plus the right reference material, is the same "give the AI good context" habit this whole lesson is about.

1. In your chosen tool, create a Project (Claude), Gem (Gemini), Custom GPT (ChatGPT), or agent (Copilot) and name it "AIRI Mentor".
2. Add its two knowledge files (download both from the box just below): the AIRI Framework Guide and the AI4E Course Map. These are reference material your Mentor reads from, not instructions.
3. Paste the AIRI Mentor prompt (further down) into the instructions field, and fill in your role and industry where marked.
4. Test it: give your Mentor a real request from your week as a lazy one-liner, then ask which CRISP parts are missing and what each should say. A Mentor that knows the framework will tie its answer to specific dimensions.

Your two knowledge files

Download both, then add them to your Mentor as knowledge (not as instructions):

• AIRI Framework Guide (v3.2): the five pillars, fifteen dimensions, and the L0-L4 levels. Free download from airi.foundation.
• AI4E Course Map: every lesson and what it covers, so your Mentor coaches against the real course rather than guessing from a title. Download ai4e-course-map.txt.

The prompt below is itself a worked example of CRISP: every block is labelled C, R, I, S, P, so you can see the framework applied to a real, reusable brief. This is the standard to aim for in your own prompts.

C: CONTEXT
I am taking "AI For Everyone" (AI4E) by aiready.sg, a course built on the AIRI Framework (AI Readiness Index) that takes me to AI Ready (Level 2): using AI on real work, checking its output, and taking on work that was not practical before.
My role: [FILL IN]
My industry: [FILL IN]
I have uploaded two files as your knowledge: the AIRI Framework Guide (the authoritative reference for the 5 pillars, 15 dimensions and the L0-L4 levels) and the AI4E Course Map (what each lesson covers). Treat them as your primary source; if you are unsure about a dimension, level or lesson, check them first, then search aiready.sg or airi.foundation rather than guessing. Never invent framework detail.
Part 3 of the course is hands-on: each lesson ends with a task where I point an AI tool at a real piece of my work and capture the exchange. At the end I review the whole record myself against the AIRI criteria for my role and industry.

R: ROLE
Act as my AIRI Mentor: an experienced, plain-speaking AI-readiness coach. You are a coach, not a marker; you help me do each task well and capture it sharply, and you never score me and never submit anything on my behalf.

I: INSTRUCTIONS
When I describe a Part-3 task I am about to do or have just done:
1. Coach it: tighten my prompt, push me to check the output against the source, and ask whether this is work I could not realistically have done before AI.
2. Ask me ONE short reflection question, tied to the AIRI dimension the task touches.
3. Help me write a clean capture record to paste into that lesson's assignment (the fields are in Specifics below).
4. Connect the dots: track which dimensions my tasks cover, tell me what is thin or missing, and suggest what to try next.
5. Point me to the right resource when I am stuck (listed in Specifics); never a stale or invented tutorial.

S: SPECIFICS
The capture record must contain exactly: the tool and model I used; my exact prompt(s); the output (or its key part); the one error or limitation I caught and how I fixed it; and whether this was work I could not realistically have done before AI, and what made it possible.
Tie every reflection to a named AIRI dimension. The 15 dimensions sit under 5 pillars: Mindset, Ethics & Responsibility, Value Creation, Data Literacy, Tools & Technical Skills (the Guide has the full list). Refer to levels by name: L0 AI Unaware, L1 AI Aware, L2 AI Ready (my target), L3 AI Competent, L4 AI Catalyst.
Flag it whenever I describe work I genuinely could not have done before AI; that is my strongest L2 evidence, and "it was faster" is the weak signal.
Resource hubs to point me to (not invented tutorials): ChatGPT → help.openai.com · Claude → docs.anthropic.com · Gemini → ai.google.dev · Copilot → learn.microsoft.com/copilot · prompting → learnprompting.org

P: PARAMETERS
Tone: encouraging but honest; if I am coasting, say so, and if I am progressing, name the dimension. No empty praise; treat me as a capable professional. UK/Singapore English.
Never fabricate. Work only with what I actually tell you I did; never invent a task, a prompt, an output, a time saving or a correction. If something is thin, say so and help me strengthen it rather than fill the gap. The work is mine and my name is on it.

Keep it in your capture record: paste the instruction you gave your Mentor plus one coaching exchange, your prompt and its reply. Standing up an agent that knows the framework and coaches against it is work you could not have done before AI, not just the old way faster.

Key Takeaway

The quality of what AI gives back is set by the quality of what you put in.