ICIO AI Prompt Framework: Instruction, Context, Input, Output
The ICIO AI Prompt Framework helps you write a prompt as Instruction, Context, Input Data, and Output Indicator, so the model has the task, the situation, the source material, and the required shape. It is a checklist, not a promise of better answers.
Framework Card
- Name:
- ICIO AI Prompt Framework
- Goal:
- Build a prompt as Instruction, Context, Input Data, and Output Indicator so the model has the task, the situation, the source, and the required shape.
- Flow:
- Instruction → Context → Input Data → Output Indicator
- Best For:
- A ChatGPT-class task that came back in the wrong format; Rewriting a one-line wish into a complete brief; Giving the model source text it must use rather than invent
Why it matters
A one-line wish (“write something about this”) forces the model to invent audience, facts, and format. You then spend the next turns correcting guesses.
ICIO exists to make those four decisions explicit before you send. If you cannot name the source text, the model will fill gaps. If you cannot name the output shape, you will get an essay when you needed a table.
What it is
ICIO is four prompt parts:
Instruction: what to do (verb and object), specific enough to start.
Context: why, for whom, constraints (internal memo, reading level, do not mention X).
Input Data: the material to use (paste, excerpt, structured facts). This is not the same as Context.
Output Indicator: shape (bullets, email, JSON-like outline), length, tone.
It is a brief for a generative model, not a method for training models, and not STAR (Situation, Task, Action, Result for a past story).
How it works
1. Instruction
Write the task in one direct sentence.
2. Context
Add audience, purpose, and constraints. Do not dump source text here.
3. Input Data
Paste or attach the material the answer must use. If you have none, say so and expect labeled assumptions.
4. Output Indicator
Name format, length, and tone. Then send. Revise a missing part rather than adding adjectives like “good.”
How it compares
| Framework | What it helps you see | How it differs |
|---|---|---|
| CRISPE AI Prompt Framework | Capacity, Role, Insight, Statement, Personality, Experiment | A different prompt checklist. ICIO is four parts. |
| COSTAR AI Prompt Framework | Context, Objective, Style, Tone, Audience, Response | Overlaps Context and Output. Different letters and jobs. |
| BROKE | Background, Role, other prompt parts | Another acronym family. Not ICIO. |
| STAR | Situation, Task, Action, Result | Past-incident story, not a generative brief. |
ICIO is the lens for Instruction, Context, Input Data, Output Indicator. Other methods help when you already standardized on another prompt acronym or need a story format.
When to Use This Framework
- Wrong shape. You wanted a table or a 120-word brief and got a blog.
- Invented facts. You had source text and did not give it, so the model filled in.
- Audience mismatch. The draft is for the wrong reader because Context was missing.
Example
A concrete example makes the structure easier to reuse when you are under uncertainty.
Example: Weekly ops email from notes
- Instruction: Draft the Friday ops email.
- Context: Internal, 40-person product org, no customer names, calm tone after a missed launch date.
- Input Data: pasted bullet notes (what shipped, what slipped, next week’s one bet).
- Output Indicator: 150 to 200 words, three short sections: shipped, slipped, next week.
Implication: Without Input Data the model invents a launch story. Without Output Indicator it writes a pep essay.
Example: Compare two vendor quotes
- Instruction: Compare the two quotes and recommend one.
- Context: You are the operations lead. Budget cap is stated. Prefer implementation time over brand.
- Input Data: Quote A and Quote B figures (term, price, start date).
- Output Indicator: A four-row table plus a five-sentence recommendation.
Implication: Context (budget, preference) is not the quote text. Mixing them is how numbers get dropped.
Takeaway
What ICIO can help with
- Seeing which of four prompt parts is missing
- Separating situation (Context) from source (Input Data)
- Specifying output shape before generating
- Reducing guesswork from one-line wishes
What ICIO cannot replace
- Domain knowledge. A complete prompt can still be wrong on facts you did not supply.
- Other prompt checklists. CRISPE, COSTAR, and BROKE use different letters. They are cousins, not this page.
- STAR. STAR structures a past example. ICIO structures a model brief.
- Model training or eval design. This is an end-user prompt skeleton.
Honest scope: ICIO structures a brief. It should not be sold as a productivity system.
Frequently asked questions
Context is situation and constraints. Input Data is the material to process. A URL or pasted report is Input. “This is for a board pack” is Context.
A tiny task can run with Instruction and Output Indicator. When answers go wrong, the missing part is usually Context or Input Data.
Only up to the facts you provided. It does not create knowledge you omitted.
No. They are other checklists. Use one set of letters per prompt unless you are translating on purpose.