A framework for prompting
Structure your prompts. Trust your output.
RICE gives every AI task a clear role, intent, context, and definition of done—so teams get useful first drafts, not costly guesswork.
The prompting brief
Give the model what it needs to succeed.
- R — Role
- Who should act?
- I — Intent
- What must happen?
- C — Context
- What informs it?
- E — Expectations
- What does done mean?
The four parts
Four things every prompt should say
01
Role
Set the expertise, perspective, and boundaries the work needs.
02
Intent
Make the desired outcome, depth, and audience explicit.
03
Context
Share sources, workflows, constraints, and consequences.
04
Expectations
Define the artifact, format, examples, and next action.
In practice
Same request. Better input.
From vague prompt to proof.
Before
“Make an executive dashboard from this spreadsheet.”
With RICE
- Role
- Executive reporting analyst.
- Intent
- Build a board-ready KPI dashboard.
- Context
- Use this spreadsheet; surface trends and anomalies.
- Expectations
- One-page, shareable HTML report.
Further reading
The complete RICE story.
RICE Framework for AI Prompting
Read the full explanation of the framework and its practical use in real prompting and agentic work.
Read the blog post