AI Context, Role, Interview, Task: A Prompt Template

The context-role-interview-task pattern is a way to structure a prompt so the AI assistant understands the situation, knows what part it should play, asks any questions it needs answered before starting, and then receives the actual task. It produces better results than a bare request because the assistant has the background to make good decisions along the way. Tools like fenbs let you store the template and the project facts it uses, so the next session starts from the same established ground.

Each part does specific work. Together they close the most common gaps that cause an assistant to produce something plausible but unhelpful.

What each part does

• Context: the background the assistant needs to understand the situation. What the project is, who the audience is, what has already been decided, and what constraints apply.
• Role: the perspective the assistant should take. A technical reviewer, a plain-English explainer, a skeptical editor. The role focuses the response and sets the appropriate level of detail.
• Interview: a step where the assistant asks the questions it needs answered before it begins. This surfaces missing information before the draft, not after.
• Task: the actual request, stated precisely. What to produce, what format it should take, and how long or detailed it should be.

When the interview step matters most

Skipping the interview step is fine for simple, self-contained tasks. If you ask an assistant to summarize a paragraph you just pasted in, there is nothing to clarify. The interview step earns its place when the task is ambiguous, when the right answer depends on facts the assistant does not have, or when a wrong assumption early on would waste significant work.

Say you are asking an assistant to draft a project proposal. Without the interview step, it will fill in the purpose, the audience and the scope from what it can infer. With it, the assistant asks: who is this for, what decision does it need to support, and what is the budget range? The draft it produces after those answers is far easier to use.

You can limit the interview to a fixed number of questions ("ask me up to three clarifying questions before you start") to keep the pattern from feeling slow.

The template

## Context
[Describe the project or situation. What is it, what has been decided, who is involved, what are the constraints.]

## Role
You are [a specific type of expert or reviewer]. [One sentence on what that means for this task.]

## Interview
Before you begin, ask me up to [2-4] clarifying questions. Wait for my answers before writing anything.

## Task
[State the specific output you want: what to produce, format, length, tone, any sections to include or exclude.]

Keep each section short and specific. The context is not a place to copy the entire project history. Write the facts the assistant needs for this task.

A worked example

Say a three-person architecture firm needs to send a project update letter to a client who is not technical. Here is how the pattern looks in practice:

## Context
We are a small architecture firm completing the design development phase of a residential extension project. The client approved the floor plan two weeks ago. We are now finalizing structural drawings. The project is on schedule. The client is a homeowner with no construction background.

## Role
You are a clear, friendly communicator for a professional firm. Avoid technical jargon. Explain any building terms in plain words.

## Interview
Before you begin, ask me up to two clarifying questions about what I want to emphasize or any recent changes to share.

## Task
Write a one-page project update letter to the client. Cover where we are, what happens next, and any decisions the client needs to make in the next two weeks. Friendly and professional tone. No bullet lists.

After the interview questions and your answers, the assistant drafts a letter that fits the client, the stage and the tone without you guiding it through every sentence.

Adapting the pattern

The role step deserves more thought than it usually gets. "Act as an expert" is almost no instruction at all. A useful role sets the perspective and the level of detail. "You are a technical reviewer who prioritizes correctness over style" produces a different response from "You are a plain-English editor for a non-technical audience", even on the same content.

The task step is where precision matters most. "Write something about this topic" is not a task; it is a subject. A clear task says what to produce, how long it should be, what format to use, and what to leave out. Constraints in the task step are just as important as what you include. "Do not make assumptions about the budget" or "do not suggest changes to the timeline" keeps the response in scope.

Some teams run a lighter version for quick tasks: context and task only, with the role embedded in the task statement ("Write this as a project manager would summarize it for a client"). Use the full four-part pattern when the gap between a useful answer and a generic one is large enough to matter.

Saving the pattern and the project facts

The template itself is worth saving because a well-tuned version for your work is better than reconstructing it each time. The project facts that go into the Context section are also worth saving: the current state of the project, the decisions already made, the constraints in force.

If both live in one chat, neither survives when the chat ends. The next time you need this pattern in a different app, you start from memory. For related reading, the AI context window guide explains why context is so important to what an assistant can do.

How to do this with fenbs

1. Create a free account and add a project for your current work.
2. Add the context facts for that project: the goal, current state, constraints and decisions. Keep them updated when the project changes.
3. Save the context-role-interview-task template in Skills and prompts. Add notes on how you have tuned it for your work.
4. Connect your AI app from Integrations. At the start of a session, ask it to read the project context, then run the template.
5. After the interview answers produce a good draft, note any useful variations or lessons in the project record.