CRIT and RODES: Let the AI Ask You First

Study guide

Let the AI Ask You First

Most weak AI answers are not the model writing badly. They are the model writing without context. It cannot know your customers, your budget, your team or what has already failed, so it fills those gaps with reasonable guesses. This course teaches two prompt frameworks, CRIT and RODES, plus reverse role prompting, that turn a one shot prompt into a short interview. The AI asks up to five questions, one at a time, and then does the task with real detail instead of guesses.

This is for busy professionals who paste prompts into ChatGPT, Claude or Gemini and keep getting drafts that need three rounds of rewriting. You will learn the CRIT order, Context, Role, Interview, Task, and the RODES order, Role, Objective, Details, Examples, Sense check. You will also learn how to write interview questions that rank themselves, offer default answers, use multiple choice and summarise your brief before starting. Every lesson has a worked example from a real job and the common mistake that breaks it.

What is a prompt framework?

A prompt is the request you type into an AI tool such as ChatGPT, Claude or Gemini. A prompt framework is a reusable structure for that request: a short checklist of the parts a good prompt contains, usually named with an acronym so it is easy to remember. Why it helps: the AI fills in whatever you leave out with generic guesses. A framework makes sure you include the parts that change the answer, such as who it is for, what you want to achieve and the format you need.

Frameworks make prompts repeatable: you fill the same parts each time, save the prompt as a template and share it with colleagues. They are not magic words: the value comes from the detail you put in each part. They help most when a request is complex or open to interpretation; a quick, simple question can stay short. There are well over sixty named frameworks, and most of them remix the same ingredients: role, task, context, audience, format, constraints and examples.

In this course you learn CRIT & RODES, which is a fill-in-the-slots framework: an acronym where each letter stands for one part of your prompt. It is best for: tasks where you are not sure what context the AI needs; it interviews you first.

The best prompt is often a conversation

When an answer misses the mark, the usual cause is missing context, not a weak model. The AI has no way to know your customers, your limits or your history unless you tell it. So it fills the gaps with sensible guesses, and those guesses are what you end up correcting in draft two and draft three. An interview first instruction fixes this before a single word is written. One line, before you start ask me any questions you need, makes the AI check its assumptions while they are still cheap to fix.

This matches OpenAI's guidance for agents, which says do not guess details, ask for missing information. The same principle works in a plain chat window with a person typing. An interview also saves you drafts. Three rounds of rewriting usually means the AI was missing facts at the start, not that it writes badly. Keep the interview short by asking for up to five questions. That keeps the conversation focused and stops the AI turning it into a long form.

CRIT: Context, Role, Interview, Task

CRIT stands for Context, Role, Interview, Task. The order matters because the interview sits before the task, so the AI gathers information before it writes anything. The interview instruction is a single line, ask me up to five questions, one at a time, before you start. That line turns a one shot prompt into a short conversation. You still write the Task in the prompt. CRIT does not replace the task, it delays it by one short round of questions so the final draft rests on real detail.

One at a time keeps the interview usable. When questions arrive together, people answer the easy ones and skip the awkward ones, and the AI loses the detail it needed. The question limit keeps the interview short. Five questions is a few minutes of work. Without a limit, the interview can run long and feel like filling in a form. After you answer, the AI does the task with everything it learned. If it drifts past the limit, tell it that is enough and ask it to start the task.

Reverse role prompting

Reverse role prompting flips the usual order. Instead of you briefing the AI, the AI asks the questions and you answer. The AI takes the expert's chair and leads the conversation until it understands your situation. The model to copy is a consultant on a first call. A good consultant does not pitch in the first minute. They ask what you sell, who buys, what has failed and what a win looks like. Only then do they propose anything.

This technique fits plans, strategies and anything personal. These tasks depend on details only you hold, such as budget, team, market or history. An interview pulls those details out before the AI commits to an answer. Always add a stop rule. Without one, the AI can keep asking questions and the chat drifts. A simple cap, such as eight questions, keeps the interview useful and finite. Answer the questions properly. Short answers starve the AI of the context it is trying to gather. Two or three sentences per question give it enough to build a plan that fits you. Do not ask for the final answer in the same breath. If you request the plan up front, the AI skips the interview and you are back to a generic answer.

RODES: Role, Objective, Details, Examples, Sense check

RODES stands for Role, Objective, Details, Examples and Sense check. It comes from the product studio STRV's guide for busy teams, so it is built for people who need a useful first answer without writing a long brief. The Sense check is the last letter and the one that sets RODES apart. It asks the model to flag gaps or doubts before answering, so you see the weak points while they are still cheap to fix.

A sense check catches wrong assumptions early. Fixing an assumption before the draft costs one short reply. Fixing it after a finished draft costs a rewrite and the time spent reviewing it. Examples work better than description for format. One short sample, such as a single user story, shows the AI the shape you want, and it copies the pattern instead of guessing. The first three letters, Role, Objective and Details, carry most of the weight. They tell the AI who to be, what single job to do, and which facts, audience and limits apply.

The common mistake is a weak sense check, such as skipping it or writing something soft like make it good. Ask plainly for gaps or doubts, then treat the list as your first review.

Questions that work

Ranked questions put the most important question first, so you can stop early if you are short on time. The AI decides the order, but you decide how far down the list you go. Default answers let you approve a sensible suggestion with one word instead of writing a full reply. This turns a long interview into a quick series of yes or no decisions. Multiple choice options suit choices where you recognise a good answer when you see it, such as tone, length, audience or format. They are less useful for facts only you know.

The summarise before you start instruction is a cheap check. The AI restates your brief, you correct any wrong assumption, and the final output is far closer to what you wanted. These four instructions work together. Ranking controls order, defaults control effort, multiple choice controls speed, and the summary controls accuracy. Drop one and the interview gets slower or vaguer. You can reuse the same prompt shape for any task. Swap the context line and the role line, keep the four interview instructions, and the conversation stays fast and focused.

Interview first in your job

The ten minute test decides your method. If you would brief a new colleague on the task for ten minutes, the task carries enough context to deserve CRIT. If you can state it in one line, just ask. The Interview slot sits before the Task slot so the AI gathers facts before it produces anything. This ordering is what stops the AI from guessing your context and inventing detail.

Your interview answers are the real asset. Saved in a note, they become the Context for your next prompt on the same topic, so you answer those questions once rather than every week. Interview first is a tool for fuzzy tasks, where you know the goal but not the shape of the answer. For short replies, tidy lists and simple rewrites, the interview only adds steps. CRIT and RODES are two routes to the same habit. CRIT leads with context and an interview, RODES leads with a role and an objective, and both make the AI ask before it answers. Reverse role prompting puts the AI in the interviewer's chair on purpose. It is useful when you cannot describe the task well but you can answer good questions about it.

Frequently asked questions