Every brief started on Briefing Fox begins the same way: state who you are. Developer, marketer, analyst, researcher, student, artist. It is the Role component of a proper brief, and on this tool it isn’t optional advice, it’s the first screen.
If naming a role were what separates a brief that finishes from one that doesn’t, that should show up in who finishes. It doesn’t. Marketers who state their role still abandon 25.0% of the time. Analysts 23.1%. Programmers 22.7%. Researchers 20.8%. Students, the group you’d expect to struggle most, abandon least, at 11.3%. Naming a role is nearly universal here and the gap between the best and worst group is still more than two to one.
The Component Nobody Skips
This isn’t a knock on the role step. It’s the easiest one, which is exactly why everyone clears it. Picking “marketer” from a list costs nothing to think about. That’s also why it explains so little: two people can pick the identical role and mean completely different things by it, and the role field can’t tell them apart.
What determines whether a brief finishes happens one screen later. Every abandonment in Briefing Fox’s own usage data occurs at the same point: someone states a goal, is shown a set of follow-up questions, and never comes back. Nobody who reaches the end of the questionnaire quits before submitting. The role is stated before the drop-off. The drop-off happens on the questions that ask for everything else.
The Three That Are Actually Being Tested
Those follow-up questions are the other three components, doing the work role can’t. Context: the specific situation a generic marketer, analyst or researcher wouldn’t share with any other person in that role. Constraints: what the output has to avoid, not just what it should include. Output format: the exact shape expected back, not “a report” but which report.
A role tells the AI which hat to wear. It says nothing about your situation, your limits, or what you’re actually going to do with the output. Those three are harder to answer than picking a label from a dropdown, which is very likely why the questionnaire, not the role step, is where people leave.
One Role, Two Briefs
Take a task an analyst brings to AI constantly: summarise this quarter for leadership. As a role-only request, that’s the whole thing. Act as an analyst, summarise the quarter.
As a full brief, the role stays the same and three more answers get attached to it. Context: one metric moved sharply this quarter and the leadership team already knows the topline; they don’t need the recap, they need the explanation. Constraints: no bullet list restating numbers that are already in the deck, no speculation past what the data supports. Output format: one paragraph, written to be read before the meeting starts, not a slide.
Same role both times. The role-only version produces a competent, generic quarterly summary, the kind every analyst’s AI output looks like. The full brief produces something written for one specific meeting. The AI didn’t get more capable between the two versions. It got the other three answers.
What the Missing Three Cost
None of this shows up as a cost until the questionnaire, because the role step never asks for it. That’s the part of the product that flatters people: everyone clears it, everyone feels like they’ve briefed the AI. The actual briefing starts on the next screen, and it’s the only screen this data says people leave from.
Three Answers, Not Four
Before your next AI task, the role is probably the part you already have. What’s worth writing down first are the other three: the one fact about your situation a stranger in the same job wouldn’t share, the one thing the output must not do, and the exact shape you expect back. Three answers, because the fourth one, the role, is the part nearly everyone already gets right.
Try Briefing Fox free at briefingfox.com
FAQ
Does naming a role in an AI prompt guarantee good output?
No. On Briefing Fox, where stating a role is the first and unavoidable step, abandonment still ranges from 11.3% (students) to 25.0% (marketers). Naming a role is necessary but nowhere near sufficient; it says nothing about the other three components of a working brief.
What are the four components of a proper AI brief?
Role, context, constraints and output format. Role defines who the AI should act as. Context supplies what it couldn’t otherwise know about your situation. Constraints define what the output must avoid. Output format specifies the exact shape you expect back.
Why do context, constraints and format matter more in practice than role?
Because role is the one component almost everyone fills in without effort, a single choice from a short list, while the other three require actually thinking through your situation. In Briefing Fox’s own data, every abandonment happens on the questions that ask for those three, not on the role step that comes before them.
What should I have ready before I brief an AI?
Assuming you already know what role you need the AI to play, have three more answers ready: one specific fact about your situation a generic version of your role wouldn’t share, one thing the output must not do, and the exact shape you expect back.