If you use LLMs for research, coding, brainstorming, writing, or feedback, you have this “lottery” feeling. Will it give you what you want or will it take you in a pointless direction? A portion of the problem is to learn how to prompt correctly. However, many times, the best prompt still delivers unusable results because of the lack of context.
Watching savvy software engineers, marketers, and other professionals use AI has clarified to me that providing context before you ask the LLM to do something is the secret to success. Which made me ask this question: Is PRFAQ the holy grail of context?
PRFAQ as Context
It sounds obvious when you see it, but the difference between a good and a great professional is that the latter insists on understanding the “why” behind a task before they work on it. A great software engineer understands the customer impact of a feature they are coding. A great content marketer understands the content strategy behind a blog post they are writing. A great design doesn’t simply design a screen. They understand the flow before and after, and the job the user is getting done.
PRFAQ is the big picture behind these and more. A well-written PRFAQ clearly articulates a vision and a strategy. That is the “instructions” that lead LLMs to perform orders of magnitude better than a raw prompt.
Here are four uses cases a PRFAQ will help you use LLMs better.
PRFAQ for Brainstorm
I use LLMs to research and brainstorm all the time—as simple as ideas for articles to write to approaches to technical architecture for a project. Recently, I had a superb example of the power of the PRFAQ. I wrote a draft of a PRFAQ for a new startup. I was collecting ideas on how to make the product viral. You don’t include that type of detail (feature lists) into a PRFAQ, so I’ve been keeping it in another document. I uploaded my PRFAQ to ChatGPT and ask me for ideas on how to improve virality. It gave me several suggestions that were fitting for the startup strategy, most of which I already had on my list, but one of them was brilliant and I hadn’t thought about it.
Without the context provided by the PRFAQ, ChatGPT would give me the average list of viral features, like “make it easier for users to share links on social media.” Blah. With the PRFAQ, it understood a specific nuance of the product and came up with a brilliant suggestion.
PRFAQ for Writing
I’ll be honest, I still struggle to get ChatGPT or Claude to write decent articles for me. The style, tone, content and everything else feels off. However, it gives me excellent ideas for opening paragraphs, suggest titles, topics to cover, etc. Sometimes, I ask it to provide a draft and I use it for inspiration, other times I ask to give me bullet points.
When I ask it to write a Product Requirements Document (PRD), a newsletter, a marketing plan, or even a boilerplate for a job description, it does a fantastic job when I provide the PRFAQ as context. This does not differ from you hiring a new employee and asking them to do the same job. Without the vision and strategy context, you’ll be going back-and-forth with your employee for weeks until they get it.
PRFAQ for Coding
Coding is not simply writing the code. It requires a technical design, an architecture, planning, and many more steps. The idea that you’ll tell an LLM what you want your app to do in plain English and it’ll magically write the entire code only works well for simple (and greenfield) projects, but not much beyond that. Every software engineer will tell you their frustration with “vibe coding.”
It’s almost a given that you should provide your AI Agent with a PRD that describes the feature you want to build. A well-written PRD will start with a good context for the user's goals, the problem, the desired outcomes and benefits, plus all the functional specs and behaviors. However, a PRD is only a slice of the full value for the end user and for the business. The PRFAQ + PRD combo delivers full context to the AI about the endgame.
PRFAQ for Evaluating
I don’t have the data to back it up, but I think one of the least used and most useful features of an LLM is to critique your work. You literally ask “Critique my article” and upload your document and the LLM will do a fantastic job of not only providing language and style feedback but also about cohesion and even point out flaws in your arguments. That’s great if you are asking to provide feedback on general knowledge, but not so much if the context of the work matters.
Asking AI to critique a PRD, a newsletter, or a marketing plan without context results in a middle of the road approach that makes whatever you are doing unremarkable. When you provide a PRFAQ and whatever else you want to evaluate, the LLM will have a specific context to understand what are the strengths and weaknesses of your artifact are.
Conclusion
Providing context before you ask an LLM (or a person) to do a task gives them the environmental setting they need to do a great job. A PRFAQ is a perfect way to provide context around a vision and the strategy for a product, business, or other types of innovation. It works during the research phase (and brainstorming), execution (writing/coding/planning), and evaluation.