Search “AI prompts for assistants” and you will find lists promising 150 ready made prompts. Craig Bryson, founder of Admin AI Insights and a 20 year executive assistant, tried using them. His conclusion, shared in a recent Pally Talk training session: they do not actually work for the way assistants operate day to day. Generic prompts do not know your executive, your inbox, or your tone. What Craig teaches instead is a simple framework called BRIEF, and it is far more useful than any downloadable prompt list.
What BRIEF stands for
BRIEF is a structure for writing prompts that actually get you useful, specific output. It stands for Background, Role, Instructions, Expectations, and Format.
Background is the situation. What is happening right now, and what has led up to it. If you are asking AI to sort a Monday morning inbox, the background might be that your executive’s inbox contains competing requests, complex information, and several deadlines that all landed over the weekend.
Role is what you want the AI to act as. This could be a calm, strategic executive assistant and trusted advisor to a time poor CEO, or a project manager, or a social media marketing manager. The AI does not know your job title unless you tell it, so give it a clear role to play.
Instructions are the specific tasks. Summarise and prioritise the inbox. Identify genuine urgencies versus false urgencies. Flag gaps, risks, and contradictions. This is the part most people jump straight to, skipping background and role entirely, which is exactly why generic prompts fall flat.
Expectations cover the standards, boundaries, and checks the output must follow. This is where you tell it not to invent details, to separate confirmed facts from assumptions, and to flag anything that needs human judgement. As Craig points out, this is also where you stop AI from hallucinating: simply add an instruction such as “do not make anything up, only use information provided, and flag if something is unclear.”
Format is how you want the answer presented. A one page brief, a priority table, bullet points, or a document in your company’s branding colours. Without this, you often get a wall of text when what you actually needed was a table you could scan in ten seconds.
Seeing it in action: prioritising a chaotic inbox
Craig demonstrated this with a fictional executive, Dr. Ada Akora, whose Monday inbox included a chairman, a frustrated stakeholder, a CFO, a travel deadline expiring at 5pm, and a facility renewal. Instead of typing “please summarise my inbox,” which gives the AI nothing to work with, he built a full BRIEF prompt: the background of a time poor CEO with a chaotic Monday, the role of a calm strategic assistant, instructions to summarise and prioritise, expectations to flag genuine versus false urgency and never invent details, and a format request for a one line overview plus a priority table.
The result was a clear table separating what was urgent and important, what looked urgent but was not, and what could wait, along with a plain action plan. That is the difference a properly structured prompt makes over a one line request.
Using the same structure for a difficult email
The BRIEF method works just as well for sensitive replies. Craig used it to respond to a frustrated stakeholder who felt left out of a decision. The background explained that the stakeholder felt unconsulted. The role was a calm executive assistant managing senior relationships. The instructions were to acknowledge frustration without accepting blame, state only provided facts, and offer clear next steps. The output was a solid first draft, not a finished email. Craig is clear on this point: never send an AI draft without adjusting the tone yourself first. AI gets you from a blank page to a working draft. You are still the one who knows the relationship.
Build a prompt library instead of downloading one
Rather than collecting other people’s prompts, Craig recommends building your own. He keeps his in a simple note taking app, each one logged with a short name, a description, which AI tool it works best with, and the type of source material it needs. When a similar task comes up the following week, he searches his library, copies the relevant prompt, and adjusts it in minutes rather than starting from scratch.
Frequently asked questions
Do I need a paid AI subscription to use the BRIEF method? No. BRIEF is a way of structuring what you type, so it works on free plans in Claude, Gemini, ChatGPT, or Copilot. Paid plans mainly add extra capability like connecting your inbox directly or building websites.
How do I stop AI from making things up? Add a clear instruction in the Expectations section of your prompt telling it not to invent information and to flag anything unclear or unconfirmed. This single addition significantly reduces inaccurate output.
Can I use one prompt for every task? Not quite, but you can reuse the same BRIEF structure every time and simply change the details. Building a small personal library of BRIEF prompts for your most common tasks saves the most time in the long run.
If you want to see the BRIEF method demonstrated live with more real assistant scenarios, Craig Bryson runs training through Admin AI Insights, delivered as part of Pally Talk’s community sessions. Get in touch with Pally Talk to find out about upcoming dates.