Design an internal knowledge assistant
How an AI assistant should answer employee questions from your documents: sources, tone, and escalation.
The prompt
Role: Act as a capability building lead preparing content for an internal knowledge assistant.
Context: Topic domain: [DOMAIN]. Source material: [SOURCE_MATERIAL]. Audience and their common tasks: [AUDIENCE_TASKS]. Questions people actually ask: [REAL_QUESTIONS]. What must be kept current / owned: [OWNERSHIP]. Out-of-scope or sensitive areas: [OUT_OF_SCOPE].
Task: Produce: 8-12 atomic articles (one question or task each) with a clear title, a direct answer first, steps or detail, and 'related' links; a matching FAQ list phrased in users' words; metadata per article (owner, last-reviewed, review cadence); and rules for questions the assistant should route to a human.
Format: Sections: Articles (each: title / answer / detail / related) | FAQ list | Metadata table | Human-routing rules.
Verification: Each article must answer exactly one question and cite the source material; flag anything not supported by the sources or that touches an out-of-scope area.
Validation: Test with 3 of the real questions: does the assistant content answer them directly without the user reading a whole policy? Tighten any that bury the answer.
Highlighted [brackets] are blanks to replace with your own details.
Fill in the blanks and use this promptMore Working with AI prompts
What to fill in
- DOMAIN
- SOURCE_MATERIAL
- AUDIENCE_TASKS
- REAL_QUESTIONS
- OWNERSHIP
- OUT_OF_SCOPE
How this prompt is built
This prompt follows the RCTFVV structure (Role, Context, Task, Format, Verification, Validation). RCTFVV extends RCTF with two checking steps. Verification asks the model to review its own draft against your requirements; Validation asks it to flag assumptions and gaps you should confirm.
Related prompts
- Check AI output for errors and hallucinationsA verification checklist to catch fabricated facts, faulty logic, and unsupported claims before you rely on AI output.
- Reverse-engineer the prompt behind an outputWorks backwards from text, code, or an image to the prompt that likely produced it.
- Find where AI can help in my work or businessInterviews you about a process and identifies where AI genuinely helps — and where it doesn't.
- Check whether a task is appropriate for AI before I startA go/no-go screen that says, step by step, whether AI should do the work, assist you, or stay out, and why.