AI Assistant Prompt Library — Collaboration Tasks

Inputs: meeting notes or transcript; meeting date; attendee list; priority tags (optional).

Instruction (to assistant): "Read the meeting notes below. Produce a concise summary (2–4 sentences), then an ordered list of action items. For each action include: task, owner (if inferred), due date (if mentioned), priority (high/medium/low), and one-sentence context. If an item is ambiguous, label it 'Needs clarification' and list the specific question to ask the team."

Safety/Checks: Ask the model to mark any items sourced from uncertain or unsourced statements. Require that the assistant appends a short confidence score (low/medium/high) for each action and a list of exact phrases from the notes it used to infer the action.

Inputs: decision description, participants, rationale, alternatives considered, timestamp.

Instruction: "Format a decision log entry with: Title, Decision Statement, Rationale (short), Alternatives Considered, Deciders, Effective Date, Next Review Date (if any). Keep language neutral and cite the sentence(s) from the input that justify the rationale."

Safety/Checks: Ensure the assistant does not invent deciders — if not explicit, prompt it to mark 'decider unknown.'

Inputs: recipient role or name, purpose, key points, desired call-to-action, tone (friendly/formal), length limit.

Instruction: "Draft an email with a clear subject line, a 2–3 sentence opening that states the purpose, 3 bullet points with supporting details, and a single clear call-to-action. Provide a subject line and two alternative opening lines with slightly different tones."

Safety/Checks: Add a checklist at the end of the draft: GDPR/privacy checks, claims that require verification, and any phrases that could be misinterpreted.

Inputs: decision question, options (short list), recommended option, supporting data, risks, stakeholders impacted, timeline.

Instruction: "Create a one-page decision memo with: Purpose, Background (3–4 lines), Options (with pros/cons), Recommendation (with 2–3 supporting facts), Implementation risks & mitigations, and Next Steps. Use headings and bullet lists for readability."

Safety/Checks: Require sources or data references for supporting facts; flag any recommendations that rely on missing data.

Inputs: research question, allowed source types (peer-reviewed, internal docs, dated news, blogs), date range, desired output length.

Instruction: "Provide a 150–300 word brief summarizing the current state of knowledge, 3–5 bullet-point implications for our team, and 3 reliable sources (with URLs or citations). Indicate confidence for each implication and note where evidence is weak or mixed."

Safety/Checks: Insist on citations. If no reliable sources are found, produce a 'what we don't know' section and suggested next steps to obtain primary data.

Inputs: question, list of internal documents or folders to prioritize, whether to include external sources.

Instruction: "Answer the question using only the listed sources. For each claim, cite the source (doc title, section, page). If the sources don't answer fully, return 'Insufficient internal evidence' and propose 2 targeted follow-up queries or data requests."

Safety/Checks: Strict source constraint prevents hallucination. The assistant should explicitly refuse to answer outside the scope.

Inputs: workshop goal, attendee roles, available time, desired outputs (e.g., prioritized list, prototype, decision), constraints.

Instruction: "Produce a time-boxed agenda with minute-by-minute segments, facilitator script prompts for each segment, required materials, and a 3-question evaluation to collect immediate feedback. Mark which segments require prework and who should prepare it."

Safety/Checks: Include accessibility and psychological safety notes (e.g., mechanism for silent input, how to handle dominant voices).

Inputs: brief description of the disagreement, stakeholders, examples of recent interactions, desired outcome.

Instruction: "Draft a neutral, facilitative script to surface assumptions and interests. Provide 5 questions a facilitator can use, and a suggested 10-minute exercise to elicit priorities. Add a short 'when to escalate' checklist."

Safety/Checks: Advise to avoid naming individuals in shared notes; recommend a private follow-up when sensitive information appears.