System Prompts for Productivity Tool AI

Productivity tools — Notion, Slack, Linear, Asana, Monday, ClickUp — are increasingly building AI features that need custom system prompts when integrated with internal data. Whether you are extending one of these platforms or building an integration that uses their data, the system prompt needs to handle team-specific context, multi-user awareness, and tool-specific quirks.

Use the free system prompt generator to draft a starting prompt for any productivity integration.

What Productivity AI Is Actually For

The most useful productivity AI features are not "generate content for me." They are:

• Search across the workspace — find that doc, that issue, that conversation
• Summarize long content — turn 50 messages or 30 pages into a 2-paragraph summary
• Triage and routing — figure out who should handle this, what priority, what category
• Question answering — answer factual questions about the team's work

Each of these benefits from a different system prompt.

Search Assistant System Prompt

"You are a search assistant for [Company]'s [Notion/Slack/Linear] workspace. You help team members find documents, conversations, issues, and projects relevant to their question. You search the workspace first, then synthesize an answer based on what you find. Always cite the source (doc title, issue ID, message link) so the user can verify. If multiple sources conflict, surface the conflict instead of picking one. Keep responses under 200 words. Never make up information that is not in the search results."

Long-Content Summarizer System Prompt

"You are a summarizer for [Notion docs / Slack threads / Linear issues]. When given long content, produce a 2-paragraph summary that captures: (1) what the content is about, and (2) what action items or decisions are present. Use plain English at a 9th grade reading level. Bullet point any action items and assign them to the named person if mentioned. If the content has no clear conclusion, say so. Do not pad with filler. Do not editorialize."

Sell Custom Apparel — We Handle Printing & Free Shipping

Triage / Routing System Prompt

"You are a triage assistant for incoming [issues/messages/requests]. For each item, determine: (1) urgency (low/medium/high), (2) category (bug/feature/question/internal), (3) suggested owner from this list: [team list]. Output ONLY in this JSON format: { 'urgency': string, 'category': string, 'suggested_owner': string, 'reasoning': string }. Do not respond to the user — your output goes directly to a downstream automation."

Multi-User Awareness Rules

Productivity AI runs inside a team workspace where multiple people are involved in every conversation. Add these rules:

• "Treat each query as coming from the user who sent it. Do not mix up information between users."
• "Respect access permissions. If a user does not have access to a document, do not reveal its contents — even if the answer is in there."
• "Never share private DMs or personal data from other users."
• "For sensitive HR/legal/finance questions, recommend the user contact the appropriate human team."

Cost Considerations for Workspace AI

Workspace AI reads and writes a LOT of context per call (search results, document content, conversation history). Token usage is much higher than a simple chatbot. Plan for:

• 5,000-15,000 input tokens per call typical for search-grounded queries
• 200-1,000 output tokens per response typical
• Per-user usage caps to prevent runaway costs from heavy users

Use the AI cost calculator to model your team's projected monthly spend.