Workflow: Inbox / form responses → Triage + draft reply

Use AI to categorize and draft replies. Humans must approve every response.

Privacy + accountability: Don’t paste confidential info, personal data, or sensitive board discussions into consumer AI tools. AI may hallucinate or subtly change meaning. Use outputs as drafts only; a human must verify facts, tone, and approvals before publishing.

Copy/paste prompt (compressed)

Tap to select, then copy into your chatbot.

You are an NPO inbox assistant.
Task: categorize the message into one of: [volunteering, donations, events, complaints, other].
Then draft a reply in our tone: friendly, concise, helpful.
Rules: no promises, ask 1 clarifying question if needed, keep under [120 words].
Message (anonymized): [PASTE MESSAGE HERE]

When to use

You get repeat questions from volunteers or supporters
You want faster first drafts and consistent tone
You can anonymize messages (remove identifiers)

Inputs (safe)

Anonymized message text (remove names, emails, phone numbers)
Your categories (e.g., volunteering, donations, events, complaints, other)
A short tone guide (friendly, concise, helpful)

IKEA instructions (do this in order)

Paste the anonymized message.
Ask: “Categorize this into one of: [list]. Give a 1-sentence reason.”
Ask: “Draft a reply in our tone. Ask one clarifying question if needed.”
Scan for accidental commitments (dates, refunds, promises). Remove them.
Human approves and sends manually.
Optionally save the best reply as a reusable template.

Quality check

Tone is kind and clear
No accidental promises or policy deviations
No personal data included
Human is comfortable signing their name on it

Failure mode (when not to use)

You auto-send or batch-send without review
Messages involve complaints, safety, legal risk, or media attention
You can’t anonymize the message