AI Meeting Bot Consent and Summary Review Checklist for Small Businesses

Why this matters

AI meeting assistants can save time, but they also create new risks: attendees may not know they are being recorded, transcripts may expose customer or employee information, action items can be assigned from misunderstood conversation, and summaries can turn tentative discussion into a false promise.

STOP AUTOMATION: do not let AI invite a bot, start recording, distribute a summary, update a CRM, create tasks, promise next steps, or send customer-facing notes without a named owner confirming consent, source notes, attendee context, privacy boundaries, and follow-up wording.

Copy/paste meeting bot review card

Field	What to capture	Review rule
Meeting type	Sales, onboarding, vendor, support, HR/interview, internal ops, customer escalation, or legal/finance-sensitive call.	Do not use the same bot rules for every meeting type.
Attendee notice	Who was told the bot is joining, how they were told, and whether the platform visibly shows recording/transcription.	Follow company policy and applicable recording/consent requirements before inviting or recording.
Bot/tool details	Tool name, workspace, transcript owner, storage location, retention setting, sharing setting, and who can access the output.	Verify settings in the tool/admin panel, not from memory.
Sensitive topics	Pricing, contract terms, health/safety, employee issues, refunds, legal claims, private customer data, credentials, or payment details.	Pause bot use or restrict summary distribution when sensitive topics exceed approved use.
Summary proof	Original transcript timestamps, speaker labels, decisions made, open questions, and evidence for every action item.	Do not turn guesses, side comments, or unresolved questions into commitments.
Follow-up decision	Send summary, hold for owner review, redact, split internal/customer notes, create tasks, update CRM, or delete/export under policy.	Record owner, due date, redactions, and where proof was saved.

Safe notice and follow-up snippets

Pre-call notice: We use [tool] to help create internal notes and action items. Please let us know before the meeting starts if you do not want the AI note taker included. Sensitive details can be handled outside the bot-recorded portion.

Internal review note: AI summary is a draft only. Confirm speaker, timestamp, decision owner, deadline, and source quote before creating tasks, sending a customer recap, or updating CRM fields.

Customer recap guardrail: Send only confirmed decisions, next steps, and dates. Do not include private internal notes, pricing assumptions, unresolved options, attendee sentiment, legal/HR details, or AI-generated interpretations.

AI review prompt

Act as a cautious small-business meeting-summary reviewer. Use only the verified notes below. Do not invent consent, recording permission, attendee intent, decisions, deadlines, prices, contract terms, approvals, customer promises, legal conclusions, or action-item owners. Return: 1) missing consent/transcript fields, 2) sensitive topics to redact or hold, 3) action items with proof quotes, 4) unsafe summary claims to remove, 5) customer-safe recap draft, and 6) STOP AUTOMATION items.

Verified notes:
- Meeting type/date/tool:
- Attendees and notice/consent proof:
- Transcript/storage/access settings:
- Sensitive topics discussed:
- Confirmed decisions with timestamps:
- Open questions/unresolved items:
- Approved customer-facing next steps:
- Owner/reviewer:
- Proof location:
- Retention/deletion rule:

Fast QA before sending the summary

Confirm attendees knew the bot was present and that sensitive portions were handled according to policy.
Check the transcript for speaker confusion before assigning owners or deadlines.
Split internal notes from customer-facing summaries.
Remove private customer data, credentials, payment details, legal/HR speculation, and unapproved pricing or schedule promises.

Disclosure: Horizon Flow is Andrew Burton's digital product catalog. This worksheet is useful without purchase; product links are labeled and UTM-tagged.