flownote / AI meeting notes
Short answer
For an in-person meeting, place a consented phone or computer centrally, reduce room noise, select the correct microphone, keep the device powered, and capture the conversation. Afterward, review speaker labels, names, decisions, and tasks before sharing a recap.
Editorial note: This guide was prepared with AI-assisted research, reviewed against the cited public documentation, and uses no customer recordings or private meeting data. Examples are explicitly fictional.
Why room-based meetings are different
Online meeting audio arrives as a relatively direct signal. In a room, one microphone must hear people at different distances while handling air conditioning, keyboard noise, side conversations, and echo. Placement matters as much as the transcription software.
The best setup is deliberately boring: one stable device near the center, a clear announcement, a short audio test, and a backup note of the decisions that matter most.
Room setup checklist
- 1
Choose the device
Use a supported phone or computer with enough battery and storage for the full meeting.
- 2
Place it centrally
Keep the microphone uncovered and away from laptop fans, projectors, coffee cups, and people typing.
- 3
Ask for consent
Explain that the meeting will be transcribed and summarized, who will receive the result, and what happens if someone opts out.
- 4
Test every side of the table
Ask people at the nearest and farthest seats to speak, then confirm that the audio level is usable.
- 5
Name important transitions
Say decisions and owners clearly: “Decision: we will use option B. Ana owns the revised timeline.”
- 6
Review before distribution
Correct speaker labels and ambiguous passages. Do not treat automatic identification as definitive.
Improve speaker separation
- Ask one person to speak at a time when decisions are being made.
- Have attendees introduce themselves early so names appear in context.
- Move the device closer for a small group instead of turning the microphone gain to an extreme.
- Repeat action items at the end with each owner’s name and due date.
- For a large room, use the room’s approved microphone system if your capture workflow supports it.
Overlapping speech and distant voices can make speaker segmentation uncertain. A reviewer should resolve important attributions from context or ask the group rather than guessing.
Match the output to the meeting
| Meeting | Useful output | Review closely |
|---|---|---|
| 1:1 | Topics, commitments, follow-up questions | Sensitive personal details and agreed confidentiality |
| Team sync | Updates, blockers, decisions, owners | Similar names and rapid handoffs |
| Client update | Progress, decisions, risks, next steps | Commercial commitments and dates |
| Interview | Themes, evidence, follow-up questions | Consent, access, and hiring-policy requirements |
| Workshop | Ideas grouped by theme plus final choices | Distinguishing proposals from decisions |
Flownote provides purpose-built output styles for several of these conversations. Select the closest style before or after capture, then edit the result into the shared record your team actually needs.
A safe fallback when recording is not appropriate
If a participant declines, the topic is unusually sensitive, or policy is unclear, stop the recording workflow. Assign a human note taker to capture only decisions, actions, owners, due dates, and open questions. Send that short recap for confirmation.
Frequently asked questions
Where should I put the phone in an in-person meeting?
Place it on a stable surface near the center of the group, away from fans, typing, and other noise. Test the farthest participant before starting.
Can AI reliably identify every speaker in a room?
Not always. Distance, similar voices, interruptions, and background noise can make labels uncertain. Review important attributions manually.
What if one participant does not consent?
Do not record them. Use manual notes or another approach approved by everyone and consistent with your organization’s policy.
Sources checked
Related guides