The recorder has to be sent to a meeting. It does not attend everything, and that is the correct default rather than a limitation.
Sending it
Open the event on the calendar and send Pippin Scribe to it from the detail panel. It joins at the start.
It works with Google Meet, Microsoft Teams and Zoom, and with NearSync's own huddles.
The status appears on the event as it changes, so you can see whether it is coming, whether it is in, and whether it has finished. If something has gone wrong with a connection, that is where you will see it rather than at the end of the call.
What people see
A participant in the list, with a name. Exactly like any other attendee.
Everybody in the call knows it is being recorded, because they can see the thing doing the recording. This is not a setting somebody can turn off.
Say so anyway. In many jurisdictions recording a conversation requires consent, and in every jurisdiction it is the difference between a colleague trusting the tool and resenting it. One sentence at the start costs nothing: this is being recorded and written up, tell me if you would rather it were not.
Anyone can ask for it to be removed, and when they do it should be removed without discussion.
Setting the meeting type
Before the meeting, on the event, choose a meeting type. One is suggested from the title, which is why titles matter considerably more than they appear to.
The type decides which framework the analysis applies and which outcomes get measured. Getting it right takes two seconds beforehand and cannot be recovered afterwards without re-running the analysis.
If nothing in the list fits, leave it blank. A generic summary is a perfectly good outcome; a discovery call scored as an engineering standup is a confusing one.
When it should not be there
One-to-ones about a person. Performance, difficulty, anything personal. A recorded one-to-one is a different conversation from an unrecorded one, and almost always a worse one, because the honest part stops happening.
Anything legally sensitive where your organisation has a considered position on recording.
When somebody says no. No further discussion required, and no note in the summary about why there is no summary.
When there was no recorder
A transcript can be pasted in afterwards and gets exactly the same treatment: summary, decisions, action items, outcomes.
That covers the cases the recorder cannot reach. A phone call. A meeting on a platform you do not control. A conversation somebody else recorded and sent you.
The analysis is only as good as what you paste in. A transcript with speaker names produces speaker-level analysis; an unattributed wall of text produces a summary and very little else.
While it is recording
Nothing to do. The call proceeds normally and nobody has to manage anything.
Afterwards the item appears in your Cockpit under Waiting, in the Distilling state, while the recording is transcribed and analysed. Longer meetings take longer, and the item moves itself into Ready to read when it is done, so there is nothing to check.
What it costs
Recording draws on your organisation's included hours, pooled across everybody: fifty a month on Standard, a hundred on Pro.
That is generous for normal use and it is finite. Two things consume it faster than people expect.
Recording recurring internal meetings by default. A weekly hour-long standup across a year is a substantial share of the allowance, and standup summaries are reliably the least-read documents any organisation produces.
Leaving the recorder in an empty room. If the meeting did not happen, the hours still ran.
The analysis afterwards spends credits on top of the hours. Both together still cost considerably less than one person re-watching the meeting, which is the honest comparison.
When something goes wrong
Four failures happen occasionally, and three of them are recoverable.
The recorder did not join. Usually a link that changed after the invitation went out, or a waiting room nobody admitted it from. The status on the event will show it. Paste the transcript afterwards if the platform produced one.
It joined an empty room. The meeting moved and the calendar did not. The hours are spent; there is nothing to recover.
The transcript has one speaker. The platform did not separate voices. The summary is still good; the per-speaker analysis is not, and the Intelligence tab will say so rather than attributing everything to one person.
Somebody objected mid-call. Remove it, and say in the follow-up message that the meeting was not recorded, so nobody is waiting for a summary that is not coming.
Before the meeting
Two minutes of preparation changes what you get out afterwards.
Write the real title. It drives the meeting type, which drives the analysis.
Set the type if the suggestion is wrong. Thirty seconds, and it is the only thing that cannot be fixed afterwards without re-running everything.
Check the attendee list. Speaker separation is better when the system knows who is expected, and the internal-versus-external labelling on the summary depends on it.
The rule that works
Record the meetings whose content will still matter in a month. Customer conversations. Decisions with consequences. Anything you will be asked about later by somebody who was not there.
Do not record the meetings whose content matters for an hour. Most internal coordination is in that category, and recording it produces a filing cabinet nobody opens.
Getting more out of it
Write the title before the meeting, not during. It drives the type, which drives the analysis.
Say names. Transcription attributes by voice imperfectly, and a conversation where people address each other by name produces markedly better speaker separation.
Say the commitment out loud. "I will send that by Thursday" gets extracted as an action item with an owner and a date. "I'll sort it" does not.
Did this answer your question?
No, ask a person