Product · Trust
Why Sprig does not send a bot to your meeting
Every meeting recorder eventually ships a bot that dials into the call. We decided not to, and the reasons are about consent and audio quality rather than engineering effort.
Every meeting tool arrives at the same fork. You can record the conversation from a device in the room, or you can send a participant that is not a person: a bot that joins the call, sits in the participant list, and captures the stream.
The bot is the easier product. It works without anyone changing how they run a meeting, it scales to calls nobody on your team attends, and it demos beautifully. We are not building it, and this is the honest version of why.
A bot changes what consent means
When a person presses record, the people in the room can see who did it and can ask them to stop. The decision has an owner, and that owner is in the conversation.
A bot moves the decision somewhere else. It is usually configured once, by someone who is not in most of the meetings it will attend, and it then joins calls on a calendar rule. The people in the room meet a participant named after a vendor and are left to work out what it is doing and who to ask. Some of them will not ask.
That is a worse consent story, and no amount of interface can fix it, because the problem is not that people were not told. The problem is that nobody in the room decided.
The person in the room starts the recording and stops it. That constraint costs us features regularly, and we have kept it every time.
A bot is a worse recording
This part is less philosophical and easier to measure. A bot receives the conference platform's mixed, compressed output stream. Some of the things that stream has already thrown away are exactly the things speaker separation needs:
- Aggressive noise suppression removes the room tone that helps distinguish two voices with similar pitch.
- Automatic gain control normalises volume per speaker, flattening the loudness differences that carry who is near the microphone.
- A single mixed channel means two people talking over each other arrive as one waveform, already summed.
Recording from a device in the room gets the microphone before any of that happens. It is not a small difference. Attribution is the thing Sprig exists to do, and the bot architecture starts by degrading its input.
What we give up
Being straight about the cost:
| What a bot does well | What Sprig asks instead |
|---|---|
| Joins a call nobody from your team attends | Someone in the meeting starts the recording |
| Records a meeting scheduled while you are away | The meeting is recorded when a participant chooses to |
| Works with no change to anyone's habits | The recorder has to be started, on a phone or a laptop |
If your requirement is coverage of meetings your team is not in, Sprig is the wrong tool and we would rather you found that out here than three weeks into a trial.
The rule underneath
Attribution before summary. A feature that cannot say who said something does not ship, and an architecture that makes attribution harder does not get built for the convenience it buys elsewhere.
That rule is why there is no bot. It is the same rule behind refusing to generate a claim that cannot cite the passage it came from, and it is the reason the pricing page says plainly that nothing is purchasable yet.
You can read the longer version of what Sprig does and does not do on the about page.