22 August 2026
Most of what a camera records in a meeting is never watched again — the useful part is the text, not the video. The realistic policy is to transcribe everything, keep the transcript on a long or indefinite schedule, and let the video expire on a short timer unless something specific — a dispute, an audit, a legal hold — says otherwise.
Somebody hits record, the call happens, and the file lands in a folder that gets opened twice: once to check nothing broke, once maybe six months later to settle an argument nobody remembers clearly. In between, the video does nothing. It costs storage, it sits behind a login, and unless it's flagged for something specific, nobody is coming back to press play.
The transcript is a different object entirely. It's text — searchable, small, and it holds every decision and commitment that made the meeting worth recording in the first place. Treat the recording and the transcript as two assets with two different lifespans, and the retention question gets a lot easier to answer.
A workable default looks like this: transcribe every meeting as it happens, keep the transcript on a long or indefinite schedule, and put a short, fixed expiry on the raw video unless a specific reason extends it.
The 90-day auto-delete you set up for video assumes nothing is ever in dispute. A legal hold breaks that assumption on contact: the moment litigation or an investigation is reasonably anticipated, every recording and transcript that falls inside the relevant date range or names the relevant people has to stop moving toward deletion, whatever the policy says.
This is where a tidy plan gets messy fast. Someone has to know which meetings are in scope before the automatic job runs, and video recordings are often the hardest part to freeze — many tools save the local recording to the device that made the call rather than to a central store, so pulling one back under hold means going to a laptop, not a server. A transcript kept centrally is the easier half of a hold to satisfy; the video is the half that takes a phone call to someone's IT department.
Run the numbers for an ordinary team and the gap between video and transcript gets concrete. A team recording, say, 150 hours of meetings a week at a modest compressed bitrate comes out to somewhere around 400 to 600 megabytes of video per hour — call it 75 gigabytes a week, close to 4 terabytes over a year, growing every week it isn't pruned.
The transcript for that same 150 hours, plain text with speaker labels, runs to a few hundred kilobytes a week — under a gigabyte for the entire year. That's not a rounding error, it's a difference of three orders of magnitude, and it's the reason keeping the text forever while the footage expires on a 60-day timer isn't a compromise. It's closer to the obvious answer.
Once the transcript is the thing you're actually keeping, what you do with it matters more than the fact that you kept it. A transcript sitting in a folder is barely better than a video sitting in a folder; a transcript that can be searched across every meeting a team has had is the actual point of keeping it. AVAY keeps that transcript centrally and lets anyone search across past meetings for what was said, by whom, in which call — which is the difference between an archive and a graveyard.
That doesn't remove the video question. If a team needs the footage itself — to review a demo, to check who was on screen, to satisfy a specific hold — someone still has to go get it from wherever it was saved, and for AVAY that's the participant's own device, not a shared cloud store. Plan the retention policy around that split: the transcript is the durable record, the video is a local, short-lived convenience that occasionally has to be retrieved by hand.
| Storage cost per year | Findability | Legal exposure | |
|---|---|---|---|
| Keep all video forever | Grows without bound, terabytes a year | High if search tooling exists | High — nothing pre-purged before discovery |
| Transcript kept, video expires fast | Small, transcript text is cheap to store | High — text search across every call | Moderate — clock starts on the delete date |
| Delete both on a fixed schedule | Lowest cost, nothing lingers | None — nothing left to search | Low if the schedule was followed before any dispute |
For almost every team, the transcript is enough — it's searchable, small, and holds the same decisions and commitments as the recording. Keep the video only when someone specifically needs to see tone, screen-share detail, or verify who said something, and even then a fixed retention window is enough.
There's no universal number, but 30 to 90 days for raw video is common for teams with no regulatory reason to hold longer — long enough to catch a dispute, short enough to bound storage. Transcripts are usually kept far longer, sometimes indefinitely, because they cost almost nothing to store and stay useful as a record.
A legal hold suspends the normal deletion schedule for any recording or transcript that falls inside the matter's date range or named participants, regardless of what your retention policy says. It has to stay preserved until the hold is lifted, which means someone needs to know which meetings are in scope before the automatic deletion job runs.
Not if your systems store them separately, which is how most transcription tools work — the transcript is generated once and kept independent of the underlying recording. Check the specific tool's settings, though, since some bundle deletion of both together by default.
It sits best with whoever already owns data retention for the rest of the company — legal, IT, or compliance — rather than being decided team by team, because a legal hold or audit won't care which team recorded the call. The policy just needs a default window for video, a longer one for transcripts, and a documented process for pausing it when a hold lands.
Treat the transcript as the record and the video as a temporary convenience: keep the text, put a short expiry on the footage, and know in advance what a legal hold does to that schedule.
AVAY is a video meeting platform that transcribes the call itself — no bot joins, because there is nothing to join. Start one at avay.ai, read how each part works in the documentation, or see what it costs.