I've been living in my observability tools for years, but my team is now asking me to evaluate meeting transcription and AI note-taking services to reduce our own operational overhead. Fireflies.ai keeps coming up. Like any good engineer, I went straight for the free tier to kick the tires.
I've set it up on a few recurring standup and post-mortem calls. The transcription accuracy is decent for our technical jargon, which is impressive. But I'm already sensing constraints that would make it unsustainable for even a moderately busy team.
What I really need from this community are the concrete, hard limits you've hit. Not just the published numbers on the website, but the practical walls. For instance, the 800 minutes of storage—is that a rolling window? Does deleting older transcripts free up space for new ones, or is it a hard cap on total lifetime uploads? More critically, the "limited transcription credits" on the free plan feels like a classic rate limit. How many minutes do you actually get per month before it stops? Is it a hard stop, or does quality degrade?
From my world, I'd instrument this like a service level objective. If the free plan is the "SLO," what's the error budget before you hit the wall and have to upgrade?
```yaml
spec:
plan: free
constraints:
- metric: storage_minutes_total
limit: 800
is_rolling: unknown
- metric: transcription_credits_monthly
limit: unknown
is_replenishing: true
```
Any data points on these limits, or other sharp edges like meeting length caps or export restrictions, would be hugely helpful. I'm trying to determine if the free tier is a viable permanent tool for a small team, or just a temporary trial that inevitably leads to a paid subscription.
-- owl
owl