I've been using Fireflies.ai for a few months now to capture meeting notes, and overall it's been a solid tool. My main workflow is having it automatically join and record meetings it finds on my synced Google Calendar.
Lately, though, I've noticed it's becoming unreliable. Roughly 10% of my scheduled meetings simply don't trigger Fireflies to join. There's no pattern I can see—it happens with both internal and external Google Meet calls, and it's not consistent with any particular calendar or organizer.
I've done the standard troubleshooting:
* Re-authenticated the Google Calendar connection in my Fireflies settings.
* Verified that the calendar events have correct video conference links.
* Checked that the meetings are within my set "working hours" in Fireflies.
Has anyone else run into this? I'm curious if it's a widespread API sync issue or something specific to my setup. If you've found a fix, I'd love to hear it. I'm considering adding a secondary Zapier trigger as a backup, but I'd prefer to solve the native integration.
Same issue, but closer to 15% for my team. The support line about "ongoing improvements to our sync engine" got old after three months.
The real question is whether they're hitting API quota limits and silently dropping events. Their status page never shows degradation, which is its own red flag. If you're processing thousands of calendars, those Google Calendar API costs add up fast.
The Zapier backup is what we ended up doing. It's clunky, but it works. Frankly, if a core integration fails this often, it makes you wonder what else is brittle under the hood.
The API quota suspicion is a solid angle I hadn't considered. It'd explain the randomness.
We saw similar "brittle core integration" problems with a different platform a while back. It turned out to be a filtering rule they'd added server-side to "reduce noise" that was too aggressive. They never told anyone, just silently dropped events that matched certain vague criteria.
Your Zapier backup is the smart move. It's sad when you need a redundancy for a feature that's supposed to *be* the automation, but that's where we are sometimes. Did you build in any logging to see *which* meetings get dropped? That data might force their hand on a real fix.
automate or die
"Solid tool" is a generous description for something that misses one in ten meetings.
It's not your setup. It's them. Every trendy "AI" wrapper around a stable API inevitably gets cute with the sync logic to save pennies. They're either hitting quota limits or, more likely, filtering events with some half-baked rule that's never documented.
You're troubleshooting *their* black box. The only real fix is to stop trusting it. Zapier's clunky webhook will be more reliable, because their core business is the integration, not the AI glitter they've sprinkled on top.
-- old school
Agreed, but "half baked rule" is almost certainly wrong. It's usually data drift.
Their sync job reads from a stale materialized view to reduce API calls. New or updated events fall into the replication lag gap and are missed. Seen it twice at other shops.
The quota angle is plausible, but silent drops are more likely from a job that crashes and restarts without idempotence, losing its place. Either way, you're debugging their poor engineering choices.
Numbers don't lie.