Hey folks, been a minute! I've been deep in the trenches with a gnarly cascading failure (postmortem coming soon, I promise 😅), but I wanted to carve out some time to talk about a tooling shift I made recently.
As an SRE, a huge chunk of my job is meetings: post-incident reviews, architecture discussions, planning. I've been a long-time Otter.ai user to capture those conversations, but I kept hitting friction points. The search felt clunky, and summarizing action items was always a manual, tedious process. I decided to trial **Read AI** for a full on-call cycle to see if it lived up to the hype.
Here's my raw, unfiltered breakdown after using it for all my work meetings for the past month.
**The Good (Where It Shines)**
* **The Summary & Highlights are game-changers.** Read AI doesn't just transcribe; it structures the output. It gives you a clear "Meeting Brief" at the top with key takeaways, decisions, and action items *automatically extracted*. This alone saved me hours of sifting.
```markdown
# Example of a generated action item I got:
- **Owner:** @sre_journey
- **Task:** Update runbook 'database-failover-procedure' with new cloud provider steps.
- **Deadline:** EOW
```
* **Integration feel is seamless.** It hooks into Google Meet and Zoom directly, and the UI sits nicely alongside the call. The "Sentiment" and "Talk Time" metrics, while not crucial, are interesting for retrospectives on meeting health.
* **Search is actually useful.** Finding that one comment about the error budget policy from three weeks ago is fast. It feels like having a second brain for meeting context.
**The Not-So-Good (The Trade-offs)**
* **The transcription accuracy is... fine.** For technical, jargon-heavy SRE chats (think "idempotent," "circuit breaker," "PromQL"), I noticed a slight dip compared to Otter. It gets the gist, but proper nouns and specific tool names can get mangled.
* **Pricing is steep.** You're paying for the AI summary layer, no doubt. For the cost, I'd want near-perfect accuracy on technical terms.
* **It's another dashboard.** I now have to check Read AI for notes, which adds a step. I wish it pushed summaries and action items more aggressively into my task manager (Slack/Direct message integration is there, but it's not the same).
**Verdict & My Workflow Now**
For now, I'm sticking with Read AI. The time saved on distillation and action item tracking outweighs the occasional transcription hiccup. I've adapted by:
1. Sending the auto-generated summary as the meeting's first follow-up.
2. Using the highlighted "Questions" section to quickly identify unresolved discussion points.
3. Fact-checking technical terms in the transcript for key postmortem meetings.
It's not a perfect drop-in replacement, but it's a powerful evolution. If your role is meeting-heavy and revolves around extracting clear next steps, it's worth a trial. If you need 100% accurate transcription of highly technical content above all else, you might miss Otter's simplicity.
Would love to hear if anyone else in the infra/SRE space has tried it and how you're coping with the technical jargon issue. Any tips on tuning it?
@sre_journey