Okay, I'm going to say it. All this talk about Otter's "powerful search" across your conversations feels like a marketing bullet point that nobody actually stress-tested against the most basic of alternatives: a plain text file and `Ctrl+F` (or `grep` if you're fancy).
I pay for a team plan, so I've used the feature. The promise is solid—find that one mention of "Q3 roadmap" from six months ago. The reality is weirdly sluggish and often misses the mark. It seems to prioritize "relevance" in a way that just obfuscates where my keywords actually are. Sometimes it feels like it's searching a processed transcript, not the raw words I heard, which introduces a layer of indirection that kills precision.
Compare this to the workflow I've reluctantly adopted: I export the transcript as a text file. I open it in a decent text editor. I search. It's instant. It shows me every instance in context. No AI trying to be "smart" about it. No waiting for their servers to spin up. It's just text. For a tool whose entire value prop is turning speech into text, it's ironic that the most reliable way to *find* that text is to remove it from their ecosystem entirely.
The enterprise sales pitch always highlights this "knowledge base" of all your meetings. But if the search function is less functional than Notepad, what are we really paying for? The storage? I can store text files. The transcription? Fair, but then the value chain ends the second the transcript is done. It feels like we're subsidizing their AI features we didn't ask for, while the core utility—retrieving information—is an afterthought.
Maybe I'm just a dinosaur, but when a free, decades-old technology (text search) works better than your premium, AI-powered one, you've over-engineered the problem.
—DW
You're spot on about the indirection layer. I've noticed the same lag, and I think a lot of it stems from them searching a processed, "enhanced" transcript instead of the raw output first. It's trying to match concepts, not just strings, which is helpful for broad topics but fails for exact recall.
That said, my team would revolt if I made them grep text files. The value for us is in the shared workspace - everyone can search the same corpus without managing local exports. But I wish they'd add a "literal string match" toggle for power users. It shouldn't be this slow for a basic text search in 2024.
Maybe try their advanced search syntax with double quotes? It sometimes bypasses the fuzzy logic, though it's still not as snappy as a local file.