Alright, I've been putting tl;dv's AI notes through the wringer for the last two months, specifically on sales discovery calls. While I love the *idea* of unlimited AI notes, I have to say... the accuracy isn't quite there for nuanced sales conversations.
Here's what I'm seeing:
* **It misses subtle objections.** If a prospect says something like, "Well, our current process is clunky, but the team is used to it," the AI often summarizes it as "Client is satisfied with current process." That's a critical misread! The real note should flag the pain point *and* the adoption hesitation.
* **Action items get invented or misplaced.** I've had it generate "Send pricing page" as a next step when pricing wasn't even discussed. Conversely, actual agreed-upon actions ("I'll share a case study on X") sometimes get lost in a generic summary.
* **The "key points" can be too generic.** It'll pull out "budget," "timeline," and "decision makers" correctly, but the *context* around each is often watered down. For sales, the devil is in those details.
Don't get me wrong, it's fantastic for getting a rough transcript and a basic outline. It saves me from starting from zero. But I've had to stop relying on it as a single source of truth. My current workflow is: use tl;dv's transcript and AI note as a first draft, then I manually scrub through the recording to highlight the *actual* key moments and rewrite the summary.
I'm curious if others are finding the same? Especially those in sales or consulting where call nuance is everything. Have you found any tricks or prompt adjustments to make the AI notes more accurate for deal-specific contexts?
Maybe I'm just expecting too much from an AI not fine-tuned on sales data? 🚀
Automate everything.