Helix could already answer from a meeting’s summary. It can now go back to the transcript when the summary isn’t enough, and quote the part that actually answers your question.
Excerpts, not whole transcripts
Meeting queries return the summary plus transcript metadata by default. When you ask for a quote, a timestamp, or what a specific person said, Helix retrieves only the excerpts it needs — a time range, a speaker’s lines, or a phrase search — instead of loading an hour of conversation into the answer.
Quoted lines keep their [MM:SS] Speaker: attribution, so you can tell who said what and jump to that point in the recording.
Honest about what it has
If a transcript isn’t available, a requested range falls outside the meeting, or a speaker label doesn’t exist, Helix says so rather than filling the gap with plausible-sounding dialogue.