Small vs large selections: how analysis scales

What changes when your scope grows past roughly a hundred discussions — the preparation wait, the progress bar, and the deeper evidence drill-down.

The Co-Pilot analyses a handful of discussions differently from a few hundred. Small selections are read directly and answered immediately; large selections are prepared first — a one-time wait — and then analysed through emergent themes with a deeper evidence drill-down. Knowing which mode you are in explains both the wait and the extra buttons.

Small selections: direct and immediate

When your scope covers a modest number of discussions — a single discussion, a session, a small event — the Co-Pilot reads the contributions directly and starts answering right away. Every claim in the answer carries its numbered citation chips inline, and each chip jumps straight to the contribution it came from. All the evidence for a claim is already in front of you.

Large selections: prepared once, then analysed by theme

Once a selection grows past roughly one hundred discussions, reading everything into a single answer is no longer possible. The app switches to a prepared analysis:

Preparation starts automatically the moment you select the large scope. A card titled Preparing this large selection appears with a live progress bar — "12 of 480 discussions ready" — and an estimate like "about 20 min".
The message box waits. While preparation runs you cannot send yet: "The assistant can start once this large selection is ready."
You do not have to watch it. Leave the page and keep working — the bell notifies you with Large selection ready when the Co-Pilot can start. Preparation is one-time: the next analysis on the same selection starts immediately.

If something goes wrong you see This selection needs attention instead — the discussions stay browsable, but the assistant needs the preparation to finish before it can work with the full selection.

After preparation, answers are built from emergent themes: the app clusters what was actually said across all discussions and reports the patterns, rather than quoting its way through hundreds of transcripts one by one.

The evidence drill-down — a large-selection feature

On large selections, a claim in an answer stands for many voices, not one. That is why some citations gain a More evidence link that opens the Evidence for this claim panel, with Supporting contributions and Pushback & different views tabs — the fastest way to judge whether a theme reflects broad agreement or a single loud voice.

On small selections you will not see this panel, and nothing is missing: every piece of evidence is already cited inline.

Saved groups make large analyses repeatable

Hand-picked large selections pair naturally with saved groups: save the selection once with Save as group, and the app prepares it in the background — the group's card shows Preparing saved group with the same discussion count and time estimate. Once ready, loading that group into any chat skips the wait entirely.

Choosing well

Pick the smallest scope that can answer your question. A session-level question answered at project level is slower and blurrier, not better.
Go large deliberately — for synthesis questions like "What are the main themes across the whole project?", the prepared, theme-based analysis is exactly what you want.
Reuse prepared selections. If you will ask several questions of the same large selection, save it as a group first.

Last updated: 2026-07-13

Wasn't helpful? Tell us what's missing