The AI summary reads like a brief

The AI summary now reads like an analyst's note: the one thing to know in a sentence, then findings that each lead to the components behind them — and every finding can take you further, to the chart it cites, to Axi, or to an alert. It also remembers: what changed since the last summary, the summaries before it, and the findings you pinned.

The story first

A summary opens on its headline, in larger type. Under it, each finding starts with a few words in bold that say what it is about, then one sentence with its numbers; a small mark gives its angle — what moved, what looks off, how things compare, the state of play. Every component a finding names is a link: hover it and its card lights up on the dashboard, click it and the dashboard scrolls there.

The title line says when the summary was written and by whom — 2 hours ago by Marie — and, when it's due a refresh, why: Board changed, Setup changed, Outdated, or written 12 days ago once it's over a week old. At the bottom, Read 7 widgets and a note · This year · France or Germany opens into what it read, what it skipped and why, and the language and model it was written with. Refresh sits beside it — in View mode too, when you point at the summary.

Set it up from its pencil

The summary's setup moved behind its pencil, like every other component's: the angles it covers, What Axi should know — up to 500 characters on what the numbers can't say on their own, which Slack reports read too — the language it's Written in, the same for every reader, and Sent — the Slack reports that deliver it.

A finding does things

Point at a finding and its buttons appear. Go to the widget takes you to the component it cites. Why? opens Axi with the finding attached and asks what's behind it — Axi answers from the dashboard, and says so when the numbers have moved since. Watch opens the Alerts panel on the KPI or chart it cites.

Your own view, your verdict

In View mode, when you've changed a filter for yourself, a line above the headline says what the summary was written for and what your view shows — and Ask Axi to brief you on it asks about your view instead. Fold the summary to one line when you'd rather it stayed out of the way: its name, its headline, when it was written. Or turn on Show on the board: each finding takes a number, and that number appears on every component it cites, so the summary reads as a layer over the dashboard. Two thumbs at the bottom say whether it helped; Something is off asks what, so summaries can get more accurate.

It remembers

A new summary is written knowing the last one: the findings it didn't have wear a small New mark the first time you see it, and one line at the bottom says what it no longer says. Earlier summaries are kept — a tab's 20 most recent, the Monday report's included — and ‹, at the bottom, steps back through them, each read as of its day. In Edit mode, Pin keeps a finding at the top of the summary and has Axi check it again every time: — still holds, or — no longer: and why.

📖 AI summary · What you can do with a finding · Set up the summary · Pin a finding · Show it on the dashboard · Earlier summaries · Where to find Axi