Insights that surface themselves

For two decades, getting an insight was pull. You went to the dashboard, scanned the tiles, noticed something off, and went digging. If you didn’t look, you didn’t know.

The emerging model is push: the number explains itself. The system reads what moved, works out the likely reason, and tells you, before you thought to ask. The dashboard stops being a place you visit and becomes a thing that speaks up when there’s something worth saying.

Why now

• Models can narrate a movement. A change used to be a number on a chart someone had to interpret. A model grounded in governed metrics can now describe it the way an analyst would: what moved, by how much, against what baseline, and whether it’s unusual.
• And propose candidate causes. Given the dimensions behind a metric, it can decompose a movement, rank the contributors, and say which most likely explain it: candidates, not verdicts, and the honest version always says so. A ranked, sourced shortlist turns a notification into the start of an explanation.

What it looks like

A regional sales lead opens her inbox Monday to a short brief that wrote itself over the weekend.

It doesn’t dump every metric; it leads with what changed. EU revenue is up ~4% week on week, driven mostly by one product line in two markets running a promotion. Below that, mid-market conversion slipped, likely cause flagged as a pricing change that shipped Thursday, stated as a candidate, not a fact. Each line links straight to the governed definition and the query behind the number.

The point isn’t the brief. It’s that she didn’t ask for it and didn’t have to know which dashboard to open. The system noticed what moved, offered a reason, and showed its work, so she checked the one finding that mattered in under a minute.

Where it’s heading

Toward continuous narration tuned to each role. The finance lead and the support manager care about different movements in the same data, so the system learns what each role attends to, drawing on the shared memory of what the business has flagged as important before, and narrates accordingly. The cadence loosens too: instead of a fixed Monday digest, a quiet running commentary that surfaces the meaningful change when it happens and otherwise stays silent. The hard skill isn’t talking more; it’s knowing when to stay quiet.

How we think about it

An insight without a cause and a source is just a notification. Anything can tell you a number went up; that’s the easy, useless part. An insight earns the name only when it carries a candidate explanation and a link to verify, so the reader can act or check rather than take it on faith. A confident narration with no traceable basis is worse than silence, because someone will believe it, the same standard we hold for proactive alerting: reach the right person with context, not noise.

Meet that bar and the inbox stops being another feed to triage. It reads like a colleague who went through the data over the weekend and tells you the one thing that changed, and why.