The read, not just the list
Open a project and Tebelis gives you a short read of what's going on — not just how many records are open, but the patterns underneath them. It looks at how work is moving (what's stuck, what's been sent back or rejected), where it's concentrated (which locations or templates keep coming up), and how your own key field values are trending over time.
Where the numbers come from
Every figure in the briefing is computed from your records — the same data the charts read. The assistant doesn't invent anything: it reads a small, pre-aggregated summary — counts, trends, and a short list of the records, people, and locations involved — and writes the plain-language version, with links to the records it mentions. It works from that summary rather than the full contents of your records, and the cost doesn't grow with how many records a project has.
Trends it can validate
For the fields you mark as outcomes — in a template's Analytics tab — Tebelis can track how an adverse value trends week to week, and check whether that value actually lines up with bad outcomes in this project's own history. When it does, the briefing says so plainly ("confirmed against your data"); when it's only a guess, it stays hedged. It always speaks in terms of adjacency — what tends to occur together — never a proven cause.
What it flags
The "Flagged by copilot" panel lists what's worth a closer look, most pressing first: a rising adverse trend, records stuck too long in review, a location that keeps coming back across send-backs and rejections. Each item links straight to the records behind it.
Honest by design
The free, deterministic read always runs — the AI narration is a layer on top of it. If AI is paused or over budget, you still get the plain triage. And everything respects who you are: you only ever see the projects and records you're allowed to.