AI outcomes stats

Five AI commitments written plainly, each backed by a mechanism a buyer can verify.

A dated model release log with versions, change summaries, category chips, and one rolled back entry under a header with a changelog link.

A split section for skeptical AI questions, heading copy left and an accordion of five blunt answers right, first item opened.

Two column split pairing heading copy and three proof points with a cited answer card whose numbered markers resolve to listed source rows.

Centered heading over a comparison table whose ruled rows pit a generic chatbot against the built-in assistant, closed by a sourcing footnote.

Centered heading over a ruled table of six policy rows, each pairing a data commitment with a plain explanation, closed by a DPA link.

Eval pass rates per release as a shadcn line chart, including the regression we published.

The three gates every AI draft passes, model, checks, human, with no autopilot setting.

Six proven prompts as cards with categories, usage counts, and mono opening lines.

An AI usage panel with per feature bars and a capacity meter that is always full, no metering.

Heading and intro above four milestone columns for days 1, 7, 30, and 90, each with a title, note, and a rising voice match meter.

Two ring framed panels under a centered heading listing what the assistant does and never does, closed by a contractual footnote.