Analysis

Per-project usage, cost, and AI-generated insights for the active billing cycle. Analysis answers two different operational questions:

• Operational: am I burning quota faster than expected, and which tools or models are responsible?
• Editorial: what are users actually asking, where is the AI struggling, and what should I add to my documents or tools next?

Usage & Cost

Counters are scoped to the current billing cycle and reset at the cycle boundary. Cost projection extrapolates linearly from the current cycle’s burn rate — early in the cycle the projection is volatile, by mid-cycle it’s a reasonable upgrade-decision signal.

Token usage broken down by model is the most useful single number for cost optimization: a tool that quietly invokes an expensive model on every turn shows up here long before it shows up on a bill.

AI-Generated Insights

The dashboard auto-generates weekly summaries from your conversations:

• Top Asked — the most-repeated user questions in the period. If a question shows up 400 times and the AI’s answer isn’t great, it’s a clear signal to add or improve a document.
• Top Issues — recurring friction patterns the AI’s content classifier flagged (failed tool calls, abandoned conversations, negative sentiment). Use as a triage queue for what to fix next.
• Key Findings — narrative summary of the broader trends across the conversation set. Reads like a weekly product review you didn’t have to write.

These insights are also surfaced on the Project Overview right panel so you see them without having to open this page.