Daisy

README excerpt

# AI Deal Desk Analyst for Contact Center Resellers

**Intelligence at Work Hackathon, 23 Oct 2026. Solo project. Everything in this repo was built during the hackathon.**

A Claude-powered Cortex Agent in Snowflake CoWork that helps non-expert resellers answer a hard pre-sales question:
*should this buyer get Amazon Connect Customer or Customer Basic, what will it cost, and when should they switch?*
A complex contact center TCO typically takes a consultant 40–80 hours. With this agent, any seller gets a defensible
first-pass answer, with sources and math, in under 15 minutes.

## What it does
- **Guided intake:** the seller gives agents, channels and capabilities. Anything unknown gets a default or a benchmark menu (Forrester TEI, AWS examples) with a recommended option.
- **Deterministic pricing:** every dollar comes from a governed, effective-dated price catalog (price version `AWS-CONNECT-2026-10-23`) via a SQL calculator. Claude never computes prices itself.
- **Deal brief:** a recommendation, side-by-side TCO, cost layers with a chart, the switch point, tagged assumptions, a seller talk track, "how we calculated this" with source URLs, and caveats.
- **Governed saved deals:** sellers save analyses. A row access policy limits reps to their own deals, and a masking policy hides discount pricing from reps; managers see everything.

Key insight from the model: **capabilities decide, not size.** Customer bundles AI in its per-minute rate, so its cost is flat at about $235 pe