Hi All, I'm looking for feedback from folks who've built signal-detection workflows in Clay. Sharing a 3-minute walkthrough of the Clay table I built for an AI connectivity platform. The architecture sits between two other tools — Codex upstream generating the seed list with a firmographic fit score, and Claude Code downstream applying tiering and account brief rubrics (not shown in this walkthrough). What's inside Clay: dynamic propensity scoring across technographic readiness signals and buying trigger detection. Each signal layer uses a Claygent-first-then-data-provider waterfall to balance LLM scraping coverage with paid enrichment when LLM confidence is low. Curious for feedback on the upstream/downstream split — specifically whether Codex earns its place at the seed list or lookalike generation step. Also open to suggestions on signal detection with LLMs versus data provider enrichment, and signal weighting. https://www.loom.com/share/3c633a40ea84442e9853f7c232277dca
