<?xml version="1.0" encoding="UTF-8"?><oembed><type>video</type><version>1.0</version><html>&lt;iframe src=&quot;https://www.loom.com/embed/94ac3555a2a24e40b660d54a74a229d2&quot; frameborder=&quot;0&quot; width=&quot;1920&quot; height=&quot;1440&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1440</height><width>1920</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1440</thumbnail_height><thumbnail_width>1920</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/94ac3555a2a24e40b660d54a74a229d2-9753357a5ae55d35.gif</thumbnail_url><duration>312.142</duration><title>Cadence walkthrough</title><description>This Loom explains how to decide whether an 18-agent collections floor should scale an AI-assisted triage, and how to test it credibly. The author locked 6 KPIs computed from 1,200 accounts over 6 months, identified data issues and cure-rate noise (36 to 49% month to month), then scored build versus extend versus buy across 8 criteria, where build led (3.6) after weighting data readiness and explainability. A guardrail-gated pilot returned a cure-rate lift and improved time to first touch by about a day and a half with zero guardrail violations, but the control cohort matched baseline and the weekly analysis showed the pilot’s lift could not be distinguished from doing nothing due to a confidence interval containing zero. Despite a go gate based on guardrails, the author recommended against scaling beyond wave 1, citing the earlier phase-one mortality impact of about 12 points.</description></oembed>