<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>AI First Data blog</title><link>https://aifirstdata.com/blog/</link><description>Practical notes from AI First Data.</description><item><title>Build an Evidence Path for Anomaly Review</title><link>https://aifirstdata.com/blog/build-an-evidence-path-for-anomaly-review/</link><guid isPermaLink="true">https://aifirstdata.com/blog/build-an-evidence-path-for-anomaly-review/</guid><pubDate>Sat, 03 Oct 2026 00:00:00 GMT</pubDate><description>A practical AI First Data note on separating an observed change from guesses about its cause, with source evidence, clear ownership, and human approval.</description></item><item><title>Frame the Question Before Running Analysis</title><link>https://aifirstdata.com/blog/frame-the-question-before-running-analysis/</link><guid isPermaLink="true">https://aifirstdata.com/blog/frame-the-question-before-running-analysis/</guid><pubDate>Sat, 03 Oct 2026 00:00:00 GMT</pubDate><description>A practical AI First Data note on turning a broad request into a permitted and testable reporting question, with source evidence, clear ownership, and human approval.</description></item><item><title>Prepare a Governed Question-to-Report Handoff</title><link>https://aifirstdata.com/blog/prepare-a-governed-question-to-report-handoff/</link><guid isPermaLink="true">https://aifirstdata.com/blog/prepare-a-governed-question-to-report-handoff/</guid><pubDate>Sat, 03 Oct 2026 00:00:00 GMT</pubDate><description>A practical AI First Data note on keeping query logic, assumptions, and approval status together, with source evidence, clear ownership, and human approval.</description></item><item><title>Review Metric Drift Before the Next Report</title><link>https://aifirstdata.com/blog/review-metric-drift-before-the-next-report/</link><guid isPermaLink="true">https://aifirstdata.com/blog/review-metric-drift-before-the-next-report/</guid><pubDate>Sat, 03 Oct 2026 00:00:00 GMT</pubDate><description>A practical AI First Data note on resolving definition differences before another dashboard inherits them, with source evidence, clear ownership, and human approval.</description></item></channel></rss>
