Episode Details
Back to Episodes
Fabric governance GPT‑5: stop manual audits and let Copilot enforce compliance
Season 1
Published 5 months, 3 weeks ago
Description
Fabric governance GPT‑5: this episode of M365.fm shows how GPT‑5 inside Microsoft 365 Copilot finally fixes Fabric governance by reasoning across Purview, Power BI, and Fabric so you can stop doing manual spreadsheet audits. Mirko Peters explains why governance breaks today: each system logs its own truth—classifications in Purview, roles and RLS in Power BI, lineage and workspaces in Fabric—without a shared reasoning layer to connect them into a single, auditable story.
Mirko starts by breaking down the gap between “data” and “logic.” Fabric, Purview, and Power BI are excellent at storing facts, but terrible at inferring relationships between those facts when you ask real compliance questions like “Which highly confidential datasets are used in reports without RLS?”. He shows how GPT‑5’s chain‑of‑thought reasoning changes this: Copilot interprets your intent, fans out across services, correlates classifications, lineage, and security config, and comes back with verified mismatches instead of raw lists you still need to reconcile.
He then contrasts old Copilot behavior with the GPT‑5 generation. Earlier models worked like helpful search: they stayed inside one product at a time and stitched text together. GPT‑5 behaves like an internal audit analyst: it decomposes your question, runs parallel reasoning threads over Fabric, Purview, and Power BI contexts, and only synthesizes an answer once the cross‑checks line up. Mirko explains why the “verbose” explanations are a feature, not a bug—they’re an audit trail of the model’s internal logic you can show to security and regulators.
The episode walks through a concrete audit scenario: proving that every Fabric table containing PII is both classified in Purview and protected by Row‑Level Security in Power BI. Mirko shows how the old way involved exporting CSVs, reconciling nearly matching names, and praying nothing was missed; then he demonstrates how a single GPT‑5 Copilot request interprets the requirement, pulls lineage from Fabric, labels from Purview, and RLS config from Power BI, and highlights only the real gaps. You’ll see how this turns multi‑week manual reviews into repeatable, on‑demand checks.
WHAT YOU WILL LEARN
Fabric governance never lacked data—it lacked reasoning. By adding GPT‑5 as a cross‑system brain on top of Purview, Power BI, and Fabric, you replace manual correlation and brittle scripts with an always‑on auditor that can explain how it reached every conclusion instead of just dumping logs at you.
WHO THIS EPISODE IS FOR
This episode is ideal for data protection officers, security and compliance teams, Fabric and Power BI admins, and architects responsible for governance across Microsoft’s data stack. It is especially valuable if you’re drowning in export‑and‑Excel audits today and need a credible path to automate evidence gathering and gap detection without losing transparency or control.
ABOUT THE HOST
Mirko Peters is a Microsoft 365 and data platform consultant focused on building governed, scalable analytics and AI platforms with Microsoft Fabric, Power BI, Purview, and Copilot. Through M365.fm, he shares governance blueprints, real‑world audit stories, and Copilot patterns that help organizations replace manual compliance work with explainable, automat
Mirko starts by breaking down the gap between “data” and “logic.” Fabric, Purview, and Power BI are excellent at storing facts, but terrible at inferring relationships between those facts when you ask real compliance questions like “Which highly confidential datasets are used in reports without RLS?”. He shows how GPT‑5’s chain‑of‑thought reasoning changes this: Copilot interprets your intent, fans out across services, correlates classifications, lineage, and security config, and comes back with verified mismatches instead of raw lists you still need to reconcile.
He then contrasts old Copilot behavior with the GPT‑5 generation. Earlier models worked like helpful search: they stayed inside one product at a time and stitched text together. GPT‑5 behaves like an internal audit analyst: it decomposes your question, runs parallel reasoning threads over Fabric, Purview, and Power BI contexts, and only synthesizes an answer once the cross‑checks line up. Mirko explains why the “verbose” explanations are a feature, not a bug—they’re an audit trail of the model’s internal logic you can show to security and regulators.
The episode walks through a concrete audit scenario: proving that every Fabric table containing PII is both classified in Purview and protected by Row‑Level Security in Power BI. Mirko shows how the old way involved exporting CSVs, reconciling nearly matching names, and praying nothing was missed; then he demonstrates how a single GPT‑5 Copilot request interprets the requirement, pulls lineage from Fabric, labels from Purview, and RLS config from Power BI, and highlights only the real gaps. You’ll see how this turns multi‑week manual reviews into repeatable, on‑demand checks.
WHAT YOU WILL LEARN
- Why Fabric, Purview, and Power BI each see only part of the governance picture—and where audits really fail.
- How GPT‑5’s chain‑of‑thought reasoning lets Copilot correlate lineage, classifications, and security into one view.
- How to phrase governance questions so Copilot can surface concrete policy and configuration gaps, not just lists.
- How GPT‑5’s detailed explanations act as an audit trail you can use with compliance and security teams.
- How to move from ad‑hoc spreadsheet audits to repeatable, Copilot‑driven governance checks across your tenant.
Fabric governance never lacked data—it lacked reasoning. By adding GPT‑5 as a cross‑system brain on top of Purview, Power BI, and Fabric, you replace manual correlation and brittle scripts with an always‑on auditor that can explain how it reached every conclusion instead of just dumping logs at you.
WHO THIS EPISODE IS FOR
This episode is ideal for data protection officers, security and compliance teams, Fabric and Power BI admins, and architects responsible for governance across Microsoft’s data stack. It is especially valuable if you’re drowning in export‑and‑Excel audits today and need a credible path to automate evidence gathering and gap detection without losing transparency or control.
ABOUT THE HOST
Mirko Peters is a Microsoft 365 and data platform consultant focused on building governed, scalable analytics and AI platforms with Microsoft Fabric, Power BI, Purview, and Copilot. Through M365.fm, he shares governance blueprints, real‑world audit stories, and Copilot patterns that help organizations replace manual compliance work with explainable, automat