Episode Details
Back to Episodes
Python Is Dead: The AI That Killed Python for Microsoft Automation
Season 1
Published 5 months ago
Description
(00:00:00) The Python Dilemma in Microsoft's AI Stack
(00:00:32) The Hidden Costs of Python in Power Automate
(00:01:41) The Pitfalls of Using Python as Glue
(00:03:52) The Power of AI-Assisted Orchestration
(00:04:29) Contained Analytics: The Right Place for Python
(00:04:48) The Manual Coding Loop: A Recipe for Disaster
(00:07:10) The Agent-Driven Approach to Orchestration
(00:12:42) Power BI Data Flows: Python's Proper Place
(00:15:51) Power Automate: Replacing Python with Office Scripts
(00:19:11) Fabric Notebooks: Containing Python in Analytics
In this episode of M365.fm, Mirko Peters challenges the long‑held belief that “Python is the language of AI” — at least inside Microsoft’s ecosystem. He explains why Python is still fantastic for data science and ML notebooks, but a terrible choice as glue code for Power Automate, Power BI, Fabric, and Microsoft 365 automation.
WHAT YOU WILL LEARN
Python isn’t “dead,” but its role inside Microsoft 365 has changed. It should power heavy analytics and ML where notebooks and data scientists live — not act as the hidden glue that keeps Power Automate, Power BI, Fabric, and Office running. When you push Python into that glue role, every small change in packages, runtimes, auth, or schemas becomes a production risk, and debugging turns into archaeology.The Microsoft stack is quietly pushing toward a different model: typed, first‑class automation close to the platform (Office Scripts, Power Fx, M, TypeScript‑style code, native connectors) plus AI that generates and maintains that code for you. In that world, Python becom
(00:00:32) The Hidden Costs of Python in Power Automate
(00:01:41) The Pitfalls of Using Python as Glue
(00:03:52) The Power of AI-Assisted Orchestration
(00:04:29) Contained Analytics: The Right Place for Python
(00:04:48) The Manual Coding Loop: A Recipe for Disaster
(00:07:10) The Agent-Driven Approach to Orchestration
(00:12:42) Power BI Data Flows: Python's Proper Place
(00:15:51) Power Automate: Replacing Python with Office Scripts
(00:19:11) Fabric Notebooks: Containing Python in Analytics
In this episode of M365.fm, Mirko Peters challenges the long‑held belief that “Python is the language of AI” — at least inside Microsoft’s ecosystem. He explains why Python is still fantastic for data science and ML notebooks, but a terrible choice as glue code for Power Automate, Power BI, Fabric, and Microsoft 365 automation.
WHAT YOU WILL LEARN
- Where Python absolutely still shines: data science, ML models, analytics notebooks, and heavy transformations
- Why Python becomes friction in Power Platform: external compute, auth overhead, cold starts, dependency drift, dynamic typing, and brutal debugging at 2:14 a.m.
- Real horror stories from Python‑powered flows: custom connectors to Functions, broken schemas, notebook orchestration, permission sprawl, and version drift that silently breaks production
- How Office Scripts (TypeScript‑style), native connectors, Copilot‑generated code, and TypeAgent‑style orchestration can replace most Python glue inside Microsoft 365
- How Copilot and Dataflow Gen2 generate M/Python anchored in your real schemas and semantic models instead of hallucinated structures
- A modern hybrid pattern: Python as the analytics and ML engine, AI + TypeScript‑like code as the orchestration layer, and agents as the air‑traffic controllers for validation, retries, and guardrails
- Quantifiable results from this shift: faster build times, lower cost, fewer defects, and dramatically simpler governance
Python isn’t “dead,” but its role inside Microsoft 365 has changed. It should power heavy analytics and ML where notebooks and data scientists live — not act as the hidden glue that keeps Power Automate, Power BI, Fabric, and Office running. When you push Python into that glue role, every small change in packages, runtimes, auth, or schemas becomes a production risk, and debugging turns into archaeology.The Microsoft stack is quietly pushing toward a different model: typed, first‑class automation close to the platform (Office Scripts, Power Fx, M, TypeScript‑style code, native connectors) plus AI that generates and maintains that code for you. In that world, Python becom