Descriptions:
Marco Casalaina, VP of Core AI Products at Microsoft, joins the ‘How I AI’ podcast to demonstrate five practical automation workflows using Warp, an AI-powered terminal application. Rather than showcasing Warp’s primary coding use case, Casalaina focuses on a less-covered strength: natural language control of any tool that exposes a command-line interface, making complex administrative operations accessible without deep CLI expertise.
The centerpiece demo is Azure resource management — assigning granular role permissions (Azure AI User, Azure AI Project Manager) to colleagues across a subscription via a single natural language prompt, a task that would otherwise require navigating the Azure web portal role-by-role for potentially an hour. Warp generates and executes the Azure CLI commands iteratively until the task completes. A document processing demo shows Warp installing PyPDF2, writing a Python script to separate and recombine odd and even pages from a scanned math test, running it, and deleting the script — an ephemeral solution that Casalaina explicitly prefers over building a permanent tool. Windows scanning automation is handled through NAPS2, an open-source scanner CLI that Casalaina pre-installed and configured as a persistent Warp rule, illustrating how a small amount of upfront setup unlocks repeated natural language control.
A recurring theme is the gap between what users think is possible on their machines and what is actually programmatically accessible. Casalaina argues that working with terminal agents fundamentally changes that mental model — and that micro-agents handling administrative friction are where the immediate, underappreciated productivity gains live for engineering leaders and technical PMs alike.
📺 Source: How I AI · Published March 23, 2026
🏷️ Format: Interview
