Descriptions:
Reed Robinson, product manager for AI at Zapier, sits down with How I AI host Claire Vo to walk through how he uses Model Context Protocol (MCP) servers to automate the most time-consuming parts of a PM’s day—meeting prep, CRM updates, and synthesizing customer feedback. Rather than treating MCPs as an abstract technical concept, Robinson frames them in practical terms: integrations that give AI tools like Claude access to data living inside other applications and the ability to act on it directly.
The conversation digs into a tension many AI power users encounter—the pull between deterministic automation workflows and instructive AI agents, even when both have access to identical tools. Robinson shares observations from Zapier’s own work on MCP server and client approaches, explaining why enterprises are increasingly drawn to role-based tool configurations where an ops admin can pre-load the right MCP capabilities for every employee who signs in.
For anyone trying to make practical sense of the MCP ecosystem, this episode offers a grounded entry point from someone building the infrastructure. Robinson addresses real limitations—long-running lookups exceeding seven minutes remain problematic for agentic MCP use—while Claire Vo draws out the workflow implications for product managers looking to reclaim time from repetitive operational tasks.
📺 Source: How I AI · Published February 02, 2026
🏷️ Format: Interview
