Aug 31, 2026 · 46 min
A PM turns Claude into a self-improving operating system
How I turned Claude into a self-improving PM assistant | Daniel Blum (PM, Melio)
The episode shows how AI workflows can move beyond isolated prompts by combining personal context, automation, feedback, and human judgment.
- 1Daniel Blum connects Slack, email, calendars, and Notion to automate planning, briefs, and follow-up.
- 2A weekly feedback loop analyzes failures and edits, then creates or improves skills instead of treating outputs as final.
- 3Melio’s shared Workstation scales personal AI practices while preserving customization, maintenance, and human judgment.
Don't miss
Daniel explains how a weekly review of failed or rewritten outputs turns feedback into new and improved skills.
The brief
Daniel Blum, a senior product manager at Melio, describes moving from basic AI tools to Claude Cowork, which unlocked a system built around his actual product work rather than isolated prompts.
His Notion dashboard gathers context from Slack, email, calendars, and meetings, while structured files teach Claude the company’s terminology, priorities, colleagues, and working patterns.
Weekly preparation and morning briefs turn work chaos into a managed inbox: the system reviews meetings, extracts follow-ups, surfaces missing context, and updates priorities.
The system improves through telemetry: a weekly review examines failed or rewritten outputs, then creates new skills or fixes existing ones based on direct and indirect feedback.
What began as Daniel’s PM setup became Melio’s shared Workstation, but the episode keeps the tradeoff visible: automation expands capacity only with maintenance, judgment, and better autonomy.
Featuring
Claude Cowork
Notion
ClaudeListen to the full episode and explore every guest, topic, and moment on PodLume.