AI PM Coaching System
A 7-agent architecture that curates 20+ PM newsletters, filters by competency gaps, and generates portfolio artifacts. 90% reduction in weekly research time.
The Problem
Everyone tells PMs to "read newsletters and listen to podcasts." Nobody tells you which ones, how to filter signal from noise, or how to actually retain and apply what you learn.
I was subscribed to 8 newsletters and 8 podcasts. Some weeks I spent 2+ hours reading everything. Other weeks, nothing. Even when I did read, I wasn't retaining much. I was trying to solve a curation problem with willpower. That doesn't scale.
The Insight
I started by doing a rigorous self-assessment using the 9-Signal PM Competency Framework (Jackie Bavaro and Gayle McDowell). I identified specific competency gaps. That changed everything. Instead of "help me be a better PM" (vague), I could target: Business Fundamentals, Technology Strategy, Influence and Presence.
And most of my newsletter subscriptions focused on Product Development and Innovation, my strongest areas. I was reinforcing what I already knew while ignoring my actual gaps.
Architecture
Input Layer (Automated)
Newsletters arrive at Gmail
Zapier detects label, AI summarizes (2000 words to 300)
Appends to Notion weekly page
Result: 85-90% data reduction
Analysis Layer (15-20 min/week)
Part 1: Web search digest (trends in gap areas)
Part 2: Newsletter analysis (filtered by priorities)
Part 3: Podcast curation (2-3 best episodes)
Part 4: Learning Log (insights mapped to competencies)
Output Layer
Weekly: Learning Log entry
Monthly: Deep dive + portfolio artifact
Quarterly: Competency reassessment
The 7 Coaching Agents
Key Design Decisions
- Summarize at ingestion, not analysis. 20 newsletters at 2000 words each = 40,000 words/week. AI tools can't process that in a single analysis pass. Reducing to 300-word summaries at the Zapier step (85% data reduction) made everything downstream viable.
- 60/40 gap-to-strength split. I almost made the mistake of focusing 100% on gaps. That would have let my strengths atrophy. 60% of weekly insights address gaps, 40% maintain competitive advantages in Product Development and Analytics.
- Artifact-oriented learning. Every Learning Log entry has an "Application" field and "Next Step" field. This forces connection to real work and portfolio creation. Without it, I'd consume content endlessly without demonstrating mastery.
Results
90%
Reduction in weekly research time
28
Learning log entries across 6 weeks
12
PM frameworks documented
15 min
Weekly time investment
7
Specialized coaching agents
85%
Data reduction at ingestion
Concrete outputs: 28 curated learning entries mapped to specific competency gaps, 12 reusable PM framework docs (KPI trees, growth diagnosis, influence models, strategy memos), 2 interview story banks with structured STAR narratives, and a competency reassessment showing measurable gap closure.
What This Demonstrates
- Systems thinking: Architected a scalable, automated solution to a messy problem
- AI fluency: Used AI at key automation points, not as a magic bullet
- Self-awareness: Identified gaps through rigorous assessment before building
- Data-driven approach: Metrics, tracking, and measurement throughout
I wrote this up as a Medium post with the full technical implementation details. The system architecture is open source on GitHub.