AI PM agent resource pack — 2026-05-09¶
Purpose: fresh source pack for building an AI product management agent with durable product-management context, workflows, and templates. This complements the existing 2026-03-19 source index instead of replacing it.
Harvest method: web fetch + human synthesis. Do not paste raw article text into prompts by default; retrieve this index first, then load the specific workflow or template needed for the task.
Quality bar¶
Sources were selected because they are widely cited, primary or near-primary, practical, and directly useful for one or more AI PM capabilities: strategy, discovery, roadmapping, prioritization, requirements, delivery shaping, measurement, or context engineering.
Fresh resources¶
| # | Source | Category | Agent-useful takeaway |
|---|---|---|---|
| 1 | SVPG — The Product Operating Model | Operating model | Product work is a way of working: principles, practices, and competencies that move companies away from project/IT/feature-output models. Use this as the top-level orientation for the agent. |
| 2 | SVPG — Product vs Feature Teams | Team model | Distinguish empowered product teams from feature teams. The agent should ask whether it is solving an outcome or merely documenting a requested feature. |
| 3 | Product Talk — Product Discovery Basics | Discovery | Discovery is decision-making about what to build. Strong teams stay continuously connected to customers and start from outcomes, opportunities, and solutions. |
| 4 | Product Talk — Opportunity Solution Trees | Discovery structure | Use outcome → opportunities → solutions → assumption tests to preserve traceability from business goals to build candidates. |
| 5 | Intercom — RICE Prioritization | Prioritization | Score Reach, Impact, Confidence, and Effort to compare unlike ideas. Treat the score as decision support, not an automatic mandate. |
| 6 | NN/g — UX Roadmaps | Roadmapping | A roadmap is a living strategic artifact organized by scope, time horizon, and themes. Roadmaps should focus on problems/outcomes, not feature release plans. |
| 7 | Amplitude — North Star Metric | Metrics / strategy | A North Star metric should align to customer value, express product strategy, and lead future business outcomes. Use input metrics to make it actionable. |
| 8 | Atlassian — Product Roadmaps | Roadmapping | Roadmaps are shared sources of truth connecting vision, direction, priorities, and progress. Tailor detail by audience and update only as often as needed to preserve trust. |
| 9 | Basecamp — Shape Up introduction | Delivery shaping | Shape before betting: define boundaries, appetite, risks, and the rough solution shape before committing team time. Useful when the agent needs to turn ambiguity into a bounded pitch. |
| 10 | DigitalOcean — Product Requirements Document | Requirements | A PRD should align stakeholders around problem, objectives, audience, scope, functional/non-functional requirements, dependencies, and delivery expectations without becoming unreadable. |
| 11 | IdeaPlan — Context Engineering for Product Managers | PM context engineering | Context engineering is deciding what information an AI system receives, when, and in what structure. PM context should include user, business, competitive, technical, and interaction context. |
| 12 | Anthropic — Building Effective Agents | Agent design | Prefer simple, composable workflows before autonomous agents. Use workflows for predictable PM tasks and agents for open-ended work with clear checkpoints and ground truth. |
| 13 | Martin Fowler / Thoughtworks — Context Engineering for Coding Agents | Context engineering | Treat files, rules, skills, tools, MCP servers, subagents, and hooks as context interfaces. Keep context small, explicit, and loaded only when relevant. |
Synthesis for the AI PM agent¶
Core operating principles¶
- Start from outcomes, not requested features. If the request is a feature, translate it into the customer problem, business outcome, and assumptions.
- Preserve traceability: business outcome → product outcome → customer opportunity → solution candidate → assumption test → requirement → release signal.
- Separate strategic artifacts from delivery artifacts. Roadmaps align on direction; PRDs align on what is being built; backlogs sequence execution.
- Treat prioritization frameworks as structured judgment. RICE, MoSCoW, Kano, and North Star inputs reveal tradeoffs; they do not replace product judgment.
- Design context as a product. The agent should load only the context needed for the PM job at hand and should cite the source or template it used.
Minimum context packet for PM work¶
Every AI PM workflow should try to collect or infer:
- Product / initiative name.
- Target user or customer segment.
- Business goal and product outcome.
- Current evidence: user research, analytics, sales/support signal, competitive signal, technical constraints.
- Decision needed now.
- Known assumptions and risks.
- Stakeholders and audience for the output.
- Source artifacts already available in the repo.
Capability map¶
| Capability | Primary sources | Output artifact |
|---|---|---|
| Product strategy framing | SVPG, Amplitude | Strategy brief, North Star tree |
| Discovery planning | Product Talk, SVPG | Discovery brief, opportunity solution tree |
| Prioritization | Intercom, Kano/MoSCoW from existing KB | Prioritization scorecard |
| Roadmapping | NN/g, Atlassian | Now/Next/Later roadmap |
| Requirements | DigitalOcean, Aha from prior index | PRD, acceptance criteria |
| Delivery shaping | Basecamp Shape Up | Shaped pitch, appetite, risks |
| Agent context design | IdeaPlan, Anthropic, Martin Fowler | Context packet, workflow routing, evaluation checklist |
Gaps to hydrate next¶
- Primary Amazon Working Backwards / PR-FAQ source.
- HEART framework primary source or reliable Google Ventures reference.
- Product analytics instrumentation and experimentation playbooks.
- Product ethics / responsible AI decision frameworks beyond NIST and OWASP.
- Customer interview repository structure and insight tagging conventions.