AI PM — Taxonomy

Canonical reference for classifying knowledge entries. Used by the source processing skill and knowledge entry template.

For full phase lineage, component descriptions, and AI-PM emphasis, see Lifecycle Framework V2.

---

Five Domains

Every knowledge entry belongs to exactly one domain.

Domain	Folder	What belongs here
Product Lifecycle	knowledge-base/product-lifecycle/{phase}/	Techniques and insights that augment a specific stage of building product
Horizontal	knowledge-base/horizontal/{horizontal_domain}/	Lifecycle-agnostic AI PM skills and knowledge areas — see Horizontal Domains below
AI Adoption & Change Management	knowledge-base/ai-adoption/	How organizations and individuals adapt to AI-native ways of working — scaling expertise, org change, driving adoption
Adjacent Disciplines	knowledge-base/adjacent-disciplines/	How AI transforms the disciplines of PMs’ close collaborators (engineering, design, analytics) — and what their shifting AI adoption means for product leadership
Software Methodology	knowledge-base/software-methodology/	How AI fundamentally changes software delivery paradigms — compound engineering, spec-driven development, vibe coding. Distinct from lifecycle-phase techniques (which augment a phase) and horizontal skills (which are lifecycle-agnostic PM tools). The test: is this about how the methodology itself is evolving?

Classification test

1. Does this technique augment a specific lifecycle phase? → Product Lifecycle (file under the phase)
2. Is it a lifecycle-agnostic AI PM skill or knowledge area? → Horizontal (then determine which horizontal domain — see below)
3. Is it about bringing others along or changing how teams/orgs work? → AI Adoption & Change Management
4. Is it about how AI transforms a collaborator discipline (engineering, design, analytics) and what that means for PMs? → Adjacent Disciplines
5. Is it about how AI fundamentally changes the software delivery methodology itself? → Software Methodology

---

Horizontal Domains

The horizontal domain is organized as a stack by delivery mechanism, with increasing capability and autonomy. Each layer is a knowledge area with its own depth. Entries sit flat in their horizontal domain folder; the horizontal_domain frontmatter field records which one.

Horizontal Domain	Slug	Folder	What belongs here
Prompting	prompting	horizontal/prompting/	Portable techniques for crafting effective instructions — works in any chat window. Prompting patterns, meta-skill patterns, writing workflows, role delineation, refinement techniques.
Context & Knowledge Management	context	horizontal/context/	Knowledge infrastructure: making non-code knowledge discoverable and usable to agents and their human coworkers — context graphs, agent-oriented KM, progressive disclosure, knowledge discoverability.
Templated AI Runtimes	runtimes	horizontal/runtimes/	Packaged, shareable, non-agentic AI tools (Custom GPTs, Google Gems, Claude Projects). Building, distributing, and managing templated AI runtimes for teams and organizations.
Agents	agents	horizontal/agents/	Building and managing knowledge agents — lifecycle management, rules, skills, templates, tools, workflows. How PMs select, onboard, train, give feedback to, and performance-manage AI agents.

Agents Sub-domains

The agents horizontal domain has its own internal organization, reflecting the distinct concerns of building, deploying, and managing AI agents.

Sub-domain	Slug	Folder	What belongs here
System Design	system-design	horizontal/agents/system-design/	Tool-agnostic patterns and techniques for designing and configuring agent systems — architecture, behavior, control mechanisms, workflow execution, knowledge capture.
Harnesses	harnesses	horizontal/agents/harnesses/	Platform-specific knowledge — setup, configuration, capabilities, best practices for specific agent tools (Claude Code, Cursor, Devin, etc.).
Managing Agents	managing-agents	horizontal/agents/managing-agents/	The human-agent management relationship — deciding what to delegate, how to delegate effectively, evaluating agent output, and managing agents as autonomous participants. See Managing Agents Sub-domains below.

Managing Agents Sub-domains

Sub-domain	Slug	Folder	What belongs here
Task-Agent Fit	task-fit	horizontal/agents/managing-agents/task-fit/	Decision frameworks for what work to delegate to agents — when delegation is worthwhile, risk assessment, the jagged frontier of agent capability.
Delegation Craft	delegation	horizontal/agents/managing-agents/delegation/	How to effectively hand off work to agents — instructions, documentation formats, defining done, checkpoints, iteration strategies.
Evaluation & Feedback	evaluation	horizontal/agents/managing-agents/evaluation/	Assessing agent output, giving feedback, iteration cost management, quality recognition, feedback loops.
Agent Selection & Onboarding	selection	horizontal/agents/managing-agents/selection/	Choosing which agents/tools for which roles, matching capability to need, initial setup and calibration.

Agents sub-domain classification test

1. Is it a tool-agnostic pattern for designing/configuring agent systems? → System Design
2. Is it specific to a particular agent platform? → Harnesses
3. Is it about the human-agent management relationship? → Managing Agents (then determine which sub-domain — see below)

Managing Agents classification test

1. Is it about deciding WHEN or WHETHER to delegate? → Task-Agent Fit
2. Is it about HOW to hand off work effectively? → Delegation Craft
3. Is it about assessing output or giving feedback? → Evaluation & Feedback
4. Is it about choosing or onboarding agents/tools? → Agent Selection & Onboarding

Horizontal classification test

1. Is it a portable prompting technique (works in any chat window)? → Prompting
2. Is it about making knowledge discoverable and usable? → Context & Knowledge Management
3. Is it about packaged, shareable, non-agentic AI tools? → Templated AI Runtimes
4. Is it about building and managing knowledge agents? → Agents (then determine which agents sub-domain — see above)

Dual attribution

Horizontal entries frequently relate to specific lifecycle phases. Use lifecycle_phase in frontmatter (optional for horizontal entries) to record the primary phase connection when one exists. An entry about how deliberate context selection improves PRD writing would live in Context with lifecycle_phase: build as a cross-reference. One idea can be attributed to multiple domains — when an insight genuinely belongs in both a horizontal domain and a lifecycle phase, create the entry where it has the most depth and add cross-references via the Related section.

---

Product Lifecycle Phases

Six phases, each with named components. Entries sit flat in the phase folder; the component frontmatter field records the most specific applicable component.

1. Discover

What problems are worth solving?

Component	Slug	Description
Problem Signal Detection	problem-signal-detection	Detecting unmet needs from data, support, sales, and customer contact
Market & Competitive Intelligence	market-competitive-intelligence	Competitive analysis, market sizing, trend monitoring
Opportunity Prioritization	opportunity-prioritization	Ranking and managing the space of possible problems
Problem Brief	problem-brief	Concise articulation of the problem, who it affects, why now

2. Frame

What does success look like, and what’s the bet?

Component	Slug	Description
Stakeholder Intent Alignment	stakeholder-intent-alignment	Refining intent, securing sponsorship, shared understanding
Success Criteria & Business Case	success-criteria-business-case	KPIs, financial modeling, defining the objective function
Roadmap Definition	roadmap-definition	Outcome-based roadmapping, strategic sequencing, scoping

3. Shape

What solution takes form, and for whom?

Component	Slug	Description
Persona & Journey Mapping	persona-journey-mapping	User models, behavioral archetypes, JTBD framing
Prototyping & Risk Reduction	prototyping-risk-reduction	Wireframes, prototypes, fat-marker sketches, assumption testing
Go-to-Market Planning	gtm-planning	Positioning, messaging, channel strategy, pricing

4. Build

How do we ship with clarity and conviction?

Component	Slug	Description
Feature Specification Writing	feature-specification-writing	PRDs, specs, technical design documents
User Story & Acceptance Criteria	user-story-acceptance-criteria	Story decomposition, acceptance criteria, story mapping
Stakeholder Communication	stakeholder-communication	Status updates, cross-functional coordination, expectation management
Scope & Priority Tradeoffs	scope-priority-tradeoffs	Scope negotiations, circuit breakers, quality/speed tradeoffs

5. Release

How do we put this into the world deliberately?

Component	Slug	Description
Release Readiness Assessment	release-readiness-assessment	Go/no-go evaluation, quality gates, support readiness
Phased Rollout Strategy	phased-rollout-strategy	Feature flags, canary deploys, segment-based rollouts
Release Communications	release-communications	Changelogs, release notes, stakeholder notifications
Launch Marketing & Enablement	launch-marketing-enablement	Launch campaigns, sales enablement, customer success briefings

6. Measure

Did it work, and what do we do next?

Component	Slug	Description
KPI & Outcome Monitoring	kpi-outcome-monitoring	North Star dashboards, OKR reviews, post-launch accountability
Customer Feedback Collection	customer-feedback-collection	NPS/CSAT, surveys, support themes, interview programs
Experiment Design & Analysis	experiment-design-analysis	A/B tests, statistical rigor, acting on results
End-of-Life & Deprecation	end-of-life-deprecation	Sunsetting, migration paths, deprecation communication

---

Entry Types

Type	Slug	What it captures
Technique	technique	A named method, practice, or procedure
Mental Model	mental-model	A framework for thinking about a class of problems
Insight	insight	An observation, principle, or non-obvious pattern

Entry Status

Status	Meaning	Promotion criteria
draft	Initial capture from one source	—
solid	Supported by multiple sources, well-articulated	Corroboration from a second independent source
canonical	Thoroughly vetted, teachable	Could teach this to someone; no open questions

Entry Origin

Origin	Meaning
sourced	Extracted from external sources
organic	From personal experience
both	Originated organically, later found external validation

---

Frontmatter Reference

Knowledge entry frontmatter fields for taxonomy:

# Required
entry_type: technique | mental-model | insight
domain: product-lifecycle | horizontal | ai-adoption | adjacent-disciplines | software-methodology

# Required for product-lifecycle entries
lifecycle_phase: discover | frame | shape | build | release | measure
component: <component-slug from tables above>

# Required for horizontal entries
horizontal_domain: prompting | context | runtimes | agents

# Optional for horizontal entries (dual attribution)
lifecycle_phase: discover | frame | shape | build | release | measure  # when the entry has a primary phase connection

# Optional
featured: true  # Worth championing organizationally

---

Placement Rules

1. Most specific component wins. If an entry clearly maps to a lifecycle component, tag it there. If it spans multiple components within a phase, use the primary one and note connections in Related.
2. Phase-level is OK. If an entry fits a phase but not a specific component, omit the component field and file it in the phase folder.
3. Cross-phase entries go horizontal. If an entry applies across 3+ phases, it’s probably a horizontal entry. Determine which horizontal domain fits best.
4. Horizontal domain entries can reference lifecycle phases. Use lifecycle_phase in frontmatter to record the primary phase connection when one exists. Use the Related section to cross-reference entries in other domains.
5. Match to delivery mechanism. A portable prompting technique → Prompting. Knowledge infrastructure → Context. A packaged, shareable tool → Templated AI Runtimes. Autonomous agent management → Agents.
6. When in doubt, ask. Don’t force-fit entries into the taxonomy. Flag uncertain placements for review.
7. Screen for adoption. Sources often embed change management insights alongside craft techniques. Extract these separately into ai-adoption.
8. Screen for horizontal domain insights. Sources about delivery workflows may contain agent management insights or knowledge management patterns. Extract these into the appropriate horizontal domain.