AI 扩展计划 / Playbooks

ABPA Starter Kit

The ABPA plan is not another note-writing track. The output is a set of business decision artifacts that can survive interviews, consulting discussions, architecture reviews, and delivery governance.

147 行abpa/README.md

ABPA-180 Starter Kit

ABPA = AI BA x Product x Architect. This folder turns docs/AI_BA_PRODUCT_ARCHITECT_180_PLAN.md from a curriculum into reusable work assets.

Purpose

The ABPA plan is not another note-writing track. The output is a set of business decision artifacts that can survive interviews, consulting discussions, architecture reviews, and delivery governance.

This folder is additive. It does not replace earlier Web3, architecture, Solidity, LLM, DSDB, AIPA, trading, finance-design, or daily-note materials. Those remain useful learning assets. ABPA only adds a BA/PM/architecture lens on top of them.

Use this folder when you want to:

start a new AI transformation case study,
convert a PRD into eval-ready requirements,
explain why an AI system should or should not be built,
prepare an AI BA / AI PM / AI Solutions Architect interview story,
turn existing AML Copilot work into an executive-ready portfolio artifact.

Folder Structure

Path	Purpose
`templates/`	Reusable blank templates for each ABPA asset
`capstone-aml/`	First filled capstone using the existing AML Copilot assets
`interview/`	Interview question bank and answer scaffolds for AI BA / AI PM / AI Architect roles
`../LEARNING_ASSET_GOVERNANCE.md`	Preservation, review, and portfolio-conversion system for all learning assets

Recommended Workflow

1. Pick a business scenario.
2. Copy the templates you need.
3. Fill evidence before opinion.
4. Map every important requirement to an eval, signal, owner, or control.
5. Summarize the decision in one page.

Evidence Hierarchy

Evidence strength	Examples	How to use
Strong	production metrics, labeled external dataset, observed user workflow, CI result, trace export	Decision basis
Medium	internal synthetic dataset, expert interview, prototype usability test, vendor trial	Hypothesis support
Weak	vendor demo, generic survey, model benchmark unrelated to workflow	Never use alone

ABPA Asset Stack

Layer	Asset	Question it answers
Problem	AI Opportunity Canvas	Is this worth doing with AI?
People	Stakeholder Evidence Map	Who can make it succeed or fail?
Workflow	BPMN + Pain Metrics	Where does work actually break down?
Requirements	Requirements-to-Eval Matrix	How will we know it works?
Data	Data Readiness Pack	Is the data fit for the AI decision?
Architecture	AI Architecture ADR Set	Which design choices are explicit and reversible?
Risk	AI Control Pack	How do we prevent and detect harm?
Operations	Operating Model + RACI	Who owns the system after launch?
Adoption	Adoption Dashboard	Are users actually changing their work?
Value	Business Case	Is the investment justified by measurable value?
Decision	Executive Decision Memo	What should leadership approve now?

Template Index

Template	Use it when
`templates/01-ai-opportunity-canvas.md`	deciding whether a scenario deserves AI investment
`templates/02-stakeholder-evidence-map.md`	mapping who has power, pain, evidence, and objections
`templates/03-bpmn-pain-metrics.md`	converting process pain into measurable workflow facts
`templates/04-requirements-to-eval-matrix.md`	making requirements testable through evals and thresholds
`templates/05-ai-control-pack.md`	designing AI risk controls and release gates
`templates/06-executive-decision-memo.md`	summarizing a decision for sponsors
`templates/07-data-readiness-pack.md`	proving data quality, labels, access, and governance readiness
`templates/08-ai-architecture-adr-set.md`	documenting AI architecture choices and reversal triggers
`templates/09-operating-model-raci.md`	assigning owners for product, evals, data, risk, operations, and adoption
`templates/10-adoption-dashboard.md`	tracking usage, trust, quality, behavior change, and value realization
`templates/11-business-case.md`	turning workflow impact into funding gates and ROI logic
`templates/12-portfolio-evidence-map.md`	mapping notes, code, PRDs, evals, and capstones into interview evidence

Current Capstone

Start here:

capstone-aml/README.md
capstone-aml/AML_ABPA_10_DAY_STARTER.md
capstone-aml/AML_30_DAY_DEEPENING_PLAN.md

Interview Pack

interview/AI_BA_PM_ARCHITECT_INTERVIEW_BANK.md

2026+ AI Expansion Tracks

These are new expansion tracks, not just ABPA starter material:

../AI_2026_EXPANSION_START_HERE.md
../AI_NEW_DEMANDS_2026_EXPANSION.md
../AI_FOUNDATIONS_CLASSIC_PAPERS_PLAN.md
../AGENTIC_ENTERPRISE_ARCHITECTURE_90_PLAN.md
../AI_GOVERNANCE_EVALOPS_RISK_90_PLAN.md
../AI_BA_PM_PRACTICE_LAB.md
../AI_ROLE_COMPETENCY_MATRIX_2026.md
../FINANCIAL_RETAIL_AI_CASE_PORTFOLIO.md
../AI_ARCHITECTURE_DIAGRAM_PLAYBOOK.md
../AI_LONG_TERM_KNOWLEDGE_GRAPH_AND_REVIEW_SYSTEM.md
../AI_INTERVIEW_PORTFOLIO_STORYLINE_PLAYBOOK.md
../AI_VENDOR_BUILD_BUY_ADOPTION_PLAYBOOK.md
../AI_REQUIREMENTS_TO_EVAL_COOKBOOK.md
../AI_OPERATING_MODEL_RACI_RUNBOOK.md
../AI_ARCHITECTURE_REVIEW_GATE_CHECKLISTS.md
../AI_CONTEXT_ENGINEERING_PLAYBOOK.md
../AI_EXPANSION_MASTER_INDEX.md
../AI_CASE_DRILL_WORKBOOK_30_DAYS.md
../AI_EXECUTIVE_COMMUNICATION_MEMO_PACK.md
../AI_PLATFORM_PM_PLAYBOOK.md
../AI_REGULATORY_RESPONSE_PLAYBOOK.md
../AI_DATA_PRODUCT_MANAGEMENT_PLAYBOOK.md
../AI_BOARD_AUDIT_COMMITTEE_GOVERNANCE_PACK.md
../AI_CAPABILITY_ASSESSMENT_RUBRIC.md
../AI_RETRIEVAL_EVAL_GRAPH_RAG_PLAYBOOK.md
../AI_PLATFORM_SECURITY_GATEWAY_LAB.md
../AI_REGULATOR_EXAM_SIMULATION_PACK.md
../AI_ADVANCED_CASE_DRILL_WORKBOOK_60_DAYS.md
../AI_MEMORY_CONTEXT_STATE_PLAYBOOK.md
../AI_MULTI_AGENT_ORCHESTRATION_PLAYBOOK.md
../AI_OBSERVABILITY_COST_SLO_PLAYBOOK.md
../AI_AGENT_PROTOCOLS_MCP_A2A_PLAYBOOK.md

It uses existing repository evidence:

docs/AML_COPILOT_PRD.md
src/aml/types.ts
src/aml/evalBaseline.ts
src/aml/failureTaxonomy.ts
src/aml/groundTruthEval.ts
src/aml/__tests__/aml.test.ts
src/aml/__tests__/p1evals.test.ts

Quality Bar

An ABPA artifact is complete only if it answers four questions:

What decision does this support?
What evidence backs it?
What uncertainty remains?
What should be tested next?