AI 扩展计划 / Playbooks

AML Capstone 入口

1. Read AML_ABPA_10_DAY_STARTER.md from top to bottom.

56 行abpa/capstone-aml/README.md

AML Copilot ABPA Capstone

This folder is the first filled ABPA case package. It uses existing AML Copilot work as evidence and does not replace or delete any earlier learning material.

Current Artifact

Artifact	Path	Status
10-day starter pack	`AML_ABPA_10_DAY_STARTER.md`	v0.1 complete
30-day deepening plan	`AML_30_DAY_DEEPENING_PLAN.md`	v0.1 complete

Evidence Reused

Evidence	Why it matters
`docs/AML_COPILOT_PRD.md`	product intent, target workflow, user value
`src/aml/types.ts`	domain objects and case structure
`src/aml/evalBaseline.ts`	baseline scoring and evaluation assumptions
`src/aml/failureTaxonomy.ts`	failure modes and quality risks
`src/aml/groundTruthEval.ts`	ground-truth evaluation direction
`src/aml/__tests__/aml.test.ts`	regression behavior
`src/aml/__tests__/p1evals.test.ts`	P1 evaluation coverage

How To Use This Capstone

Read AML_ABPA_10_DAY_STARTER.md from top to bottom.
Pick one unfinished section and turn it into a standalone portfolio artifact.
Use the templates in ../templates/ when a section needs more depth.
Keep evidence links close to each claim.
Add new versions beside old ones when learning evolves.

For a structured next step, use AML_30_DAY_DEEPENING_PLAN.md to deepen the starter into data readiness, workflow, eval, architecture, operating model, adoption, business case, and interview rehearsal assets.

Recommended Next 30 Days

Week	Focus	Output
W1	Data readiness	AML Data Readiness Pack v0.1
W2	Architecture choices	AML Architecture ADR Set v0.1
W3	Operating model	AML Operating Model + RACI v0.1
W4	Adoption and value	AML Adoption Dashboard + Business Case v0.1

Portfolio Storyline

Interview angle	Story to tell	Evidence to show
AI BA	translated ambiguous AML pain into eval-ready requirements	opportunity canvas, stakeholder map, requirements-to-eval
AI PM	balanced adoption, quality, risk, and unit economics	adoption dashboard, business case
AI Solutions Architect	selected bounded architecture with controls and observability	ADR set, control pack
Financial domain expert	understood AML workflow, case review, audit, and risk controls	PRD, failure taxonomy, eval tests

Preservation Rule

Do not delete older AML, Web3, architecture, LLM, DSDB, AIPA, or daily-note assets when improving this capstone. Add indexes, cross-links, versions, and review notes instead.