{ "datasetId": "ai-product-manager-readiness-matrix-v1", "schemaVersion": "1.0.0", "createdAt": "2026-07-20", "language": "zh-CN", "license": "https://creativecommons.org/licenses/by/4.0/", "licenseScope": "The framework, field structure, Chinese descriptions, and fictional practice brief created by DragonAI. Third-party source material is excluded and remains under its original terms.", "purposeZh": "用可观察产物而不是课程时长或工具数量判断 AI 产品能力准备度", "assessmentBoundaryZh": "这是学习与作品复盘工具，不是职业资格、招聘市场统计或就业结果预测。不同岗位和行业应调整维度深度。", "sources": [ { "evidenceId": "google-ml-problem-framing-20260720", "title": "Machine Learning Problem Framing: Overview", "publisher": "Google for Developers", "url": "https://developers.google.com/machine-learning/problem-framing/problem-framing" }, { "evidenceId": "google-pair-mental-models-20260720", "title": "People + AI Guidebook: Mental Models", "publisher": "Google PAIR", "url": "https://pair.withgoogle.com/guidebook-v2/chapter/mental-models/" }, { "evidenceId": "google-pair-data-collection-evaluation-20260720", "title": "People + AI Guidebook: Data Collection + Evaluation", "publisher": "Google PAIR", "url": "https://pair.withgoogle.com/guidebook-v2/chapter/data-collection/" }, { "evidenceId": "nist-ai-rmf-core-20260720", "title": "AI Risk Management Framework Core", "publisher": "National Institute of Standards and Technology", "url": "https://airc.nist.gov/airmf-resources/airmf/5-sec-core/" }, { "evidenceId": "google-responsible-ai-introduction-20260720", "title": "Introduction to Responsible AI", "publisher": "Google for Developers", "url": "https://developers.google.com/machine-learning/guides/intro-responsible-ai" } ], "levelDefinitions": [ { "level": "L1-recognize", "nameZh": "识别", "criterionZh": "能解释关键概念、输入输出和常见失败，但尚未独立留下可复核产物。" }, { "level": "L2-execute", "nameZh": "执行", "criterionZh": "能在限定场景完成一次可重复练习，并保存输入、输出、判断和失败记录。" }, { "level": "L3-deliver", "nameZh": "交付", "criterionZh": "能连接该维度的多个产物与上下游决策，用预设指标说明取舍，并处理异常或反证。" }, { "level": "L4-operate", "nameZh": "持续运行", "criterionZh": "能在真实约束下持续跟踪用户与业务结果，维护责任、依赖、监测、复审和调整记录。" } ], "dimensions": [ { "dimensionId": "problem-framing", "nameZh": "问题定义与价值判断", "decisionQuestionZh": "用户任务是否需要 AI，成功和停止条件是什么？", "minimumObservableArtifactZh": "一页问题卡：用户、任务、非 AI 基线、成功指标、失败成本和排除范围", "minimumValidationActionZh": "同时比较 AI、规则或人工流程，说明为什么选择当前方案", "advancedEvidenceZh": "真实用户验证、业务基线、成本收益和停止实验记录", "evidenceIds": [ "google-ml-problem-framing-20260720" ] }, { "dimensionId": "product-strategy", "nameZh": "产品策略与优先级", "decisionQuestionZh": "为谁创造什么差异化价值，先验证哪项假设，什么暂时不做？", "minimumObservableArtifactZh": "机会与定位说明：目标用户、替代方案、价值假设、关键风险、优先级依据和阶段路线图", "minimumValidationActionZh": "用用户证据和非 AI 替代方案挑战价值假设，并说明至少一次取舍", "advancedEvidenceZh": "采用、留存、任务结果、单位经济性和路线图复盘记录", "evidenceIds": [] }, { "dimensionId": "user-interaction", "nameZh": "用户交互与信任校准", "decisionQuestionZh": "用户怎样理解能力边界、纠错、反馈和人工接管？", "minimumObservableArtifactZh": "关键流程图：正常路径、失败提示、反馈入口、人工确认和退出路径", "minimumValidationActionZh": "让目标用户完成任务并复述系统能做什么、不能做什么", "advancedEvidenceZh": "可用性测试、过度信任观察、反馈闭环和解释策略记录", "evidenceIds": [ "google-pair-mental-models-20260720" ] }, { "dimensionId": "data-context", "nameZh": "数据、上下文与知识边界", "decisionQuestionZh": "模型依据来自哪里，数据是否允许、代表且可追溯？", "minimumObservableArtifactZh": "数据与上下文清单：来源、用途、许可或同意、质量、切片、更新和删除规则", "minimumValidationActionZh": "抽查输入到输出的依据链，并记录资料不足、冲突和过期情况", "advancedEvidenceZh": "版本化数据集、检索质量分析、覆盖缺口和数据治理记录", "evidenceIds": [ "google-pair-data-collection-evaluation-20260720" ] }, { "dimensionId": "prototype-system", "nameZh": "原型验证与系统边界", "decisionQuestionZh": "模型、提示、检索、工具、权限和人工分别承担什么？", "minimumObservableArtifactZh": "可运行原型和系统图，标注 API、数据流、权限、错误分支、延迟与成本", "minimumValidationActionZh": "用正常、边界和失败输入各跑一次，保存请求、响应和错误处理", "advancedEvidenceZh": "威胁分析、可靠性预算、可观测性、降级与回滚演练", "evidenceIds": [ "nist-ai-rmf-core-20260720" ] }, { "dimensionId": "evaluation-experimentation", "nameZh": "评测与实验决策", "decisionQuestionZh": "怎样证明新方案在关键任务和切片上更好且风险可接受？", "minimumObservableArtifactZh": "版本化评测集、指标卡、阈值、逐条结果和发布结论", "minimumValidationActionZh": "在看结果前固定门槛，对失败分类并比较至少一个基线", "advancedEvidenceZh": "回归测试、人工评审一致性、线上实验和生产漂移监测", "evidenceIds": [ "nist-ai-rmf-core-20260720", "google-pair-data-collection-evaluation-20260720" ] }, { "dimensionId": "delivery-adoption", "nameZh": "协同交付与产品采用", "decisionQuestionZh": "跨团队依赖怎样推进，产品怎样发布并进入用户工作流？", "minimumObservableArtifactZh": "责任与依赖计划、发布范围、采用方案、支持流程和反馈回收路径", "minimumValidationActionZh": "完成一次发布就绪检查或预演，确认负责人、依赖、培训、支持和停止条件", "advancedEvidenceZh": "分阶段发布、采用漏斗、支持问题、利益相关方决策和发布后复盘", "evidenceIds": [] }, { "dimensionId": "governance-operations", "nameZh": "治理、风险与持续运营", "decisionQuestionZh": "谁对上线、监测、事件、供应商变化和复审负责？", "minimumObservableArtifactZh": "责任与运行清单：风险、负责人、人工监督、监测、事件响应、回滚和复审日期", "minimumValidationActionZh": "选择一个高影响失败，演练发现、升级、处置、通知和恢复", "advancedEvidenceZh": "风险登记、供应链审查、生产事件复盘和定期独立评估", "evidenceIds": [ "nist-ai-rmf-core-20260720", "google-responsible-ai-introduction-20260720" ] } ], "assessmentMethodZh": { "method": "per-dimension-evidence-state", "description": "L1-L4 只描述单个维度的证据状态，不生成总体等级或总分。先确认目标岗位或项目的必需维度，再逐维检查产物和验证记录。没有可复核产物时只能记为 L1；不能用其他维度抵消关键维度缺口。", "outputFields": [ "dimensionId", "requiredForTarget", "evidenceState", "artifactUrls", "validationRecordUrls", "reviewer", "reviewedAt", "nextEvidenceNeeded" ] }, "fictionalPracticeBrief": { "scenarioZh": "为内部客服搭建知识问答助手", "requiredArtifactsZh": [ "问题卡与非 AI 基线", "机会定位、优先级与阶段路线图", "带人工转接的用户流程", "知识来源和更新清单", "带错误分支的可运行原型", "评测集、指标卡和发布结论", "责任依赖、发布与采用计划、反馈入口", "责任、监测、事件与回滚清单" ], "noteZh": "该示例为虚构练习结构，不代表特定企业流程或推荐阈值。" }, "title": "AI 产品经理八维能力与产物矩阵 v1", "description": "把问题定义、产品策略、用户交互、数据与上下文、原型系统、评测实验、协同采用、治理运营八类能力映射为可观察产物和验证动作。", "url": "https://course.dragonai.tech/datasets/ai-product-manager-readiness-matrix-v1.json", "creator": "烛龙智元内容研究组" }