ai-efficiency-auditlisted

# 提效证据生成器 · AI Efficiency Audit > 造器图纸 #16 · 审计类工具 · 难度 ★★☆ > 下次老板提 AI 提效,方案你先拍在桌上。 This skill turns vague "we should use AI" pressure into an auditable, defensible efficiency case. It inventories the user's real workflow, scores each task for automation potential with an explicit rubric, quantifies hours saved with honest ranges (never promises), and produces a manager-ready one-page proposal with a pilot plan and rollback conditions. As a final step it can **forge the audit into a standalone team self-check tool** (造器模式). The credibility of the output depends on three disciplines. Follow them strictly: 1. **Baseline first** — never claim savings without first establishing current time spent. No baseline, no number. 2. **Ranges, not promises** — all savings are estimated as conservative–optimistic ranges with stated assumptions. 3. **Pilot before rollout** — every proposal ends in a small reversible pilot with pre-agreed success metrics, not a big-bang adoption. ## Workflow ### Stage 1 — Workflow intake (建立基线) Collect the task inventory. Two modes, let the user pick: - **快速模式 (5 minutes)**: Ask the user to brain-dump their typical week in plain language. You extract the task list yourself, then confirm it back as a table. - **审计模式 (recommended for team proposals)**: Walk through the structured inventory, one category at a time: 例行产出 (reports, emails, docs) → 信息处理 (reading, summarizing, searching) → 协调沟通 (meetings, follow-ups, status syncs) → 核心专业工作 →