起步顺序Order of work 先给一条审计流程定基线,在内部证明回报,再把方法打包给客户。Baseline one audit workflow, show the return inside the firm, then package the method for clients.
公开事实Public facts 已核对Checked
- 7532025 年调研的企业领袖Business leaders in the 2025 survey生成式 AI 采用指数Generative AI adoption index
- 83%加拿大受访者已超过规划阶段(AI in Finance 2026)Canadian respondents beyond planning, AI in Finance 2026KPMG 报告KPMG report
- 40全国办公室Offices across Canada合伙人 700 多名Over 700 partners
- 700+合伙人Partners总部在多伦多 333 Bay StreetHeadquartered at 333 Bay Street, Toronto
高采用,低证据High use, low proof 判断Judgment
- 93% 的企业在用 AIof businesses use AI
- 2% 见到可衡量回报see measurable returns
- KPMG 加拿大 2025 年调研,753 位企业领袖KPMG Canada survey, 753 business leaders, 2025
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Foteini Agrafioti
每个 AI 应用挂到事务所目标,各有一位负责人。Link each AI use to a firm goal; one owner each.
Erik Brynjolfsson
预期会有滞后;跟踪先行指标与基线。Expect a lag; track leading indicators and baselines.
Ziad Obermeyer
盘点所有模型;财务场景写清真实目标。List every model; name real target in finance work.
Ethan Mollick
在真实审计任务中测试,记录失败处。Test on real audit tasks; record where AI fails.
Andrew Ng
一条流程,三到六个月出数字。One workflow, numbers in three to six months.
Thomas Davenport
这是流程再造,由业务线负责人负责。Process redesign, owned by a practice leader.
Andrew Ng:六个月内出一个实测的胜利。Show a measured win within six months.
Erik Brynjolfsson:回报有滞后;别过早砍掉。Returns lag; do not cut early.
→ 按季度看先行指标;十二个月时评判。Quarterly leading indicators; judge at twelve months.
六位专家是谁、怎样思考、有什么铁律 →Who the six experts are, how they think, their rules →
三个目标Three priorities 建议Recommendation
给一条流程定基线Baseline one workflow
一条审计流程的工时、返工与复核时间。Hours, rework and review time on one audit workflow.
在内部证明回报Show the return inside
对比 AI 辅助与常规工作。Compare AI-assisted and standard work, side by side.
把方法打包Package the method
把基线方法做成帮助客户测回报的服务。Turn the baseline method into a client offer for returns.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 审计平台 AI 环节Audit platform AI steps 提效Efficiency | 在集中式审计应用平台中加入可测量的 AI 环节。Add measured AI steps to the central audit application hub. | 平台已上线(公开)· AI 环节为建议Platform live (public) · AI steps proposed |
| AI 回报测量工具包AI return measurement kit 增收Revenue体验Experience | 给客户 AI 项目用的基线与对照方法。A baseline-and-comparison method for clients' AI projects. | 建议Proposed |
| 财务智能体Agentic AI in finance 提效Efficiency降风险Risk | 智能体准备财务与审计工作,由专业人员复核。Agents prepare finance and audit work; professionals review. | 建议 · 人工复核Proposed · people review |
| 可信 AI 治理Trusted AI governance 降风险Risk增收Revenue | 把现有框架打包成可重复交付的客户服务。Package the existing framework into a repeatable client service. | 框架已公开 · 打包为建议Framework live (public) · packaging proposed |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 选定一条审计流程One audit workflow chosen
- 基线:工时、返工Baselines: hours, rework
- 点名业务线负责人Practice owner named
试点Pilot
- AI 辅助与常规并行AI-assisted run beside standard
- 记录失败案例Failure log kept
- 首份对比报告First comparison report
上线Go live
- 环节接入审计平台Steps join the audit platform
- 专业人员复核输出Professionals review outputs
- 上线即监控Monitoring from day one
扩展Extend
- 十二个月回报复盘Twelve-month return review
- 客户方法打包Client method packaged
- 年度重排优先级Annual re-ranking
治理与合规Governance 建议Recommendation
- AI 使用的职业与审计准则:向法律与风险部门确认Professional and audit standards on AI use: confirm with counsel and risk
- 客户数据进入 AI 工具:向法律顾问确认保密条款Client data in AI tools: confirm confidentiality terms with counsel
- 所有 AI 辅助的审计输出由专业人员复核签字Professionals review and sign all AI-assisted audit outputs
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 使用 AI 环节的项目团队Engagement teams using AI steps · 已有基线的流程数Workflows with a baseline |
|---|---|
| 过程Process | 每个流程环节的工时Hours per workflow step · 复核后的返工Rework after review · 记录的失败案例Failures logged |
| 结果Outcome | 相对基线的回报Return against baseline · 客户反馈的质量Client-reported quality |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
测得太晚Measuring too late
现在不定基线,回报将永远证明不了。Without a baseline now, the return can never be shown.
节省被吞掉Savings that vanish
起草省下的工时可能在复核中被吃回去。Hours saved on drafting may reappear as extra review.
没验证就建议Advising before proving
客户购买前可能先看事务所自己的结果。Clients may ask for the firm's own results before buying.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 点名业务线负责人;选一条流程Name practice owner; pick one workflow |
|---|---|
| 第 3–4 周Weeks 3–4 | 收集基线Collect baselines |
| 第 5–10 周Weeks 5–10 | AI 辅助与常规并行运行Run AI-assisted beside standard |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: 计入复核时间后对比显示没有收益,或客户不信任事务所自测。the comparison shows no gain after review time, or clients do not trust a firm measuring itself.
出处Sources
本页是基于公开资料的分析,与所述企业没有合作关系,不代表其官方立场或内部计划;企业数字多来自其自述或媒体转述;「专家组」是依据六位专家公开观点的推演,不是他们的原话,也不代表他们或其雇主的立场。资料整理日期:2026-11-14。This page is an analysis based on public information. It has no affiliation with the company and does not represent its official position or internal plans. Company figures come mostly from its own statements or media reports. The expert panel is an inference from six experts' public views, not their words, and does not represent them or their employers. Compiled 2026-11-14.
想给你的公司也做一份?Want one for your company?
同样的方法:公开事实、专属判断、12 个月路线和前 90 天。公司大小都可以聊,中文沟通。Same method: public facts, a company-specific judgment, a 12-month roadmap and the first 90 days. Any size of business; we can talk in Chinese or English.