起步顺序Order of work 先把加拿大有效的做法打包,复制到美国一条业务线,再打通理赔。Package what works in Canada, copy it to one U.S. line, then take claims end to end.
公开事实Public facts 已核对Checked
- 600+Intact Lab 中的数据与 AI 专家Data and AI specialists in Intact Lab数据科学家、精算师、工程师Data scientists, actuaries, engineers
- 90%+机器学习支撑的加拿大个人险定价占比Canadian personal-lines pricing supported by ML商业险定价约 60%About 60% of commercial lines pricing
- 24%Underwriting Advisor 带来的报价处理量提升Quote volume lift from Underwriting Advisor56% 的核保字段自动预填56% of underwriting fields prefilled
- 4thEvident 2025 保险 AI 指数排名Evident 2025 insurance AI Index rank在保险公司中的排名Ranked among insurers
出海复制的三个层级Three tiers for copying abroad 判断Judgment
- 原样复制:Underwriting AdvisorCopy as is: Underwriting Advisor:本地负责人、基线、数据接入a local owner, a baseline, a data feed
- 先适配:理赔智能体Adapt first: claims agent:本地规则、流程负责人、边界测试local rules, a process owner, edge-case tests
- 严格审查:定价与反欺诈模型Review hard: pricing and fraud models:目标与代理指标、按客户群看结果target versus proxy, results by customer group
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Thomas Davenport
复制是每个国家的流程改变;按业务线点名负责人。Copying is a process change in each country; name owners per line.
Erik Brynjolfsson
按基线核算 2 亿加元+;为新市场预算再培训。Count the C$200M+ against baselines; budget retraining for new markets.
Foteini Agrafioti
Lab 与各国团队的权责写成一页纸。Write one page on who owns what between the Lab and each country.
Ethan Mollick
放大前,在真实案件上测出理赔智能体的失败点。Test where claims agents fail on real files before widening.
Ziad Obermeyer
定价与反欺诈:写清真实目标与代理指标;按人群检查。For pricing and fraud, name the real goal and the proxy; check by group.
Andrew Ng
六个月内把加拿大工具复制到美国一条业务线。Copy the Canadian tool into one U.S. line within six months.
Andrew Ng:把已验证的工具快速复制到美国一条线。Copy the proven tool to one U.S. line fast.
Thomas Davenport:复制前先理顺当地流程。Fix the local process before copying.
→ 先在一条线复制工具,并先适配当地流程。Copy the tool to one line; adapt the process there first.
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
打包加拿大打法Package the Canadian playbook
把工具、数据管道与培训做成可交付的工具包。Tools, data pipes and training as a kit another country can take.
复制到美国一条业务线Copy to one U.S. line
先复制 Underwriting Advisor,配本地负责人与基线。Underwriting Advisor first, with a local owner and baseline.
打通理赔Take claims end to end
报案、定损、反欺诈与赔付,例外由人处理。Report, assessment, fraud and payment, with a person on exceptions.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 美英复制工具包U.S. and U.K. replication kit 提效Efficiency增收Revenue | 标准工具包加本地负责人,把部署率提到约 30% 以上。A standard kit and local owners to lift deployment beyond about 30%. | 建议 · 管理层视其为增长杠杆Proposed · management names this a lever |
| 理赔智能体,报案到赔付Claims agent, report to payout 提效Efficiency体验Experience降风险Risk | 覆盖受理、定损与反欺诈的智能体,例外由人决定。Agents for intake, assessment and fraud checks; a person decides exceptions. | 建议 · 完整审查Proposed · full review |
| 气候风险定价Climate-risk pricing 降风险Risk增收Revenue | 深化风险定价;CEO 把 AI 视为应对气候损失的核心。Deepen risk-based pricing, where the CEO places AI against climate losses. | 建议Proposed |
| 通向 2028 的价值台账Value ledger to 2028 提效Efficiency | 用一本台账追踪向 5 亿加元+目标的进度。One ledger tracking progress toward the C$500M+ target. | 建议Proposed |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 打法工具包初稿Playbook kit drafted
- 价值台账与基线Value ledger and baselines
- 权责页Ownership page
试点Pilot
- 美国一条线复制工具One U.S. line copies the tool
- 对照组Control group
- 本地负责人Local owner named
上线Go live
- 理赔智能体上一条业务Claims agent on one book
- 例外由人决定Person decides exceptions
- 上线即监控Monitoring from day one
扩展Extend
- 第二条美英业务线Second U.S. or U.K. line
- 跨业务线复用Reuse across lines
- 年度重排优先级Annual re-ranking
治理与合规Governance 建议Recommendation
- 加拿大模型清单对齐 OSFI E-23(2027-05-01 生效);美英规则请法务确认Canadian model inventory aligned to OSFI E-23 (effective 2027-05-01); confirm U.S. and U.K. rules with counsel
- 定价与反欺诈模型:完整审查Pricing and fraud models: full review
- 理赔例外由人决定A person decides claim exceptions
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 每周使用工具的核保人Underwriters using the tool weekly · 已采用工具包的业务线Lines running the kit |
|---|---|
| 过程Process | 报价处理量Quote throughput · 理赔周期Claim cycle time · 例外率Exception rate |
| 结果Outcome | 对照 5 亿加元+目标的价值Value against the C$500M+ target · 赔付率对照基线Loss ratio against baseline |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
工具包水土不服A kit that does not travel
各国数据与流程不同;复制工具,适配流程。Data and processes differ by country; copy the tool, adapt the process.
定价中的代理指标Pricing proxies
模型可能优化风险的代理指标;按客户群检查。A model may optimise a proxy for risk; check results by customer group.
价值重复计算Double counting value
模型众多时收益易重复计算;用同一本台账。With many models, gains can be counted twice; use one ledger.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 点名工具包负责人;定基线Name the kit owner; fix baselines |
|---|---|
| 第 3–4 周Weeks 3–4 | 按国家盘点模型与工具Inventory models and tools by country |
| 第 5–10 周Weeks 5–10 | 起草工具包;选定美国一条线Draft the kit; pick one U.S. line |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: 美英部署率低是因为数据与规则不同,而不是复制得慢。the low U.S. and U.K. rate reflects different data and rules rather than slow copying.
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.