起步顺序Order of work 先选一类理赔,把任务串成流程并测边界,再推广到顾问。Choose one claim type, link its tasks into a journey, test the edges, then spread to advisers.
公开事实Public facts 已核对Checked
- C$1.60T2025 年末管理资产Assets under management, end of 2025财富与资产管理规模Wealth and asset management scale
- 32,1512025 年末全职等效员工Full-time equivalent employees, end of 2025不含亚洲合资公司Excludes Asian joint ventures
- +5.9%2025 年保险收入同比增长Insurance revenue growth in 2025保险收入 23,982 百万加元C$23,982M insurance revenue
- Nov 2026获评加拿大最佳数字保险公司Named Best Digital Insurer in Canada同时获最佳移动保险 AppAlso best mobile insurance app
从任务级到流程级From task level to journey level 判断Judgment
- 任务级,目前在用Task level, used today:单据到达A document arrives → 智能体做一个任务Agent does one task → 其余由人完成A person runs the rest
- 流程级,建议Journey level, proposed:报案Claim reported → 智能体推进到赔付Agent carries it to payment → 人处理例外A person decides exceptions
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Foteini Agrafioti
把流程挂到一个业务目标;点名谁负责交接。Tie the journey to one business target; name who owns the handoffs.
Erik Brynjolfsson
先增强理赔员;对照基线与对照组度量。Augment adjusters first; measure against a baseline and a control group.
Ziad Obermeyer
健康险理赔:写清真实目标与代理指标;不由模型单独拒赔。In health claims, name the real goal and the proxy; no denial by model alone.
Ethan Mollick
放大之前,先测出智能体在杂乱医疗单据上的失败点。Map where agents fail on messy medical documents before widening.
Andrew Ng
从一类三到六个月内出数字的理赔开始。Start with one claim type that shows numbers in three to six months.
Thomas Davenport
这是理赔流程的端到端改变;点名流程负责人。This changes the claims process end to end; name the process owner.
Erik Brynjolfsson:让理赔员在回路中,以增强为主。Keep adjusters in the loop and augment them.
Andrew Ng:把最简单的理赔全自动,快出数字。Fully automate the simplest claims to show numbers.
→ 仅对最低风险理赔全自动;不做自动拒赔。Full automation only for the lowest-risk claims; no automated denials.
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
选定一条理赔流程Pick one claim journey
选一类健康险理赔,画出每个任务与交接,定基线。Choose a health claim type, map each task and hand-off, set a baseline.
把任务串成流程Link tasks into a journey
从报案到赔付的智能体流程,例外由人处理。One agent flow from report to payment, with a person on exceptions.
再赋能顾问Equip advisers next
把同一批能力复用为顾问助手。Reuse the same skills as an assistant for advisers.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 健康险理赔受理Health claim intake 提效Efficiency体验Experience | 读取并核对医疗单据,把干净的理赔往下传。Read and check medical receipts, then hand clean claims onward. | 建议 · 基于任务级智能体Proposed · builds on task-level agents |
| 核保辅助Underwriting support 提效Efficiency降风险Risk | 汇总申请并标出缺口,由核保人决定。Summarise applications and flag gaps; underwriters decide. | 建议Proposed |
| 报案到赔付智能体Report-to-payment agent 提效Efficiency体验Experience降风险Risk | 贯穿整个理赔的智能体流程,例外由人处理。One agent flow across the claim, with a person on exceptions. | 建议 · 完整审查Proposed · full review |
| 顾问助手Adviser assistant 增收Revenue体验Experience | 为加拿大与亚洲约 99,000 名顾问提供助手。An assistant for the roughly 99,000 advisers across Canada and Asia. | 建议Proposed |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 选一类理赔并画流程One claim type, mapped
- 基线与负责人Baselines and owner
- 模型清单与分级Model inventory, tiers
试点Pilot
- 受理智能体处理真实理赔Intake agent on live claims
- 对照组Control group
- 边界案例记录Edge-case log
上线Go live
- 任务串到赔付Link tasks to payment
- 例外由人决定Person decides exceptions
- 上线即监控Monitoring from day one
扩展Extend
- 第二类理赔Second claim type
- 顾问助手试点Adviser assistant pilot
- 年度重排优先级Annual re-ranking
治理与合规Governance 建议Recommendation
- 模型清单对齐 OSFI E-23(2027-05-01 生效)Model inventory aligned to OSFI E-23 (effective 2027-05-01)
- 不由模型单独拒赔No claim denial by a model alone
- 例外与涉及客户权益的步骤由人决定A person decides exceptions and customer-affecting steps
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 经智能体流转的理赔数Claims routed through the agent · 每周使用助手的理赔员Adjusters using the assistant weekly |
|---|---|
| 过程Process | 直通率Straight-through rate · 报案到赔付时长Time from report to payment · 例外率Exception rate |
| 结果Outcome | 单笔理赔成本对照基线Cost per claim against baseline · 理赔客户满意度Customer satisfaction on claims |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
错误沿流程传递Errors travel along the journey
一张读错的单据可能一路走到赔付;每个交接点都要核对。One misread receipt can reach payment; check at each hand-off.
更快不等于更公平Faster is not fairer
若模型把成本当作需求的替代,部分客户会吃亏。If the model learns cost as a stand-in for need, some members lose out.
理赔员绕开它Adjusters route around it
没有培训与流程再造,大家会沿用旧流程。Without training and redesign, people keep the old process.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 点名流程负责人;定基线Name the journey owner; fix baselines |
|---|---|
| 第 3–4 周Weeks 3–4 | 画出一类理赔的任务与交接Map tasks and hand-offs for one claim type |
| 第 5–10 周Weeks 5–10 | 风险分级初稿;用真实单据测试Risk tiers drafted; test on real receipts |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: 任务级智能体已拿走大部分价值,串成流程只增加风险、收益不大。task-level agents already capture most of the value, and linking them adds risk without much gain.
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.