起步顺序Order of work 先量成员来电与表单,再回答简单问题并由人兜底,最后复用交易台的自动化。Measure member contacts first, then answer the simple ones with a person behind, then reuse the desks' automation.
公开事实Public facts 已核对Checked
- 650,000+计划成员Plan members遍及 1,000 多家参保雇主Across 1,000+ participating employers
- ~170养老金服务与技术团队规模People in pension services and technologyAI 是该团队的核心工作之一A I is one of the team's core jobs
- 10–20%OMERS Ventures 投向 AI 原生企业的活动占比Share of Ventures activity in AI-native firms含 Cohere、WaabiIncludes Cohere and Waabi
- 20262026 年新任 Ventures 负责人明确 AI-firstNew Ventures head chose an AI-first focusLaura Lenz,聚焦加拿大 AI-first 公司Laura Lenz, Canadian AI-first companies
从 65 万成员到 AI 可办的工作From 650,000 members to work AI could finish 判断Judgment
- 65 万以上成员650,000+ members — 已知known
- 每年有多少人来问How many ask each year — 未公开,需测量not public — to be measured
- 其中 AI 能办完多少How many AI could finish — 未公开,需测量not public — to be measured
- 后两层公开资料没有数字,先把它们量出来。No public numbers for the last two stages. Measuring them is the first job.
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Erik Brynjolfsson
先做增强:帮新手像老手一样回答;上任何聊天机器人之前先定基线。Design augmentation first: help new staff answer like experienced ones; set baselines before any chatbot.
Ethan Mollick
用真实的退休问题测试;养老金答错代价高,边界摸清前由人复核。Test on real retirement questions; a wrong pension answer is costly, so a person reviews until the edge is known.
Ziad Obermeyer
退休答复影响成员决策:先定义什么是好答案,并按成员群体分别检查。Retirement answers shape people's decisions: define a good answer, and check results across member groups.
Foteini Agrafioti
养老金服务团队与投资交易台需要一页纸,写清谁负责哪项 AI 工作。The pension-services team and the investing desks need one page on who owns which AI work.
Andrew Ng
选一种表单或问题,三到六个月内出数字。Pick one form or question type that can show numbers in three to six months.
Thomas Davenport
把初级团队自发的自动化变成有负责人的共享工作流;这是流程改变。Turn junior teams' ad hoc automation into shared workflows with named owners; that is process change.
Ethan Mollick:初级团队自发的自动化是最好的信号;保持自由。Keep desks' self-started automation free.
Thomas Davenport:标准化为全机构、有负责人的工作流。Standardise into owned, shared workflows.
→ 自发工作放在沙盒;验证有效的升级为有负责人的工作流。Sandbox for self-started work; promote proven ones into owned workflows.
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
量清成员需求Size the member demand
按类型统计来电与表单,以及其中多少很简单。Count contacts and forms by type, and how many are simple.
回答成员,由人兜底Answer members, with a person behind
从简单问题开始;涉及权益的一律升级给人。Start with simple questions; escalate anything about entitlement.
让交易台自动化可复用Make desk automation reusable
把自发的做法升级为有负责人的共享工作流。Promote self-started work into shared, owned workflows.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 成员问答助手Member Q&A assistant 体验Experience提效Efficiency | 回答简单的退休问题,由人复核并接手升级件。Answer simple retirement questions; a person reviews and takes escalations. | 建议Proposed |
| 表单与请求处理Forms and requests processing 提效Efficiency | 读取并预填成员表单,提交前由员工确认。Read and pre-fill member forms; staff confirm before anything is filed. | 建议Proposed |
| 可复用的投研工作流Reusable analysis workflows 提效Efficiency增收Revenue | 把初级团队的自动化变成全机构可复用的工作流。Turn junior teams' automation into workflows the whole institution can reuse. | 建议 · 基于公开表态Proposed · builds on public statements |
| 尽调中的 AI 检查AI check in due diligence 降风险Risk增收Revenue | 在组合公司尽调中评估 AI 机会与风险。Assess AI opportunity and risk in portfolio-company due diligence. | 建议Proposed |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 统计来电与表单Count contacts and forms
- 模型清单与分级Model inventory and tiers
- 权责一页纸Ownership page
试点Pilot
- 一类问题,人工复核One question type, human review
- 对照组测量Control group measured
- 失败边界测试记录Failure-edge test log
上线Go live
- 助手上线处理简单问题Assistant live for simple questions
- 升级给员工Escalation to staff
- 上线即监控Monitoring from day one
扩展Extend
- 表单处理评审Forms processing review
- 交易台工作流升级Desk workflow promotion
- 年度重排优先级Annual re-ranking
治理与合规Governance 建议Recommendation
- 模型清单;与法律顾问确认 OSFI E-23 适用性Model inventory; confirm OSFI E-23 applicability with counsel
- 涉及福利权益的答复由人复核A person reviews any answer about benefit entitlement
- 按成员群体分别检查结果Results checked by member group
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 使用助手的成员数Members using the assistant · 每周活跃员工数Weekly active staff |
|---|---|
| 过程Process | 无需升级即办完的比例Answers finished without escalation · 处理时长Handling time · 复核通过率Review pass rate |
| 结果Outcome | 对照基线办完的联系数Contacts finished against baseline · 成员满意度Member satisfaction |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
没量清就开建Building before sizing
为没人统计过的需求做聊天机器人,可能什么也解决不了。A chatbot for demand nobody counted may solve nothing.
养老金答错Wrong pension answers
一次自信的错误答复可能改变退休决策。One confident wrong answer can change a retirement decision.
隐藏的自发自动化Hidden desk automation
没人负责的脚本会悄悄出错;纳入清单。Unowned scripts break silently; bring them into the inventory.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 点名来电基线负责人Name the owner of the contact baseline |
|---|---|
| 第 3–4 周Weeks 3–4 | 按类型抽样来电与表单Sample contacts and forms by type |
| 第 5–10 周Weeks 5–10 | 估算简单问题占比;分级初稿Size the simple share; draft tiers |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: 成员问题太多样、太个人化,难以自动化,AI 能办完的比例其实很小。member questions are too varied and personal for automation, so the share AI can finish turns out small.
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.