英文旁白、无字幕的 3–4 分钟管理层简报。3–4 minute executive brief, English narration, no captions. 在 YouTube 打开Open on YouTube ↗
起步顺序Order of work 先立规则,再做两个职能,最后铺到全公司。Rules first, two functions second, the whole firm last.
公开事实Public facts 已核对Checked
- C$60.4B管理资产与收费资产合计Assets under management plus fee-earning assets2025 年 11 月 30 日;上年同期 536 亿At 30 Nov 2025; C$53.6B a year earlier
- 6242025 财年员工Employees in fiscal 2025其中 603 人在加拿大603 in Canada
- 820,000+服务的投资者Investors served三条业务线Across three business lines
- AI CoE正在组建的内部 AI 专家中心In-house AI Centre of Expertise being built来自公开招聘信息,非 AGF 官宣From public job postings, not an AGF announcement
按顺序爬四级Four rungs to climb in order 判断Judgment
- 1 · 先立规则1 · Rules first:负责人、风险分级、审查Owners, risk tiers, review
- 2 · 两个职能2 · Two functions:先做投资与分销Investing and distribution go first
- 3 · 一套共用工具3 · One shared kit:智能体与检索,复用Agents and search, reused
- 4 · 全公司4 · Whole firm:其他职能过关后加入Others join through the gate
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Foteini Agrafioti
写一页纸:AI 中心、平台团队与投资团队谁负责什么。Write one page: who owns the AI centre, the platform team, and the investment desks.
Erik Brynjolfsson
团队精简:预算要含流程再造与培训;先量每项任务工时基线。Lean firm: budget for process redesign and training; baseline hours per task.
Ziad Obermeyer
研究助手要写清优化什么;所有模型与提示库先建清单。For research copilots, name what they optimise; inventory every model and prompt kit.
Ethan Mollick
让分析师用真实公司资料测试,并记录出错处。Let analysts test the copilot on real company files and log where it fails.
Andrew Ng
选一个投研任务(如汇总披露文件),三到六个月内出数字。Pick one research task, like summarising filings, with a number in three to six months.
Thomas Davenport
这是投资流程的改变,还是一个工具?点名流程负责人。Is this an investment-process change or a tool? Name the process owner.
Andrew Ng:早点交付一个看得见的成果,赢得信任。Ship one visible win early to earn trust.
Foteini Agrafioti:规则落到一页纸之前不扩大。No scaling before rules are on one page.
→ 试点走轻量规则;扩大须过中心的关。Pilot under a light rule set; scaling waits for the centre's gate.
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
先定治理再扩大Settle governance before scale
负责人、风险分级与审查规则,由中心统一写定。Owners, risk tiers and review rules, written once for the centre.
先赢一个投研流程Win one research workflow
对披露文件做检索与摘要,分析师核对来源。Search and summaries over filings; analysts check sources.
延伸到顾问分销Extend to advisor distribution
复用同一套工具做客户洞察与内容起草。Reuse the same kit for client insights and content drafts.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 带检索的投研助手Research copilot with search 提效Efficiency体验Experience | 在披露与笔记上做检索,让分析师更快找到来源。Retrieval over filings and notes so analysts find sources faster. | 建议 · 基于在招的团队Proposed · builds on a team being hired |
| 分销洞察助手Distribution insight assistant 增收Revenue体验Experience | 为顾问生成客户洞察与内容草稿,使用前审阅。Client insights and content drafts for advisors, reviewed before use. | 建议Proposed |
| 共用智能体与流程工具Shared agent and workflow kit 提效Efficiency | 一套可复用的智能体与流程,供各职能使用。One reusable set of agents and workflows for every function. | 建议Proposed |
| 模型与提示库清单Model and prompt inventory 降风险Risk | 持续更新的工具、负责人与风险分级登记册。A living register of tools, owners and risk tiers. | 建议Proposed |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 一页权责One page of ownership
- 风险分级与审查规则Risk tiers and review rules
- 每项任务工时基线Baseline hours per task
试点Pilot
- 披露文件检索Research search on filings
- 对照组测量Control group measured
- 记录失败案例Failure log kept
上线Go live
- 分销助手上线Distribution assistant live
- 分析师核对来源Analysts review sources
- 上线即监控Monitoring from day one
扩展Extend
- 向新职能开放工具Open the kit to new functions
- 按职能过关评审Gate review per function
- 年度重排优先级Annual re-ranking
治理与合规Governance 建议Recommendation
- 模型、提示库与负责人清单Inventory of models, prompt kits and owners
- 每份投研输出由分析师核对来源Analysts check sources on every research output
- 与法律顾问确认 OSFI E-23 是否适用Confirm whether OSFI E-23 applies, with counsel
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 各职能周活跃使用率Weekly active use by function · 对披露文件的检索次数Searches run on filings |
|---|---|
| 过程Process | 来源核对通过率Source-check pass rate · 每项投研任务时间Time per research task · 通过准入的职能数Functions through the gate |
| 结果Outcome | 相对基线返还的工时Hours returned vs baseline · 人均营收Revenue per employee |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
规则未立就扩大Scaling before the rules
没有负责人的快工具会带来影子使用。Fast tools without owners create shadow use.
研究里的自信性错误Confident research errors
助手可能错引;分析师必须核对来源。Copilots can cite wrongly; analysts must check sources.
小团队被摊薄A small team stretched thin
精简团队跑不了六个试点;只选两个职能。A lean firm cannot run six pilots; pick two functions.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 点名中心负责人;起草一页权责Name the centre's owner; draft one-page ownership |
|---|---|
| 第 3–6 周Weeks 3–6 | 起草风险分级;量投研工时基线Draft risk tiers; baseline research hours |
| 第 7–12 周Weeks 7–12 | 选定首个投研任务;定审查规则Pick the first research task; set review rules |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: 分析师过度信任助手,或精简团队撑不起两个职能并行。analysts trust the copilot too much, or the lean team cannot support two functions at once.
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.