量身 AI← 案例库← Case library
案例 · AI 落地方案CASE · AI DEPLOYMENT PLAN · 资本市场基础设施 · 多伦多Capital markets infrastructure · Toronto

TMX 集团TMX Group

让 AI 的自主程度跟随出错的代价Let AI autonomy follow the cost of being wrong
基于公开资料的分析,与该企业没有合作关系。Based on public information; no affiliation with the company.
视频制作中,上线后会出现在这里。Video coming soon — it will appear here.
一句话判断The verdict AI 横跨研究产品与实时市场系统;两者出错的代价天差地别。A I spans research products and live market systems; the risk of an error differs hugely between them.
起步顺序Order of work 先在出错代价低处落地,再做仅告警的监控,最后才触及交易与清算。Ship AI first where errors are cheap, then monitoring in alert-only mode, and touch trading and clearing last.

公开事实Public facts 已核对Checked

  • Evolve
    战略支柱:用数据、AI 与技术打造新解决方案Strategy pillar: use data, AI and technology to build new solutions写在 2025 年 MD&A 中Stated in the 2025 MD&A
  • 300+
    Verity 服务的基金数Funds served by Verity2025 年 10 月收购,并入 TMX DatalinxAcquired October 2025, folded into TMX Datalinx
  • Splunk
    AI 预测分析系统监控的合作方Partner for A I predictive analytics on system monitoring首席技术官称预测分析是未来最有价值的能力Its CTO calls predictive analytics the most valuable future capability
  • 2,185
    2025 年末员工数(全职当量)Employees at end of 2025, full-time equivalent年报口径Annual-report basis

自主程度应跟随出错的代价Autonomy should follow the cost of being wrong 判断Judgment

  • 研究工具Research tools:面向基金客户的数据产品与研究工作流Data products and research workflows for fund clients(每个数字都能追溯到来源吗?Is every number traceable to source?)
  • 披露审核Disclosure review:为上市公司公告与披露审核提供辅助Announcement and filing review support for listed companies(谁对最终判读签字?Who signs the final reading?)
  • 系统监控System monitoring:对市场系统做预测与预警Prediction and early warning on market systems(一次误报的代价是什么?What does a false alarm cost?)
  • 交易与清算Trading and clearing:任何可能暂停或改变实时市场的事Anything that can pause or change a live market(模型失效时的后备方案是什么?What is the fallback if the model fails?)

专家组裁决Expert panel 依据公开观点的推演Inference from public views

Foteini Agrafioti

技术与业务战略对齐Strategy alignment

把 Evolve 挂到点名的产品与运营目标;一页纸写清 Datalinx AI、Verity 与平台的负责人。Tie Evolve to named product and operations targets; one page on who owns Datalinx A I, Verity and the platform.

Erik Brynjolfsson

价值与度量Value & measurement

收购的价值随流程再造而来;定好客户使用与分析师时间基线,并为整合出预算。Value from an acquisition arrives with process change; baseline client usage and analyst time, and budget integration.

Ziad Obermeyer

算法在优化什么What it optimizes

异常检测预测的是「异常」而非「有害」;定义目标,并按市场板块测误报。Anomaly detection predicts unusual, not harmful; define the target and test false alarms by market segment.

Ethan Mollick

人怎么用How people use it

运维人员最清楚监控在哪失灵;信任新告警之前先回放历史事件。Operations staff know where monitoring fails; replay past incidents before trusting new alerts.

Andrew Ng

节奏与方法Pace & method

Splunk 监控是快赢点:一个系统,三个月出数字,再延伸到清算。Splunk monitoring is the quick win: one system, a number in three months, then extend to clearing.

Thomas Davenport

整家企业怎么变Whole-enterprise change

TMX 既卖数据又运营市场;AI 同时改变产品与运营。各点名一位负责人。T M X sells data and runs markets; A I changes both product and operations. Name an owner for each.

保留的分歧Disagreement kept · 快速延伸还是先审查Extend fast vs. review first

Andrew Ng:快速延伸监控;它已与 Splunk 一起运行。Extend monitoring fast; Splunk already runs.

Ziad Obermeyer:市场里的异常检测在影响任何动作前,须先审查目标与误报。Review false alarms before any action.

→ 先仅告警;审查通过前不对市场系统做自动动作。Alert-only first; no automatic action on market systems until review passes.

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

1

让 Datalinx 成为 AI 产品线Make Datalinx the A I product line

Verity 数据加预测模型,有负责人与客户基线。Verity data plus predictive models, with an owner and a client baseline.

2

在事件停摆市场前预测Predict incidents before they stop a market

以仅告警模式延伸 Splunk 监控,并用历史事件评分。Extend Splunk monitoring in alert-only mode, scored against history.

3

加快披露审核,人工签字Speed disclosure review, people signing

助手标出缺漏;每一项由审核人签字。An assistant that flags gaps; a reviewer signs every item.

场景组合Scenarios 建议Recommendation

场景Scenario做法Approach状态Status
Verity 数据加预测模型Verity data plus predictive models
增收Revenue
把 Verity 数据集与预测模型结合,服务基金客户。Combine Verity datasets with predictive models for fund clients.建议 · 基于 Verity 收购Proposed · builds on the Verity deal
智能投研工作流Intelligent research workflows
增收Revenue体验Experience
研究管理工作流,每个数字附来源。Research-management workflows with sources on every number.建议Proposed
市场系统异常检测Anomaly detection for markets
降风险Risk提效Efficiency
对交易与清算系统做仅告警的检测,用历史事件评分。Alert-only detection on trading and clearing systems, scored on past incidents.建议 · 完整审查Proposed · full review
披露审核助手Disclosure review assistant
提效Efficiency
标出公告与披露中的缺漏,审核人签字。Flag gaps in announcements and filings; a reviewer signs.建议Proposed

12 个月路线12-month roadmap 建议Recommendation

奠基Foundation

第 1–3 月Months 1–3
  • 客户使用基线Client usage baseline
  • 模型清单与风险分级Model inventory and risk tiers
  • 一页权责One page of ownership
◆ 分级定稿Tiers final

试点Pilot

第 4–6 月Months 4–6
  • 仅告警监控回放Alert-only monitoring replay
  • 统计误报False alarms counted
  • Verity 数据试点Verity data pilot
◆ 指标达成Targets met

上线Go live

第 7–9 月Months 7–9
  • 告警面向运维上线Alerts live for operators
  • 每项动作由人决定Person decides every action
  • 上线即监控Monitoring from day one
◆ 风险合规签字Risk sign-off

扩展Extend

第 10–12 月Months 10–12
  • 清算准入评审Clearing gate review
  • 披露助手试点Disclosure assistant pilot
  • 年度重排优先级Annual re-ranking
◆ 进入下一年Next year begins

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
  • 告警只提示人;审查通过前不对实时市场做自动动作Alerts advise people; no automatic action on live markets until reviewed
  • 数据产品中的每个数字可追溯到来源Every number in a data product traces to source

三层指标Three-layer metrics 建议Recommendation

采用Adoption客户对 AI 数据产品的使用Client use of A I data products · 每周使用告警的运维人员Operators using alerts weekly
过程Process按板块的误报率False-alarm rate by segment · 对历史事件的提前量Alert lead time on past incidents · 来源追溯覆盖率Source-trace coverage
结果Outcome对照基线的 Datalinx 收入Datalinx revenue against baseline · 对照基线的事件时长Incident minutes against baseline

目标值由企业自定。Targets are the company's to set.

风险Risks 判断Judgment

告警疲劳Alert fatigue

误报太多会让运维学会忽略真正的告警。Too many false alarms train operators to ignore the real one.

买了数据,流程没变Acquired data, unchanged process

Verity 的价值需要新工作流,而不只是新数据集。Verity's value needs new workflows, not just a new dataset.

靠近实时市场的自动化Automation near live markets

一次自信的错误动作可能让市场暂停;保持由人决策。A confident wrong action can pause a market; keep people deciding.

前 90 天First 90 days 建议Recommendation

第 1–2 周Weeks 1–2点名 Datalinx AI 与监控的负责人Name owners for Datalinx A I and monitoring
第 3–4 周Weeks 3–4盘点模型;定客户使用基线Inventory models; baseline client use
第 5–10 周Weeks 5–10风险分级;收集历史事件用于回放Risk tiers; collect past incidents for replay
第 13 周Week 13首次复盘First review
今天只需要一个决定:One decision today: 指定 Datalinx 产品线中 AI 的负责人name the owner of A I in the Datalinx product line

我们可能错在:We could be wrong if: 客户更要原始数据而非 AI 衍生结果,或现有监控已能抓住重要事件。clients prefer raw data to A I-derived outputs, or current monitoring already catches the incidents that matter.

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.

Telegram @johnsunlei邮件 Email [email protected]