起步顺序Order of work 先定义断网与服务指标,再做网络预警,最后做客服助手。Define outage and service measures first, then network early warning, then care-centre assistants.
公开事实Public facts 已核对Checked
- C$10B三年网络可靠性投入Network reliability investment over three yearsCEO Tony Staffieri 宣布;含更大力度使用 AICEO Tony Staffieri; includes greater use of AI
- 2022促成该计划的全国大断网年份Year of the national outage behind the plan罗杰斯 CEO 事后宣布该计划Rogers' CEO announced the plan afterwards
- 5多伦多市中心 5G AI 交通试点路口数Downtown Toronto intersections in a 5G AI traffic pilot2024 年 3 月启动试点Pilot announced in March 2024
- C$20.6B2024 财年营收Revenue in fiscal 2024员工约 25,000 人About 25,000 employees
一个还没有尺子的承诺A promise with no ruler yet 判断Judgment
- 三年three years:100 亿加元可靠性计划C$10B reliability programme
- 其中多少是 AI、要防什么:未公开How much is AI, and what it must prevent: not public
- 先量清 AI 部分,再定断网指标;目标由罗杰斯自定。Size the AI part first, then set outage measures. Targets are Rogers' to set.
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Foteini Agrafioti
网络、客服与媒体一页权责;每个 AI 事项挂到可靠性或服务目标。One page of ownership across network, care and media; tie each AI item to a reliability or service target.
Erik Brynjolfsson
客服端按增强设计,优先帮新手;量处理时长与重复来电基线。In care, build for augmentation and help new agents first; baseline handle time and repeat contacts.
Ziad Obermeyer
断网预测要写清真实目标(客户影响)与代理指标(告警数)。For outage prediction, name the real target, customer impact, versus the proxy, alarms raised.
Ethan Mollick
用真实工单测出客服助手在哪里失败;由坐席复核。Test the care assistant on real tickets to find where it fails; agents review.
Andrew Ng
选一段网络做预警,三到六个月内出数字。Pick one network segment for early warning, with numbers in three to six months.
Thomas Davenport
这是网络运维方式的改变,不是外挂;点名网络运维负责人。This changes how network operations run, not a bolt-on; name a network-operations owner.
Andrew Ng:在一段网络上先启动预警,快速学习。Start early warning on one segment; learn fast.
Ziad Obermeyer:先定目标,再谈预警。Define the target before alerts.
→ 先跑影子模式;工程师确认预警后才有自动动作。Shadow mode first; engineers confirm alerts before any automatic action.
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
让承诺可度量Make the promise measurable
先定义断网与服务指标并设基线,再动手建设。Define outage and service measures, with baselines, before building.
网络故障预警Early warning on the network
预测故障、规划容量,工程师确认预警。Predict faults and plan capacity; engineers confirm alerts.
一次联系解决更多问题Resolve more in one contact
为客服坐席提供助手与工单分流。An assistant and ticket routing for care agents.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 网络故障预警Network early warning 降风险Risk体验Experience | 先在一段网络上做故障预测,再逐步扩大。Predictive AI for faults on one segment first, then wider. | 建议 · 基于公开承诺Proposed · builds on a public commitment |
| 容量规划Capacity planning 提效Efficiency降风险Risk | 预测负载,在需要之前扩容。Forecast load so capacity is added before it is needed. | 建议Proposed |
| 客服助手与工单分流Care assistant and ticket routing 提效Efficiency体验Experience | 建议答案并分流工单,由坐席确认。Suggest answers and route tickets; agents confirm. | 建议Proposed |
| 媒体内容推荐Media recommendations 增收Revenue体验Experience | 为 Sportsnet 做个性化推荐与广告定向。Personalised recommendations and ad targeting for Sportsnet. | 建议Proposed |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 断网与服务指标Outage and service measures
- 设定基线Baselines set
- 模型清单Model inventory
试点Pilot
- 影子模式预警Early warning in shadow mode
- 真实工单上的客服助手Care assistant on real tickets
- 对照组测量Control group measured
上线Go live
- 预警触发动作Warning drives action
- 坐席确认关键步骤Agents confirm key steps
- 上线即监控Monitoring from day one
扩展Extend
- 更多网络段More network segments
- 容量规划Capacity planning
- 年度重排优先级Annual re-ranking
治理与合规Governance 建议Recommendation
- 自动动作之前由工程师确认预警Engineers confirm alerts before any automatic action
- 涉及客户的步骤由人确认A person confirms customer-affecting steps
- 与法律顾问确认适用的行业规则,并建模型清单Confirm applicable sector rules and model inventory, with counsel
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 周活跃客服坐席数Weekly active care agents · 覆盖的网络段数Network segments covered |
|---|---|
| 过程Process | 预警精确率与召回率Alert precision and recall · 客服处理时长Care handle time · 坐席确认率Agent confirmation rate |
| 结果Outcome | 相对基线的断网时长Outage minutes vs baseline · 每个问题的重复联系数Repeat contacts per issue |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
AI 被计入却未被度量AI counted, not measured
没有断网指标,AI 部分无法证明价值。Without outage measures, the AI part cannot show its value.
预警疲劳Alert fatigue
误报太多会让工程师不再理会。Too many false alarms teach engineers to ignore them.
客服自动化惹恼客户Care automation upsets customers
必须由坐席确认,客户要能找到真人。Agents must confirm, and customers must reach a person.
前 90 天First 90 days 建议Recommendation
| 第 1–2 周Weeks 1–2 | 点名负责人;起草断网指标Name the owner; draft outage measures |
|---|---|
| 第 3–6 周Weeks 3–6 | 确定基线;盘点模型Fix baselines; inventory models |
| 第 7–12 周Weeks 7–12 | 选定一段网络与两个客服任务Pick one segment and two care tasks |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: AI 在该计划中占比本就很小,或预警来得太晚、防不了什么。the AI share of the programme is already small, or alerts arrive too late to prevent anything.
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.