起步顺序Order of work 先选一个门店流程,加上由人批准的动作并量化,再复制到下一个。Pick one store workflow, add a person-approved action, measure it, then copy to the next.
公开事实Public facts 已核对Checked
- 2,5042025 年末门店数Stores at the end of 20252024 年末为 2,492 家2,492 at the end of 2024
- 1B每年交易次数Transactions a year约 10 亿次Roughly one billion
- ChatGPTPC Express 的 ChatGPT 渠道:转化率第二高PC Express channel: second-highest converting仅次于 Digital Flyer;上线约半年After Digital Flyer; about six months live
- Shakudo统一企业 AI 平台的合作方Partner for one enterprise A I platform2026 年 5 月 4 日宣布;数据留在 Loblaw 自有云Announced May 4, 2026; data stays in Loblaw's cloud
从「提建议」到「办动作」From advice to action 判断Judgment
- Robin 现状Robin today:门店数据Store data → Robin 排优先级Robin ranks priorities → 店长自行处理Manager acts alone
- 建议方案Proposed:门店与供应商数据Store and supplier data → 起草补货与损耗预警Draft reorder and waste alert → 店长批准Manager approves → 收益对照基线记录Gain logged against baseline
专家组裁决Expert panel 依据公开观点的推演Inference from public views
Erik Brynjolfsson
先设基线和对照门店,再数收益。Baseline and control stores before counting any gain.
Ethan Mollick
先测出智能体在真实货架与后仓会读错什么。Test where agents misread real shelves and back rooms.
Thomas Davenport
这是补货流程的改变,不只是工具;点名负责人。This changes replenishment, not just a tool; name its owner.
Andrew Ng
一个流程、少数门店,三到六个月出数字。One workflow, a few stores, numbers in three to six months.
Foteini Agrafioti
每个智能体挂到门店目标;权责一页纸。Tie each agent to a store target; one page of ownership.
Ziad Obermeyer
药房与定价:写清真实目标与代理指标。For pharmacy and pricing, name the target and the proxy.
Andrew Ng:先做一个流程。Start with one workflow.
Thomas Davenport:统一在一个平台上建。Build on one platform.
→ 一个流程,建在共用平台上。One workflow, on the shared platform.
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 Robin 在批准下办事Let Robin act, with approval
起草补货与损耗预警,由店长批准。Drafts for reorder and waste; the manager approves.
逐店证明收益Prove the gain, store by store
设基线与对照门店,收益只算一次。Baselines and control stores, so the gain is counted once.
给个性化与药房把关Gate personalisation and pharmacy
先过隐私与定价公平检查;与法律顾问、临床治理确认。Privacy and pricing-fairness checks first; confirm with counsel and clinical governance.
场景组合Scenarios 建议Recommendation
| 场景Scenario | 做法Approach | 状态Status |
|---|---|---|
| 自动补货草稿Auto-reorder drafts 提效Efficiency | Robin 依据库存与销售规律起草订单,店长批准。Robin drafts orders from stock and sales patterns; the manager approves. | 建议 · 基于已上线 RobinProposed · builds on live Robin |
| 损耗预警Waste and shrink alerts 提效Efficiency降风险Risk | 在亏钱之前提示负毛利商品与后仓积压。Flag negative-margin items and back-room backlog before they cost money. | 建议Proposed |
| 忠诚度个性化Loyalty-led offers 增收Revenue降风险Risk | 用 PC Optimum 数据做个性化,先过隐私与定价公平审查。Use PC Optimum data for offers, after privacy and pricing-fairness review. | 建议 · 完整审查Proposed · full review |
| 药房助手Pharmacy assistant 体验Experience降风险Risk | 在 Shoppers 药房网络做用药提醒与随访支持。Medication reminders and follow-up support across the Shoppers network. | 建议 · 临床审查Proposed · clinical review |
12 个月路线12-month roadmap 建议Recommendation
奠基Foundation
- 选定一个流程与负责人Pick one workflow and owner
- 基线与对照门店Baselines, control stores
- 模型清单与风险分级Model inventory and risk tiers
试点Pilot
- 试点门店补货草稿Reorder drafts in pilot stores
- 店长测试边界Managers test the edges
- 首份三层指标报告First three-layer report
上线Go live
- 扩大到更多门店Roll out to more stores
- 加入损耗预警Waste alerts added
- 上线即监控Monitoring from day one
扩展Extend
- 忠诚度个性化隐私审查Privacy review for loyalty offers
- 药房场景立项Pharmacy scoping
- 重排下年优先级Re-rank next year
治理与合规Governance 建议Recommendation
- 定价与忠诚度数据:隐私与公平审查,法律顾问参与Pricing and loyalty data: privacy and fairness review, with counsel
- 药房 AI:与临床治理确认Pharmacy A I: confirm with clinical governance
- 库存与影响顾客的动作由人批准A person approves stock and customer-affecting actions
三层指标Three-layer metrics 建议Recommendation
| 采用Adoption | 每周活跃使用 Robin 的店长数Weekly active managers using Robin · 每店生成的草稿数Drafts created per store |
|---|---|
| 过程Process | 草稿被批准比例Share of drafts approved · 每项任务的店长用时Manager time per task · 缺货率Out-of-stock rate |
| 结果Outcome | 对照门店的损耗差异Waste against control stores · 被提示商品的毛利Margin on flagged items |
目标值由企业自定。Targets are the company's to set.
风险Risks 判断Judgment
收益重复计算Gain counted twice
没有对照门店,季节性销售会被算到智能体头上。Without control stores, seasonal sales will be credited to the agent.
店长不采纳草稿Managers overrule drafts
草稿若不贴合真实货架,信任会很快下降;先测试。If drafts miss real shelf conditions, trust drops fast; test first.
定价里的代理目标Proxy goals in pricing
优化毛利不等于对顾客公平的价值。Optimising margin is not the same as fair value for customers.
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 workflow owner; fix baselines |
|---|---|
| 第 3–6 周Weeks 3–6 | 盘点在用模型;选试点门店Inventory live models; pick pilot stores |
| 第 7–12 周Weeks 7–12 | 在真实货架上测 Robin 草稿Test Robin drafts on real shelves |
| 第 13 周Week 13 | 首次复盘First review |
我们可能错在:We could be wrong if: 店长不采纳草稿,或对照门店在没有草稿时也取得同样的收益。store managers ignore the drafts, or control stores show the same gain without them.
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.