GARY AIOS / 我的 AI 工作方式

看清增长质量，决定资源投向。

财务、渠道、库存和会议信息散在不同文件里。我用 AIOS 先确认事实，再解释变化，把结论落到具体行动。

AI 准备证据、挑战假设、整理后续工作。我决定取舍，也承担结果。

Chief of Staff｜最后只回答三个问题

01 / 增长质量

增长是否带来利润？

收入、增速、利润、利润率，是否达到预期。

02 / 经营判断

哪些变化值得干预？

区分短期波动与结构性问题，抓住关键偏差。

03 / 资源取舍

下一份资源投向哪里？

明确加码、调整或收缩，以及需要我决定的事。

三大引擎，驱动经营闭环

确认事实 → 形成判断 → 推动行动

01 / TRUTH ENGINE

确认事实

核准来源、指标口径与当前周期。

经营

八渠道日报 · 售罄与库存

财务

P&L · G2N · Plan / 获批预测

战略

年度重点 · 营销与商品资源

权威来源 · 版本与情景 · 数据契约

交付：可追溯的事实与显式数据缺口

02 / JUDGMENT ENGINE

形成判断

解释变化、驱动因素、风险与假设。

核心结果

Net Sales · NS YoY · OI · OI%

领先 / 诊断

GMV / OPS · 流量 · 转化 · 客单 · 售罄 · 退货

经营影响

偏差规模 · 原因证据 · 收入与利润影响

经营背景 · 角色与技能 · 挑战假设

交付：关键结论，分清事实、推断与假设

03 / ACTION ENGINE

推动行动

明确行动、负责人、时间与升级事项。

每周纠偏

OPS / ST 看板 · 月末预测 · 行动

每月验证

P&L 回顾 · 对账 · 预测准确性

每季取舍

战略回顾 · 资源建议 · 决策备忘

固定节奏 · 执行追踪 · 人工审批

交付：行动清单与需要 Gary 决策的事项

三项经营杠杆

让判断落到资源配置

营销投入

钱投在哪

渠道与人群的投入取舍

折扣

如何让利

平衡销量、利润与价格纪律

库存 / OTB

货配在哪

平衡增长机会与资金占用

最后用四个结果验证

财务实际值以 P&L 为准；GMV / OPS 是经营信号，不是 Net Sales。

Net Sales净收入

NS YoY收入同比增长

OI经营利润

OI%经营利润率

应用案例 / 每周电商经营回顾

成交有缺口，先判断是否值得加投。

示意情境 · 非实际业绩

假设某渠道成交低于上一轮获批预测，流量下降，转化率相对稳定。我的第一反应不是直接加预算，而是确认缺口来自哪里。

01 / TRUTH ENGINE

确认事实

对齐截止日和 PF，检查日报、投放与商品数据，排除缺数或口径变化。

02 / JUDGMENT ENGINE

形成判断

流量可能是主要驱动，但还要查流量质量、客单、退货和可售库存。转化稳定本身不能证明加投有效。

03 / ACTION ENGINE

推动行动

若商品供给与增量回报支持，提出小范围调整；若超出团队权限，由我审批，并写清负责人、复盘时间和停止条件。

关键取舍：这笔新增投入，能否带来值得承担风险的收入与利润？

输出：经营摘要＋优先行动＋待决策事项。数据不足时，保留待验证假设，不把它写成结论。

执行之后，再回到结果

我按固定顺序做分析：先验证，后解释；通过检查，才讨论行动。资源调整之后，还要看结果是否支持原来的判断。

01 →监测信号

02 →验证事实

03 →诊断原因

04 →决策行动

05 ↶验证并学习

用效果修正下一轮资源配置，而不是以“报告完成”结束。

长期价值

少花时间找资料，
把时间用在判断和取舍。

每次用完，我都会把有效的方法和人工修正留下来。下一次不是从头提问，而是从已经验证过的知识继续。

少找资料

来源与版本有记录，先从正确文件开始。

少争口径

实际、计划和预测分开，冲突不被静默合并。

行动更清楚

结论落实到负责人、时间与预期影响。

经验留得住

保留有效方法，修正无效判断。

AI 看清选项，我决定取舍。

看清局势

事实发生了什么
选项可以怎么做
风险可能失去什么

提供依据

GARY

主导经营

方向做什么、不做什么
资源钱与货投向哪里
责任批准行动、承担结果

决策与责任不外包；AI 的行动以授权为边界。

↶ 我复盘结果，AI 沉淀经验。

口径、来源与适用范围

本页整合线上 AIOS 介绍与经营体系地图，展示工作方法，不代表所有专题均已上线，也不展示实际业绩。

Net Sales、OI、Expenses 与 eMarketing 财务实际值以 P&L 为准。OI% = OI / Net Sales。Plan、PF、Actual 分开比较；PF 为上一轮获批预测，-pre 不替代正式预测。任何非零 Un-reconciled Difference 均需突出显示。

营销投入是费用资源，折扣是价格杠杆，库存是可售资源与资金占用，OTB 是采购额度；四者不能简单加总。领先信号相对于最终财务结果而言，不等于已证明的因果关系。

OTB / 供应、完整营销成本与归因、历史库存仍是需要补齐的数据。预测准确性回顾须锁定结果发生前获批的预测版本及对应关账实际值；缺少渠道 P&L 时不估算渠道 Net Sales 或 OI。

当前折扣分析只实施前端 OPS 层：VIP / JD-SOP 不追踪，Total 另排除 RED；不能等同于财务折扣或利润改善。

内容依据：线上 AIOS 介绍，以及现有知识库的 Source of Truth、KPI Definitions、Knowledge Architecture 与工作流。线上原版 ↗

GARY AIOS / HOW I WORK WITH AI

Understand growth quality. Decide where resources go.

Finance, channel, inventory and meeting information often live in different files. I use AIOS to confirm the facts, explain what changed and turn conclusions into specific actions.

AI prepares evidence, challenges assumptions and organizes follow-up. I make the trade-offs and own the outcome.

Chief of Staff | Three questions that matter

01 / Growth quality

Is growth delivering profit?

Are revenue, growth, profit and margin meeting expectations?

02 / Business judgment

What needs intervention?

Separate short-term noise from structural issues and material gaps.

03 / Resource choices

Where should the next resource go?

Decide where to invest, adjust or pull back—and what needs my approval.

Three engines. One operating loop.

Confirm → Explain → Act

01 / TRUTH ENGINE

Get the facts straight

Confirm sources, definitions and the current cycle.

Operations

Eight channel reports · Sell-through · Inventory

Finance

P&L · G2N · Plan / Approved Forecast

Strategy

Annual priorities · Marketing and product resources

Source authority · Versions and scenarios · Data contracts

Output: traceable facts and explicit data gaps

02 / JUDGMENT ENGINE

Explain what changed

Identify drivers, risks and assumptions.

Outcomes

Net Sales · NS YoY · OI · OI%

Signals

GMV / OPS · Traffic · Conversion · Basket · Sell-through · Returns

Impact

Size of gap · Evidence · Revenue and profit implications

Business context · Personas and skills · Assumption checks

Output: key findings; facts, inferences and hypotheses kept separate

03 / ACTION ENGINE

Decide who does what

Set actions, owners, timing and escalation.

Weekly

OPS / ST dashboards · Month-end outlook · Actions

Monthly

P&L review · Reconciliation · Forecast accuracy

Quarterly

Strategy review · Resource proposals · Decision memo

Operating cadence · Follow-through · Human approval

Output: action tracker and decisions needed from Gary

Three business levers

Turn judgment into resource choices

Marketing

Where to invest

Allocate spending across channels and audiences

Discount

How much to concede

Balance volume, profit and pricing discipline

Inventory / OTB

Where to place stock

Balance growth opportunities and capital tied up

Validate against four outcomes

P&L governs financial Actual. GMV / OPS are operating signals, not Net Sales.

Net SalesNet revenue

NS YoYRevenue growth

OIOperating income

OI%OI / Net Sales

In practice / Weekly ecommerce review

A sales gap is not an automatic reason to spend more.

ILLUSTRATIVE · NOT ACTUAL RESULTS

Suppose a channel is below its approved previous forecast. Traffic is down while conversion is relatively stable. Before adding budget, I need to understand the gap.

01 / TRUTH ENGINE

Confirm

Align the cut-off and PF. Check daily reports, media and product data. Rule out missing records or changed definitions.

02 / JUDGMENT ENGINE

Explain

Traffic may be the main driver. But traffic quality, basket value, returns and available stock still matter. Stable conversion alone does not justify more spend.

03 / ACTION ENGINE

Act

If supply and incremental returns support it, propose a limited adjustment. I approve moves beyond team authority, with an owner, a review date and stop conditions.

The trade-off: will this investment deliver revenue and profit that justify the risk?

Output: executive readout + priority actions + decisions needed. Missing evidence stays an open hypothesis, not a conclusion.

After action, return to the outcome

I follow a consistent order: validate first, then explain. Actions come after the analysis passes review. Once resources move, I check whether results support the original judgment.

01 →Monitor

02 →Validate

03 →Diagnose

04 →Decide & act

05 ↶Validate & learn

Use the result to inform the next allocation. Finishing a report is not the finish line.

Long-term value

Less time finding files.
More time making trade-offs.

After each use, I retain the methods and human corrections that proved useful. The next task starts from validated knowledge, rather than a blank conversation.

Less searching

Registered sources and versions point to the right file.

Fewer definition debates

Actual, Plan and Forecast stay separate. Conflicts stay visible.

Clearer follow-through

Each conclusion has an owner, timing and expected impact.

Learning retained

Keep effective methods and correct weak judgments.

AI clarifies the options. I make the choices.

Clarify the landscape

FactsWhat is happening
OptionsWhat we could do
RisksWhat we could lose

Inform judgment

GARY

Lead the business

DirectionWhat to pursue—and what not to
ResourcesWhere to allocate capital and stock
AccountabilityApprove actions. Own outcomes.

Decision authority and accountability stay with me. AI acts only within authorization.

↶ I review outcomes. AI retains the learning.

Definitions, sources and scope

This page combines the public AIOS introduction and the operating map. It describes a working method, not actual results or a claim that every analytical module is live.

P&L governs Net Sales, OI, Expenses and financial eMarketing Actual. OI% = OI / Net Sales. Keep Plan, PF and Actual separate. PF is the immediately preceding approved forecast; a -pre file cannot replace a formal forecast. Highlight every non-zero Un-reconciled Difference.

Marketing is a spending resource; Discount is a pricing lever; inventory is saleable stock and tied-up capital; OTB is purchasing capacity. They cannot simply be added together. Leading signals are relative to final financial outcomes, not proven causal effects.

OTB / supply, complete marketing costs and attribution, and historical inventory remain data gaps. Forecast accuracy requires a forecast approved before the outcome and matching closed Actual. Do not estimate channel Net Sales or OI without channel P&L.

Current discount analysis covers the front-end OPS layer only. VIP / JD-SOP are excluded from tracking; RED is additionally excluded from Total. This is not equivalent to financial discount or profit improvement.

Based on the public AIOS introduction and the existing Source of Truth, KPI Definitions, Knowledge Architecture and workflows. Original online page ↗

看清增长质量，决定资源投向。

Chief of Staff｜最后只回答三个问题

增长是否带来利润？

哪些变化值得干预？

下一份资源投向哪里？

三大引擎，驱动经营闭环

01 / TRUTH ENGINE

02 / JUDGMENT ENGINE

03 / ACTION ENGINE

三项经营杠杆

营销投入

折扣

库存 / OTB

最后用四个结果验证

成交有缺口，先判断是否值得加投。

确认事实

形成判断

推动行动

执行之后，再回到结果

少花时间找资料，把时间用在判断和取舍。

少找资料

少争口径

行动更清楚

经验留得住

AI 看清选项，我决定取舍。

看清局势

主导经营

Understand growth quality. Decide where resources go.

Chief of Staff | Three questions that matter

Is growth delivering profit?

What needs intervention?

Where should the next resource go?

Three engines. One operating loop.

01 / TRUTH ENGINE

02 / JUDGMENT ENGINE

03 / ACTION ENGINE

Three business levers

Marketing

Discount

Inventory / OTB

Validate against four outcomes

A sales gap is not an automatic reason to spend more.

Confirm

Explain

Act

After action, return to the outcome

Less time finding files.More time making trade-offs.

Less searching

Fewer definition debates

Clearer follow-through

Learning retained

AI clarifies the options. I make the choices.

Clarify the landscape

Lead the business

少花时间找资料，
把时间用在判断和取舍。

Less time finding files.
More time making trade-offs.