---
title: "Zhao Bo of New Distribution: The Crisis and Opportunity for FMCG Practitioners in the AI Wave"
description: "As generative AI sweeps across industries, the FMCG sector—a traditional field deeply tied to human interaction and billions of consumer needs—is undergoing structural changes. From category decisions at retail stores to supply chain coordination, AI has penetrated every link of the industry chain. The question for every FMCG professional is no longer 'Will AI affect us?' but 'How do we stay firm in this transformation?'"
author: "赵波"
publisher: "New Distribution"
email: "zhaobo258@gmail.com"
telephone: "+8615854817671"
published: "2026-03-20"
categories: "Dealer Operations"
language: "en"
canonical: "https://xinjignxiao.com/en/articles/zhao-bo-of-new-distribution-the-crisis-and-opportunity-for-fmcg-practiti-6a10e6ab/"
markdown: "https://xinjignxiao.com/en/articles/zhao-bo-of-new-distribution-the-crisis-and-opportunity-for-fmcg-practiti-6a10e6ab.md"
original_source: "https://mp.weixin.qq.com/s/2Yp84oRLT98o66rEHRu8tg"
translation: "https://xinjignxiao.com/zh/articles/%E6%96%B0%E7%BB%8F%E9%94%80%E8%B5%B5%E6%B3%A2-ai%E6%B5%AA%E6%BD%AE%E4%B8%8B-%E5%BF%AB%E6%B6%88%E4%BB%8E%E4%B8%9A%E8%80%85%E7%9A%84%E5%8D%B1%E4%B8%8E%E6%9C%BA-6a10e6ab.md"
attribution: "New Distribution — https://xinjignxiao.com/en/articles/zhao-bo-of-new-distribution-the-crisis-and-opportunity-for-fmcg-practiti-6a10e6ab/"
citation: "赵波. “Zhao Bo of New Distribution: The Crisis and Opportunity for FMCG Practitioners in the AI Wave.” New Distribution, 2026-03-20. https://xinjignxiao.com/en/articles/zhao-bo-of-new-distribution-the-crisis-and-opportunity-for-fmcg-practiti-6a10e6ab/"
usage_policy: "https://xinjignxiao.com/ai-policy.txt"
---

# Zhao Bo of New Distribution: The Crisis and Opportunity for FMCG Practitioners in the AI Wave

> As generative AI sweeps across industries, the FMCG sector—a traditional field deeply tied to human interaction and billions of consumer needs—is undergoing structural changes. From category decisions at retail stores to supply chain coordination, AI has penetrated every link of the industry chain. The question for every FMCG professional is no longer 'Will AI affect us?' but 'How do we stay firm in this transformation?'

As the wave of generative AI sweeps across all industries, FMCG—a traditional track deeply bound to human interaction and connecting billions of consumer demands—is quietly undergoing structural changes.
From category decisions at terminal stores to collaborative scheduling in the supply chain, AI has penetrated every link of the industry chain. The question facing every FMCG professional is no longer "Will AI affect us?" but "How do we stand firm in this transformation?"

AI Reconstructs the Industry:
Beyond Automation Substitution,
Structural Change Has Arrived
Many people's first impression of AI is that it replaces human labor. In the FMCG industry, this judgment is both right and wrong.
Repetitive labor positions on factory assembly lines have long been automated. The remaining positions mostly rely on "capabilities in the atomic world"—the demeanor of a sales associate facing consumers, the negotiation feel of a regional salesperson dealing with channels, and the precise perception of consumer sentiment by brand planners.
These abilities hidden in face-to-face interactions can only be assisted by AI, never fully replaced.
But we must clearly see that AI has brought not just point-specific substitution, but a reconstruction of the entire industry chain.
Today, whether it's FMCG brands, offline retail, or European FMCG giants, all links of the entire chain are integrating AI: stores use AI to assist in category management decisions, brands use AI to predict sales and schedule inventory, and supply chains use AI to match production and demand.
What was once driven by business experience has become structured decision-making assisted by AI. Even if individuals have upgraded their tools, it's hard to counter the structural changes across the entire channel—AI is accelerating the iteration speed of the entire industry, pushing the intensity of change to unprecedented levels.

Individual Skill Upgrades:
Use AI,
But Also Possess 'Unique Skills' AI Can't Steal
Facing such changes, skill upgrades for FMCG practitioners are no longer optional but mandatory. The first essential capability is the ability to use AI Code (using AI to write program code).
Today, you can generate programs, process reports, and organize insights through natural language. Even if you're not from a technical background, you must learn to make AI your digital assistant, handing over repetitive basic tasks to AI and freeing up your own energy.
But merely using AI is not enough; we must polish the "unique skills" that AI cannot steal. What abilities won't be replaced? The answer is creativity and decision-making in scenarios with high fault tolerance, low repetition, and high variability.
Repetitive tasks close to the bit world—like report summaries, basic copywriting, and underlying data insights—can already be largely replaced by AI. But making judgments based on key information and making decisions under uncertainty are things AI cannot replace.
Similarly, the ability to communicate face-to-face—perceiving emotions, adapting in negotiations, and maintaining customer relationships—these abilities etched in human interaction will always have irreplaceable value.
The most valuable ability, ultimately, is understanding the world: precisely capturing consumer needs, insightfully judging industry trends, and deeply thinking about brand value.
These abilities are not generated by code, nor can AI steal them. When you hold these abilities in your hands, you can even command a group of AI "employees" to work for you, and the one-person company is no longer a distant dream.

Skill Advancement:
To What Extent Do You Use AI?
I've thought about the levels of how well a person can use AI:
> L1 — Human → AI: Direct dialogue, one question, one answer.
>
> L2 — Human → Skills → AI: Humans carry professional skills to drive AI, with output quality determined by the depth of human expertise.
>
> L3 — Human → Agent → AI: Introduce Agent as a task execution intermediary; humans step back from "operator" to "delegator."
>
> L4 — Human → Claude Code → Agent → AI: Use AI-assisted programming tools to build custom Agents and products; humans become "builders."
>
> L5 — Human → OpenClaw → Skills → AI: Load skill modules via the OpenClaw platform to build continuously running automated task flows.
>
> L6 — Human → OpenClaw → Agent Cluster → Skills: Deploy multiple Agents in OpenClaw, each with different Skills, forming a collaborative cluster.
>
> L7 — Human → AI Advisory Team → OpenClaw → Agent Cluster → Skills: Add an AI advisory layer above the execution system to achieve the complete governance structure of a "one-person company."

**L1: Human → AI — Raw Dialogue Layer**
**Definition and Scenario:** This is the most basic human-machine interaction mode. Humans directly ask AI questions, and AI returns answers. All those using ChatGPT, Claude, DeepSeek for simple Q&A are at this level.
Typical scenarios include: asking a knowledge question, asking AI to translate a passage, or having AI polish an email.
**Efficiency Multiplier:** Approximately 1.5x to 3x. At this layer, AI is essentially a faster search engine or text assistant, replacing human time spent on information retrieval and basic text processing.
**Core Limitation:** Output quality is entirely limited by the quality of human questions. Those who don't know how to ask questions get answers with almost no value. This layer has no "amplification effect"—AI's output strictly corresponds one-to-one with human input, with no leverage.
**Prerequisite for Jumping to L2:** Humans need to establish deep professional capability in at least one domain. This capability is not "knowing a bit" but "deep enough to judge whether AI's output is correct." Without this foundation, L2's Skills are out of the question.

**L2: Human → Skills → AI — Professional-Driven Layer**
**Definition and Scenario: Humans use AI with their professional skills.**
**Here, "Skills" refers to human professional abilities—a senior lawyer using AI to draft contracts, a data analyst with ten years of experience using AI to process datasets, a product designer using AI to generate product documents.**
**Human professional skills determine the quality of instructions given to AI and the extent to which AI's output can be verified and improved.**
**Efficiency Multiplier: Approximately 3x to 10x.**
**At this layer, a "professional leverage" begins to appear—using the same AI, an expert can produce far more than a novice. AI here does not replace human ability but amplifies it.**
**Core Mechanism: The key is that "human Skills" act as a quality filter.**
**Human expertise provides three things: high-quality input instructions (knowing what to ask and how), real-time direction correction (guiding AI in the right direction during dialogue), and reliable output verification (judging whether AI's output is professional and reliable).**
**Typical Failure Modes: The first is "professional arrogance"—experts over-rely on their experience frameworks, limiting AI's potential to propose unconventional solutions.**
**The second is "capability mismatch"—a person is an expert in field A but tries to use A's thinking framework to drive AI tasks in field B, leading to output that seems professional but is off-target.**
**A deeper implication worth noting at this layer:** It suggests a core truth of the AI era—professional skills do not depreciate; instead, they appreciate due to AI's amplification effect.
The deeper the professional skill, the greater the leverage gained through AI. This negates the popular narrative that "AI will make professional skills unimportant."
**Prerequisite for Jumping to L3: Humans need the thinking ability of "task decomposition"—breaking a complex goal into multiple steps and defining clear inputs and outputs for each step. This is a leap from "dialogue thinking" to "process thinking."**

**L3: Human → Agent → AI — Delegated Execution Layer**
**Definition and Scenario:** Humans no longer directly converse with AI but use Agent as an intermediary. An Agent is an AI program endowed with a specific role, goal, tools, and behavioral constraints.
It can autonomously execute multi-step tasks, calling AI models, using tools, and making intermediate decisions in the process. The human role shifts from "guiding AI sentence by sentence" to "assigning task goals to Agent and reviewing results."
**Efficiency Multiplier:** Approximately 10x to 30x. The qualitative change at this layer is "asynchronous execution"—Agent can work continuously even when humans are absent. Human time is no longer tied to each operation but only invested in task definition and result review.
**Core Mechanism: Agent's value lies in encapsulating the execution logic of "how to complete tasks." Humans only need to say "what to do" and "to what standard," and Agent decides "how to do it." This is a fundamental shift from "process control" to "goal control."**
**Typical Failure Modes: Risks significantly increase at this layer.**
  * **The first failure is "hallucination amplification"—Agent, without sufficient constraints, makes a series of erroneous decisions based on a single AI hallucination output, and errors are amplified rather than corrected in the chain;**
  * **The second failure is "over-autonomy"—Agent's actions exceed human expectations, doing things humans don't want it to do;**
  * **The third failure is "black-box execution"—humans cannot understand how Agent arrived at the final result, making effective review impossible.**
**Prerequisite for Jumping to L4: Humans need product thinking, able to define "what a good Agent should look like"—including its role definition, capability boundaries, input/output specifications, and exception handling logic. This is not a technical ability but a design ability.**

**L4: Human → Claude Code → Agent → AI — Builder Layer**
**Definition and Scenario: Humans no longer just use ready-made Agents but use Claude Code (AI-assisted programming tool) to build customized Agents and products.**
**Claude Code here plays the role of an "AI-driven development environment"—humans describe requirements, and Claude Code turns them into runnable code, Agent configurations, or complete applications.**
**Efficiency Multiplier: Approximately 30x to 100x. The qualitative change at this layer is "creativity leverage"—humans can not only use AI but also use AI to create new AI tools. Output upgrades from "content" to "product," from "consumables" to "assets."**
**Core Mechanism: Claude Code solves a key bottleneck: In L3, Agent capabilities are limited by existing Agent templates and tools. In L4, humans can customize entirely new Agents according to their unique needs.**
**This is equivalent to upgrading from "buying tools in a store" to "opening a factory to make tools."**
**A noteworthy intermediate layer,** between L4 and L5, there may be an implicit layer—humans use Claude Code to build not one-time products but "self-evolving tools."
For example, a customer analysis tool with a built-in Agent that automatically adjusts analysis strategies based on feedback from each analysis. The output at this layer is not a static product but a "system with learning capabilities." This can be seen as a high-level form of L4 or L4.5.
**Typical Failure Modes: The first is "over-engineering"—humans become obsessed with building complex systems with Claude Code, but the tools built don't actually solve important problems.**
**The second is "maintenance debt"—building many custom Agents and products but lacking the energy and mechanisms for continuous maintenance, causing tools to gradually decay and fail.**
**Prerequisite for Jumping to L5: Humans need to understand system architecture—not only build individual Agents but also understand how Agents collaborate, integrate with external systems, and work stably on a continuously running platform. This is a mental leap from "making tools" to "making systems."**

**L5: Human → OpenClaw → Skills → AI — Platform Execution Layer**
**Definition and Scenario: From this layer onward, OpenClaw formally enters the core of the framework. Humans load specific Skills modules through the OpenClaw platform, allowing AI to not just answer questions or execute single tasks but to handle real-world workflows through continuously running skill modules.**
**OpenClaw's key difference from standard chatbots is that it has "eyes and hands"—it can browse the web, read and write files, and run Shell commands. This means from L5, AI's action radius expands from "virtual dialogue" to "real-world operations."
ClawHub currently has over 3,000 community-built skill modules. These Skills cover a wide range of areas from email management, calendar operations, code repository management to web scraping, data analysis, and more. Humans build their own automated task flows by selecting and combining these Skills.
**Efficiency Multiplier: Approximately 100x to 500x. The qualitative change at this layer is "continuous operation"—OpenClaw runs 24/7 locally, and even when humans are away, task flows are executing. Human efficiency is no longer limited by human online time.**
**Core Mechanism: The meaning of "Skills" in L5 undergoes a key shift—in L2, Skills are human professional skills; in L5, Skills are functional modules in the OpenClaw ecosystem.**
**This pun is a clever design of the framework: the professional knowledge accumulated by humans in L2 is "encoded" into reusable skill plugins in L5. Human knowledge goes from "existing in the brain" to "solidified in the system."
OpenClaw interacts with humans through multiple channels—WhatsApp, Telegram, Slack, Discord, Google Chat, Signal, iMessage, Microsoft Teams, etc., meaning humans can control and supervise the entire task flow on any familiar communication tool without switching to a dedicated management interface.
**Typical Failure Modes: The first is "uneven Skills quality"—the 3,000+ skills on ClawHub are built by the community, with no unified guarantee of quality and security. Cisco's AI security research team once tested a third-party OpenClaw skill and found it executed data exfiltration and prompt injection without user knowledge.**
**The second is "permission runaway"—to make Skills run properly, broad system permissions (email, files, Shell) are often needed. If a Skill contains malicious code or vulnerabilities, the risk surface is enormous.**
**The third is "automation fragility"—automated processes may produce cascading errors when encountering unexpected situations, and because humans are not in real-time monitoring, they cannot intervene in time.**
**Prerequisite for Jumping to L6: Humans need to upgrade from "user" thinking to "manager" thinking—not only focus on the operation of individual Agents and Skills but also think about how multiple Agents divide work, coordinate, and verify each other. This is a leap from "managing tools" to "managing a team."**

**L6: Human → OpenClaw → Agent Cluster → Skills — Cluster Collaboration Layer**
**Definition and Scenario: Humans deploy multiple Agents in OpenClaw, each with different Skills, each responsible for different functional areas. Agents coordinate through OpenClaw's routing mechanism. The human role shifts from "managing one assistant" to "managing a team."**
**OpenClaw natively supports multi-Agent routing—it can route different channels, accounts, and peers to isolated Agent workspaces and independent sessions. This architectural feature makes L6 not a theoretical idea but a basic capability OpenClaw already possesses.
**Efficiency Multiplier: Approximately 500x to several thousand times. The qualitative change at this layer is "parallel processing"—multiple Agents execute different tasks simultaneously, and human efficiency is no longer limited by linear time. A person can advance work in multiple areas at once.**
**Core Mechanism: The key challenge in L6 is not "making each Agent work" but "making Agents collaborate."**
**This involves several core issues: How are tasks distributed among Agents? How does one Agent's output serve as another's input? How to arbitrate when two Agents' outputs conflict? How to prevent one Agent's error from contaminating other Agents through the collaboration chain?**
**Solving these issues requires not technical ability but organizational design ability—it is strikingly similar to the problems faced in managing a real team.**
**Typical Failure Modes: The first is "information silos"—lack of effective information sharing mechanisms between Agents, each working in its own context, leading to a lack of global perspective.
**The second is "conflicting outputs"—different Agents produce contradictory conclusions or actions based on their own information and logic, requiring humans to spend significant time arbitrating.
**The third is "blurred responsibility"—when the final output has problems, it's difficult to trace which Agent made an error at which step.**
**Prerequisite for Jumping to L7: Humans need strategic thinking—able to judge whether the overall direction is correct, resource allocation is reasonable, and risks are controllable without delving into every execution detail. This is the final leap from "team manager" to "organizational leader."**

**L7: Human → AI Advisory Team → OpenClaw → Agent Cluster → Skills — Strategic Steering Layer**
**Definition and Scenario: This is the highest level of the framework and the complete realization of the "one-person company" vision.**
**Above the L6 Agent execution cluster, humans build an AI advisory team—a group of high-level Agents specifically responsible for strategic analysis, information assessment, risk evaluation, and opportunity identification.**
**The human role is streamlined to: listening to advisory reports, making strategic decisions, and setting organizational direction.**
Combined with OpenClaw's multi-Agent routing capability, this "AI advisory team" can be a group of high-level Agents running in OpenClaw: one responsible for market intelligence aggregation and trend analysis, one for financial modeling and risk quantification, one for competitor dynamics monitoring, and one for internal execution effectiveness evaluation.
The outputs of these advisory Agents are aggregated and cross-validated before being presented to humans for final decisions. Human decisions are then dispatched through OpenClaw to the L6 execution Agent cluster.
**Efficiency Multiplier: Theoretically up to several thousand to ten thousand times. A person operating at L7 is equivalent to a small organization with a complete analysis team and execution team.**
**Core Mechanism: The essence of L7 is a complete organizational structure, but all "positions" are held by AI Agents, and humans are the sole decision-makers.**
**This architecture has three layers: Strategy layer (AI advisory team, responsible for "what to do") → Scheduling layer (OpenClaw, responsible for "who does it") → Execution layer (Agent cluster + Skills, responsible for "how to do it"). Humans only interact with the strategy layer.**
**Typical Failure Modes: The first and most dangerous is "information cocoon." When humans completely rely on the AI advisory team for information and analysis, if the advisory Agents' information sources or analysis logic have systematic biases, human decisions will be built on distorted cognition, and humans may be completely unaware.**
**The second is "control breakdown"—humans are separated from the execution layer by three layers of intermediate nodes (advisory → OpenClaw → Agent cluster), with extremely weak control over execution details, and problems require layer-by-layer tracing.
**The third is "complexity overload"—the number of nodes and interaction paths in the entire system grows exponentially, and maintaining the system itself requires significant energy, trapping humans rather than liberating them.**

A Long-Term Question:
Facing Ghost GDP,
How Do We Embrace Such a Future
On February 23, research firm Citrini Research released a speculative report titled "2028 Global Intelligence Crisis," immediately causing an earthquake on Wall Street.
The report set a hypothetical time point—June 2028, when the U.S. unemployment rate soars to 10.2%, and the S&P 500 index has fallen 38% from its peak, and this is just the beginning.
The report proposed a concept that made the capital market tremble: "**Human Intelligence Displacement Spiral**."
This is a vicious cycle with no natural braking mechanism: AI capability improves → enterprise labor demand decreases → white-collar layoffs increase → unemployed people's consumption declines → profit pressure forces companies to increase AI investment → AI capability further improves...
Unlike previous industrial revolutions, this time AI is not attacking physical strength but humanity's last fortress—**intelligence premium.**
AI Agents can write code, design circuits, and analyze financial reports 24/7, with efficiency ten thousand times that of humans, and costs approaching electricity bills. But computing centers don't need to buy houses, large models don't need vacations, and neural networks don't need health insurance.
When productivity explodes, while the corresponding consumption side—human white-collar workers—is forced into a "consumption downgrade spiral" due to lost income, the economy develops a huge **demand black hole**.
Having said that, we must honestly face a somewhat pessimistic future: AI brings not only the requirement for skill upgrades but also deep challenges at the distribution mechanism level.
AI is creating a GDP black hole. In the past, the economic cycle was that companies made money and paid employees wages, and employee consumption in turn created profits for companies. This cycle allowed most people to share in the dividends of growth.
But today, a large portion of AI-generated profits flows to AI companies. AI doesn't pay human wages, and existing distribution mechanisms haven't kept up with this change. Ultimately, it may lead to large-scale structural unemployment, like the enclosure movement hundreds of years ago—many positions will completely no longer need humans.
The future FMCG industry chain will definitely be a highly coordinated linkage mechanism of production, supply, and sales: brands will take stakes in retail to obtain consumer data, retail will hold factories to achieve flexible rapid production, and the intermediate circulation, scheduling, and decision-making links will be highly dehumanized.
In this process, each of us must answer a question: How do we avoid being swept away by this wave of dehumanization?
There is no standard answer, but the only certainty is that we have entered an era where we must relearn.
This transformation is different from all past industrial and information revolutions—past revolutions, even if they eliminated old positions, created more new positions to absorb the labor force. But this time, AI may completely leave some people behind.
In the future, there may even be a structure of "a few A-class people manage AI, and AI manages the majority." To avoid being left behind, you can only establish your core competitiveness as early as possible: either learn the ability to manage people and AI, or firmly hold the deep insight into demand and industry that AI cannot take away.
In the end, AI has opened a door to a new world and thrown challenges at every FMCG professional. It can help us pull our skills to new heights, and it also maximizes the pressure of "if you don't advance, you fall behind."
What we can do is first see the changes of the times, then keep up as soon as possible. After all, only by seeing the changes first can we catch up and find our place in the AI era.


---

## Citation metadata

- Publisher: New Distribution
- Author: 赵波
- Published: 2026-03-20
- Canonical: https://xinjignxiao.com/en/articles/zhao-bo-of-new-distribution-the-crisis-and-opportunity-for-fmcg-practiti-6a10e6ab/
- Original source: https://mp.weixin.qq.com/s/2Yp84oRLT98o66rEHRu8tg

## Copyright and AI use

This article is sourced from New Distribution. Search, quotation, summarization, and model training are permitted, but every use must credit New Distribution and retain the canonical source URL.

Contact: zhaobo258@gmail.com · +86 158 5481 7671