Hello everyone, I'm Zhao Bo from New Distribution. Thank you to the brand owners, distributors, retailers, and service providers from across the country who have traveled all the way to Chengdu to attend the 11th China FMCG Conference and the 6th China FMCG Distribution and Retail Conference. Today, I want to discuss a topic with you: "How should the FMCG industry embrace change in the AI era?" FMCG is likely to be one of the first industries where AI is deployed on a large scale and delivers real results. This is not a judgment I made while sitting in my office. During the Spring Festival, I locked myself in a hotel room and spent almost every day tinkering with AI, researching AI agents, and figuring out how to make AI not just chat with you, but actually get work done. The more I experimented, the stronger my feeling became: AI is no longer just a "talking tool"; it is becoming digital employees, digital assistants, and digital systems. For the FMCG industry, this is not a small opportunity—it's a structural change. Many people still view AI as something that writes copy, makes PPTs, or generates images. But I want to say that what truly matters is not these surface-level capabilities, but the fact that AI is rewriting consumer decisions, reshaping the connection between brands and channels, and transforming the internal logic of business operations. Why I Say FMCG is the Frontline Battlefield for AI Transformation Let me start with the conclusion: FMCG is naturally suited for AI intervention. Why do I say that? Because FMCG has several very typical characteristics. First, high frequency. Consumers make many FMCG purchase decisions every week—beverages, snacks, alcohol, condiments, daily chemicals—these happen every day. Second, low average transaction value. Most purchases are not high-value, and the cost of trial and error is low. Consumers won't research for two weeks like they would for a house or a car; many decisions are made in seconds. Third, low involvement. A large portion of FMCG consumption is not deep deliberation but quick judgment. Whoever is more convenient, more accessible, and more like the "right answer" in that scenario is more likely to be bought. Fourth, strong context. FMCG is strongly correlated with time, place, weather, mood, physical state, festivals, and social relationships. Whether you stayed up late, exercised, drank, hosted guests, took care of kids, or just passed by a convenience store, these all affect your purchase choices. When these four characteristics are combined, they create exactly the battlefield where AI excels. AI's strongest capability is not creating demand out of thin air, but matching, recommending, judging, and ranking within vast amounts of information. And FMCG is precisely an industry with extremely high decision frequency, high data density, and fast feedback loops. When a consumer is recommended something today, buys today, repurchases today, or churns today, a feedback loop is immediately formed. So, once AI enters FMCG, it won't slowly penetrate; it will quickly change many things that were previously taken for granted. What AI is Changing Not Just Efficiency, But Consumer Decision-Making Power In the past, the core logic of the FMCG industry was "people finding products." Brands advertised, channels distributed, terminals displayed, salespeople pushed, and consumers compared, chose, and ordered in front of the shelf. But in the AI era, a change is happening: decision-making power is shifting from "people choosing products" to "AI choosing products for people." This change has already begun, but many people haven't truly realized it yet. Today, consumers' decision entry points are no longer just shelves, stores, or e-commerce search boxes. Voice assistants, chatbots, recommendation engines, automatic replenishment, and platform algorithms are already making more and more pre-judgments on behalf of consumers. Imagine a scenario. You drank too much last night, and this morning you feel unwell. You ask AI: What should I drink now? In the traditional logic, you think for yourself. In the AI logic, AI will directly recommend a more specific answer based on your physical state, historical preferences, and contextual needs. I gave an example at the event. For someone like me who has long been concerned about uric acid, gout, and liver health, AI's recommendation for me would definitely be different from that for a young woman. For me, soda water, turmeric-based drinks, and fresh lemon drinks are not just about taste; they tie together physical state, functional needs, and consumption decisions. This is the most fundamental change in consumption decisions in the AI era: In the past, brands faced a group of people; in the future, brands will face each individual person. So, several very obvious changes will occur in the FMCG industry next. 1. Experience is being rewritten In the past, when we did consumer insights, we focused more on demographic profiles. Male, female, 18 to 25 years old, white-collar, student, mom, late-night snack crowd, sports crowd. That won't be enough in the future. In the future, we need to face individual profiles. When it comes to drinking water, some people drink to quench thirst, some for social reasons, some for weight loss, some to relieve a hangover, some are worried about uric acid, and some just want to take a photo. AI doesn't understand you by "large groups"; it understands you by your state at this very moment. 2. Pricing is being rewritten In the past, prices were labels—9.9, 12.9, 19.9—written on the shelf, and everyone saw the same thing. In the future, prices will increasingly become variables. Different channels, different demographics, different times, and different trigger conditions will make pricing and promotional rhythms increasingly intelligent. What consumers see will no longer be a static price tag, but a dynamic result of negotiation. 3. Loyalty is being rewritten In the past, brand loyalty was often loyalty to a trademark, to advertising, or to habit. In the future, it's more likely to be loyalty to a system. Whichever system understands me better, saves me more time, and can consistently help me make choices that are "almost always right," I'll be more willing to stay in that system. So I later condensed this judgment into one sentence: In the AI era, the basic requirements of FMCG consumers for a brand are no longer just cheap and good, but three things—you understand me, help me make choices, and don't overstep. Whoever balances these three points better will more easily win the consumer's mind in the next phase. What Really Matters Is Not "Whether to Use AI" But What Problem You're Trying to Solve In recent years, many companies have been talking about digitalization and AI. But I've always had a strong feeling: Many companies don't lack tools; they lack problem awareness. Over the past decade, many companies have spent tens of millions, even hundreds of millions, on digitalization, with unsatisfactory results. Why? It's not because there aren't enough systems, reports, or tracking points, but because from the very beginning, they didn't think clearly: What problem am I trying to solve? If the problem isn't clearly defined, no amount of data will help; if the model isn't thought through, the system becomes more complex; if the scenario isn't identified, adding more skills only creates more chaos. The same goes for AI. It's not about having more skills, more expensive models, or bigger systems. The most important thing is to start from the problem. For example, a distributor: Is your most critical problem slow inventory turnover, weak store sell-through, low salesperson efficiency, high receivables risk, or substandard terminal displays? You need to first know where your pain points are.

  • First define the problem, then define the model.
  • First set the goal, then match the data.
  • First close the loop on one scenario, then replicate to more scenarios. This order cannot be reversed. If reversed, you'll end up with a bunch of beautiful nonsense. Why I Always Say Tacit Understanding Is the Most Valuable Asset Through my recent tinkering with AI, I've had a particularly deep feeling: AI is powerful, but AI has no tacit understanding. It knows a lot of knowledge, even most public information in the world, but it doesn't know hot or cold, doesn't know pain, doesn't sense the shift in atmosphere in social settings, and doesn't know whether a boss saying "let's take another look" means hesitation, perfunctoriness, or lack of money. These things, humans have them; AI doesn't. So I always say, what's truly valuable isn't just data, but the tacit understanding you form on the front lines. You've seen customers frown. You know why frontline salespeople can't push through; you know why small shop owners say they won't stock up but are actually afraid they can't sell; you know that a brand may look like it's selling well, but the terminal hasn't truly formed consumer mindshare. This tacit knowledge is often hard to articulate, but you can feel it. And it's precisely these things that determine whether AI has real value. So for experienced people in the industry, the biggest opportunity isn't competing with AI on computing power, but quickly converting your scenario perception, problem awareness, and business understanding into capabilities that AI can call upon. You identify a real problem, then immediately use AI to solve it—this process itself is product development. Don't try to build a massive system from the start. First make a power drill, first make a small skill, first solve a high-frequency problem. That's enough. How Companies Can Implement AI I Believe It Should Be Done in Three Steps FMCG companies can implement AI in three stages: the crutch stage, the organ stage, and the system stage. 1. Crutch Stage First, treat AI as a tool. Write copy, make PPTs, summarize reports, categorize public opinion, diagnose stores, provide basic product selection suggestions, and issue inventory warnings. In this stage, don't deify AI, and don't restructure the company right away. The goal is simple: make people willing to use it, make suggestions reviewable, and make benefits quantifiable. 2. Organ Stage Next, turn AI from an external tool into an "organ" within your business process. For example, automatic replenishment suggestions, out-of-stock handling suggestions, salesperson route recommendations, promotional spending suggestions, abnormal store warnings, and reconciliation assistance. At this point, AI is no longer just giving you a report; it's entering your daily workflow. 3. System Stage Further along, it's the system stage. AI begins to run through five levels—consumer, point of sale, city, supply chain, and headquarters—forming a true operational loop.

At the consumer level, do scenario recognition and demand judgment.

At the point-of-sale level, do store profiling, SKU optimization, and replenishment suggestions.

At the city level, do channel scheduling, sell-through prediction, and cost efficiency.

At the supply chain level, do demand forecasting, inventory allocation, and production planning coordination.

At the headquarters level, do panoramic insights, resource allocation, and risk control. In this stage, AI is no longer just saving you time; it's starting to influence the entire company's operating model. There Are Principles for AI Implementation; Not Everything Can Be Automated I don't advocate for companies to immediately connect AI to core transaction chains and execute automatically. That's too risky. The approach I prefer is to first assess, then implement gradually. Whether something can be done by AI depends on at least three factors: First, the degree of automation potential. Can the rules be clearly stated, and can the boundaries be clearly defined? Second, the degree of fault tolerance. How costly is a mistake? Third, the degree of variability. If rules change every day—today this way, tomorrow that way—then hard automation isn't suitable. Additionally, I think any company using AI must clarify four things: Task contract, acceptance criteria, constraints, and feedback loop. What you want it to do and what you don't want it to do. What counts as doing it right? What areas are off-limits? How do you correct it when it goes off track? If these aren't clear, AI will definitely "hallucinate" and go off track. It's not that it's deliberately trying to trick you; it's that you haven't defined the boundaries clearly. AI's Impact on Distributors and Salespeople Is Not Minor Tinkering, But a Rewriting of Roles I think many people haven't truly paid attention to this yet. First, distributors In the past, a large portion of distributors' profits came from price differences, information asymmetry, and local relationship advantages. But once AI enters the industry on a large scale, channel transparency will increase. Where products are sold, how much is sold, who sells fast, who has high inventory, who executes well—many things will become increasingly visible. This means that the real value of distributors in the future will no longer mainly be "I know more than you," but rather: I have stronger fulfillment.

I have stronger terminal services.

I can do regional operations.

I can do data collaboration.

I can help brands with more detailed localized execution. So distributors will likely differentiate into several roles in the future: Regional fulfillment service providers, terminal operations service providers, and data collaboration service providers. Whoever can upgrade in these directions will have more room to grow. Second, salespeople Salespeople will be affected more directly. In the past, many salespeople relied on running around, drinking, relationships, and experience. In the future, these things won't completely disappear, but they won't be enough. Future salespeople will be more like people who execute and fine-tune strategies within an algorithmic framework. Routes: recommended by the system.

Orders: AI gives suggestions.

Displays: real-time recognition via phone photos.

Terminals: AI does store profiling.

Customer management: AI gives strategic suggestions. So a good salesperson in the future won't just be someone who can run around, but someone who can do three things:

  • Read data Be able to understand the store profiles and suggestions the system gives you.
  • Use tools Be proficient in SFA, BI, photo recognition, and data assistants.
  • Translate Translate data and suggestions into language that bosses, stores, and customers can all understand. In short, the core competitiveness of future salespeople will no longer be purely physical effort, but the synergy of "people + tools." In Conclusion: AI Is Not Replacing FMCG People But Redefining Value Creation I've never believed that AI will wipe out everyone in the industry overnight. But I'm very certain that AI will quickly compress repetitive work without value, low-quality communication, and low-level decision-making. It will force us to re-answer several questions:

What do we understand about consumers better than others?

What products can we make better than others?

In what scenarios can we provide better solutions than others?

Are we making money through relationships or through capability? So, for brand owners, distributors, and retailers alike, the most critical things going forward are, I believe, very clear: First, build a solid foundation for products and data infrastructure.

Second, enter AI decision entry points as soon as possible, rather than sticking to traditional shelf thinking.

Third, rewrite marketing and pricing systems, moving from "flood irrigation" to individualized matching.

Fourth, drive organizational upgrades so AI truly enters the daily operational cycle. This change has no end point. It's not a one-off deal, nor does it end with a single project. It's more like an infinite game. It's not about who shouts the loudest today, but who starts earlier, iterates faster, and can continuously turn real problems into real results. So I'll end with the same sentence: The key isn't "whether to do AI," but "how to start the first step as soon as possible." While embracing AI, don't lose the most fundamental and simple things. Product strength is the foundation, supply chain is the hard currency, and long-term trust is the most expensive asset. AI won't create these things for you. But whoever learns to use AI to amplify these things first will have a better chance of winning the next phase. PS: For those interested in the content of the on-site speeches, please follow the recent posts on the "New Distribution" WeChat official account. We will compile and publish all speakers' speeches for our readers. Click Read Original to see more about the 11th China FMCG Conference and the 6th FMCG Distribution and Retail Conference...