How FMCG Professionals Can Use AI Well

Author | Zhao Bo

Editing | Zhang Yuwei Layout | Ge Changren Artificial intelligence has been a hot topic for a while now, and many people are discussing it, but very few have actually applied it at the enterprise level — AI still seems like just a chat tool, let alone something applied to the FMCG industry. Because Xinjingxiao works in the content industry, I personally use AI somewhat more frequently, so I’d like to share some of the experiences and feelings from my recent use of it.

First, you should install some AI tools on your phone — things like Doubao, Ernie Bot, Tongyi, Kimi, Yuanbao, and if you’re a bit more advanced and have the ability to get over the wall, you can install ChatGPT, Gemini, the integrated version of Poe, and Perplexity — so you can use them anytime. Personally, ChatGPT is still my favorite. The benefit of having it on your phone is that you can ask AI questions at any moment; it’s like an all-knowing treasury of knowledge that can greatly extend your knowledge scope and refine your ideas and concepts. Let me first share some of my feelings from using AI over this period: for the articles I’ve written in the past six months, AI has basically been deeply involved in the content of nearly all of them. My overall feeling is: AI is an extremely powerful and effective productivity tool that can help us solve a great many problems in information gathering, data analysis, content processing, and copywriting. You can think of it as an all-knowing and highly efficient work assistant. It can quickly handle work that has no technical content but is nonetheless very time-consuming — things like information lookup, data analysis, and work reports. But note that because this assistant doesn’t understand what you’re trying to do and isn’t deep enough in your field, it can only help you complete relatively junior work. My definition of AI is: a recent college graduate serving as your assistant. So, based on this positioning, the way we communicate with AI cannot follow the same expression style we’d use with a mature colleague. First, you need to know what you want, and you need to be able to describe clearly what you want. When you communicate with it, you need to use an expression style it can understand, and the key to that expression is: the Prompt. So what is a Prompt? I threw this question to GPT-4o, and here’s how it answered me: 1. What is a Prompt? • A Prompt is the instruction or cue you give to AI, telling it what you want. • You can think of a Prompt as a letter you write to AI, telling it in the letter what kind of reply you want to receive. 2. Why is the Prompt important? • AI generates its response by understanding your Prompt. If the Prompt is written clearly, AI can more easily understand your needs and thus give you an accurate answer. 3. How do you write a good Prompt? • Be clear and explicit: use simple, straightforward language to describe the question you’re asking or the content you want generated. • Be specific and detailed: the more specific the description, the better it helps AI understand your intent. For example, instead of just saying “write an article,” say “write an article about the importance of environmental protection, around 300 words, suitable for middle school students to read.” • Provide context: give the necessary background information. For instance, if you want AI to help you write an email, tell it the purpose of the email and who the recipient is. 4. Example explanation Suppose you want AI to help you write an article about environmental protection — you could write the Prompt like this: • Bad Prompt: Write an article. • Good Prompt: Write an article about the importance of environmental protection, around 300 words, suitable for middle school students to read. • Too brief: the Prompt is too short and lacks sufficient information. For example, “Help me write a story.” • Too complex: the Prompt is too complicated, and AI might not be able to understand it. For example, “Help me write a long scientific research article about the multifaceted impacts of climate change on different regions.” 5. Tip • If your first attempt at a Prompt doesn’t produce a satisfactory result, don’t be discouraged. You can revise the Prompt, add more detail, until you get a result you’re happy with. If it’s work you do regularly, you can lock the Prompt in as a fixed template, and even turn it into a model exclusively for yourself — I built one exclusively for Xinjingxiao’s writing.

I use GPT-4o, Gemini, and Claude-3 interchangeably. Honestly, each has its own strengths and weaknesses: GPT-4o is still the most powerful overall, Gemini is the friendliest for Chinese-language interaction, and Claude-3 is the strongest at interpreting data and research reports. Among domestic models, Kimi has the strongest reading comprehension, Doubao is the friendliest to use, and as for the rest — frankly speaking, when it comes to applying them to FMCG, every single domestic large model, without exception, leaves something to be desired in terms of accuracy.

To sum up, I think AI can help our work in roughly the following areas: 1. Ask questions online and let AI answer for you — especially for knowledge blind spots or information you need to look up on the fly. You can go straight to AI and ask; it’s best to use Poe, which lets you cross-check the accuracy of information. If timeliness is required, domestic large models will be faster. 2. Help you read books and articles, interpret papers and industry reports, and summarize and consolidate content and viewpoints for fast learning. Using Kimi in WeChat to read articles is particularly fast, Doubao responds the quickest, and GPT has the highest accuracy. This is especially useful for interpreting industry reports — we interpret a dozen or more industry reports every day, and GPT can quickly help summarize and extract the information within them, which is highly efficient. Sometimes you simply don’t have time to read a book but feel it’s important — you can just ask AI; if it’s newly published, converting it to PDF and having AI summarize it is also very convenient. 3. Write reports, write PowerPoint decks, process large volumes of data, and become a work assistant. For FMCG professionals, this is probably the most practical use — you feed it the materials you need to process, and it can quickly analyze, summarize, and lay out the framework and content for you. AI is also excellent at things like transcribing meeting recordings into text and organizing them into meeting documents. 4. Write Python code to automatically batch-process tedious, time-consuming, repetitive work. Note that you don’t need to know how to code for this — you just need to tell AI that you’re a coding novice who wants to write a piece of Python code but doesn’t know how, and ask it to walk you through it step by step, and it will teach you one step at a time. I myself am a coding novice, and by following GPT’s hand-holding guidance, I quickly set up a Python environment on my computer, and then, based on the requirements I gave it, had GPT write the code for me. That’s how, in under an hour, I hand-built a Tetris game on my computer. And it’s not just games — for anything you want to build, as long as it’s not too complex, GPT can basically make it happen. For example, batch file conversion and so on — it can write and run the code for you quickly and well. Personally, I think this is the single most powerful boost to efficiency. 5. Use Coze’s workflows to automate your business processes. Coze is very powerful — while it requires some coding foundation, a lot of the entry-level work, plugins, apps, and workflows have already been built by others and can be used directly. I recommend managers study it and apply it within their own companies. 6. At an even more advanced level, use open-source large models and the Ollama framework for local deployment to build your own data and knowledge base, making content retrieval and work support more convenient. I tried building a database myself — for sensitive data that isn’t convenient to feed to AI, some companies, or more senior executives, can deploy a large model on their own computer for convenient retrieval. I tried building a local large model myself using AnythingLLM, and based on the feedback, the results were pretty good.

Having covered everything I can think of in terms of how AI can help boost our personal efficiency, let me now talk about how, at the individual level, to cultivate the underlying skill of using AI well: First, stay curious — dare to imagine, dare to ask, dare to explore. A good question matters more than an answer. Whatever you think of or see, ask about it. AI really is all-knowing, and it can greatly help expand the boundaries of our information and our capabilities. Don’t be afraid of asking something wrong, and don’t treat it as something sacred — you can ask it anytime, and not just ask once, but keep following up. For example: what do you think of this matter, and what books would you recommend? Then keep following up on those specific books. A while back, during the Nongfu Spring PR incident, I had GPT combine the perspectives of four books — The Crowd, The Laws of Imitation, The Ape and the Sushi Master, and Emotions — to analyze the mechanisms behind the public’s mindless behavior. GPT wrote me several thousand words at length — there’s too much content, so I won’t post it all here. Then I had GPT draw on some viewpoints from Weak Ties, written by a professor at Xiamen University in China, to offer some valuable advice on handling the crisis PR. But I clearly noticed that GPT’s comprehension ability when reading Chinese-language books was considerably weaker than with English-language books — though it still gave some decent points. After that, I had GPT recommend ten books on crisis PR, and had it quickly summarize the three most important issues and five key areas to watch in crisis PR. Combining the books mentioned earlier (The Crowd, The Laws of Imitation, The Ape and the Sushi Master, Emotions, Weak Ties) along with the ten crisis-PR-related books it recommended, we can distill the three most important issues and five key areas of focus in corporate public-opinion crisis PR:

The three most important issues

1. Emotion management and public opinion guidance: In crisis PR, how to manage and defuse the public’s negative emotions and prevent the spread of panic and misunderstanding is one of the key issues. Correctly guiding public opinion and establishing a positive emotional feedback loop is crucial for restoring corporate reputation.

2. Control and use of information dissemination: Effectively controlling the spread of information during a crisis and preventing the spread of false or harmful information, while at the same time leveraging weak-tie network mechanisms to rapidly convey the company’s position and information, is another core issue.

3. Long-term reputation management: Crisis PR is not just about the short-term response to public opinion — more important is how to maintain and rebuild the company’s long-term reputation after the crisis, ensuring the recovery of the corporate image and brand value.

Five key areas of focus

1. Transparency and rapid response: Providing transparent, accurate, and timely information during a crisis is crucial for building public trust. A rapid response helps reduce uncertainty and speculation and stabilizes emotions.

2. Strategic use of social media and digital platforms: In the digital era, social media and other digital platforms are key battlegrounds for crisis communication and PR response. Formulating an effective social media strategy, monitoring online public opinion, and actively engaging in dialogue are focus areas that must not be neglected.

3. Leadership and internal communication: In a crisis, strong leadership and clear internal communication mechanisms can ensure consistency of information and team cohesion. The leader’s behavior and decisions will directly affect the company’s crisis-response outcome.

4. Emotional resonance and public participation: Building resonance with public emotion, and understanding and responding to their concerns, can effectively ease negative emotions. Through mechanisms for public participation and feedback, companies can better adjust their strategy and restore trust.

5. Long-term crisis management plans and reputation recovery strategy: Crisis PR should not be limited to emergency response — it should also include formulating long-term crisis management plans and reputation recovery strategies. This includes regular risk assessments, formulating preventive measures, crisis drills, and ongoing brand building and shaping a positive image. By comprehensively considering these issues and focal points, a company can respond to a public-opinion crisis more thoroughly and effectively — not only protecting and restoring brand reputation, but also strengthening the company’s competitiveness and market position through the process of crisis management. For a media professional, that basically covers the essential material needed to write a professional article on Nongfu Spring’s crisis PR handling. But I felt it still wasn’t enough, so I then had GPT lay out an outline for me. From seeing the WeChat article to accumulating the source material, it basically took about half an hour.

Title: “The Fuse: When Corporate Public-Opinion Crises Meet the Emotional Spark Point”

Outline:

### Introduction - Briefly describe the ubiquity and severity of corporate public-opinion crises - Introduce the theme: how companies manage public emotion during a public-opinion crisis, and avoid an emotional spark point igniting a bigger crisis

### The complexity of corporate public-opinion crises - Outline the different types of public-opinion crises a company may face (e.g., product failures, ethical controversies, etc.) - Discuss the speed and breadth of information dissemination in the social media era, and its effect on intensifying a crisis

### The power of emotion - Analyze the mechanisms and effects by which emotion spreads through a group, as discussed in books like Emotions - Discuss how an emotional spark point can intensify public reaction within a short period, escalating a public-opinion crisis

### From The Crowd to Weak Ties: the art of guiding public opinion - Explore the susceptibility of crowd psychology and emotion, drawing on The Crowd - Draw on the role of weak-tie networks in information dissemination as proposed in Weak Ties, and how to use this mechanism for effective communication in crisis PR

### Key strategy: emotion management and public-opinion guidance during a crisis - Transparency and rapid response: build trust, reduce uncertainty - Active use of social media: monitor public opinion, respond quickly, build positive dialogue - Strengthen leadership and internal communication: ensure consistency of information, improve team execution - Emotional resonance and public participation: understand and respond to public emotion, build empathetic connection

### Long-term perspective: reputation recovery and brand rebuilding - Learning from the crisis: assess the effectiveness of the crisis response, summarize lessons learned - Reputation recovery strategy: long-term communication plan, actively reshaping brand image - Ongoing risk management and prevention: establish effective crisis early-warning and response mechanisms

### Conclusion - Emphasize the central importance of emotion management and public-opinion guidance in public-opinion crisis PR

- Summarize how, by combining theory and practice, companies can be provided with effective strategies and solutions for handling a public-opinion crisis

This outline combines the theories and viewpoints of the books mentioned, and is intended to help you write a comprehensive and in-depth article exploring strategies for managing public emotion and guiding public opinion during a public-opinion crisis.

Of course, Xinjingxiao is not a media outlet that covers this kind of content — otherwise, a piece like this could be written in under an hour. My purpose here was to test GPT’s ability to quickly summarize and produce a draft; I had no intention of actually writing this article. Rather, I wanted to give everyone a reference for how to apply GPT to turn information into usable content.

First, you need to properly train your own ability to write prompts — this is the core factor that determines how efficiently you use AI. Work that happens on a regular basis can be locked in by designing a small custom model, but for everyday questions, the ability to ask well is the single most important factor in whether you can use AI well. The basic framework, broadly speaking, has a foundation, a role, skills, a task, and constraints. That is to say, you need to tell AI what role it’s playing, what skills it has, what task it needs to handle, and where the boundaries are. Of course, writing prompts takes practice — there’s a lot of material online about it, and there are also many standardized prompts. Coze in particular will automatically generate prompts for you, which you can use directly.

Second, you need to master some basic capabilities. Here, I think Liang Jiangjun’s summary, in his WeChat public account, of the three capabilities for mastering AI is very well put:

Access rights: whoever can access AI’s higher-level permissions, and has the right to use greater computing power, can quickly widen the gap with others — and note, this is a class gap. In the future, whoever has access rights has, in effect, human rights.
Your level of expertise:
- Capability in the relevant field
- Overall competence
  
  Cognitive models Breadth of knowledge
- Your ability to use and command AI
  
  Asking questions Judging questions
Your capacity for imagination: who can be more creative with AI.

I deeply agree with these three capabilities. I also asked GPT-3.5 what capabilities are needed to use AI well, and here’s what GPT answered:

1. Systems thinking: Systems thinking helps you understand the full picture of a problem, identify its key points and causal relationships, and ask more targeted, deeper questions.

2. Logical thinking: Logical thinking helps you analyze and organize information, form clear thinking, and ask reasonable, accurate questions.

3. Communication skills: Communication skills help you exchange and share with others, understand a problem from their perspective, and avoid misunderstandings and unnecessary disputes.

4. Innovative thinking: Innovative thinking helps you come up with new questions, find new angles and solutions to a problem, and discover new points of value in a problem.

5. Critical thinking: Critical thinking helps you examine a problem, challenge existing viewpoints and assumptions, and ask more meaningful and creative questions.

6. Language expression: Language expression helps you use clear and accurate language to express your question, so others can better understand what you’re asking.

7. Analytical ability: Analytical ability helps you break a problem down, find its individual components, and form a more specific, actionable question.

In short, asking good questions requires the combined application of multiple capability models, and requires continuous learning and practice — through reflection and feedback, continuously improving your own methods and skills. It seems the challenge for people here really is significant. Third, sometimes AI also talks nonsense — says things that are off the mark, and this is especially true of domestic AI. You must have a basic level of critical thinking and judgment in your own field of expertise, so I’d recommend cross-checking across multiple platforms — and on this front, Poe can really help you resolve a lot of issues.

On AIGC’s impact on our generation

Nvidia founder Jensen Huang said: the “iPhone moment” for the AI industry has arrived! In my view, it’s his iPhone moment — but for the vast majority of people, it more closely resembles the miserable beginning faced by farmers under Britain’s Industrial Revolution. The evolution of human civilization: primitive civilization → agricultural civilization → industrial civilization → information civilization → AI civilization — and the pace of social development and change gets faster with each stage. For a middle-aged ride-hailing driver who lost his previous job, Level-4 autonomous driving is about to take away even his opportunity to sell his time — in that situation, how is he supposed to relearn a skill to face the coming AI era? This is truly a very difficult thing, and it’s the challenge our generation must face. Industrial civilization’s machinery replaced the workshop owners of the handicraft era; AI civilization’s artificial intelligence is replacing the white-collar workers in offices. The underlying logic is: anything that is a general, standardized skill can be replaced. During Britain’s Industrial Revolution, it took two full generations to fully adapt to the new mode of production. The fate of the first generation to be replaced was, in fact, extremely miserable. It’s unclear what kind of challenges today’s white-collar workers will face next. Their occupations were forced to change, they sold off their land, and they crowded into slums. And that’s when a true proletariat truly emerged.

This time it could be even more brutal than the Industrial Revolution — a small number of people will get suddenly rich, while most people will become the truly destitute class. In the industrial era, you could still go work in a factory. In this era, we might genuinely become people of no use.

But as for the future, I’m actually still quite optimistic. From a broad logical standpoint, the individual has always been dependent on something. In the agricultural civilization era, the means of production people depended on was land; in the industrial era, people’s survival depended on the value created by companies; in the information era, we depend on internet platforms; in the AI civilization stage, people’s productive power will surely come from AI. Do you see the logic here? For the individual, as civilization progresses, the means of production one can command actually keeps growing — whether it’s information or energy, both keep growing larger and larger, and the possibility of any one individual becoming a super-individual keeps growing too. New media, e-commerce, and influencers are all super-individuals who rose up on the back of platforms — and likewise, I believe that in the AI era, even more super-individuals will rise. The core question is: how will you use AI well.

How do you make AI your assistant rather than the rival of your career? AI truly does present precise, complete information, putting the information and knowledge we want right at our fingertips. Work that used to depend on organizing knowledge, inductive analysis, and statistics can all now be replaced by AI. In other words, low-skill work has already been completely replaced by AI. This is not alarmism — it’s something that has happened in just the past year or two. Starting this year, fresh college graduates may find it harder to land a job, because their cognitive frameworks and educational experience contain no AI-related experience at all, while AI can already do the work better than they can. Our biggest challenge today is learning how to make AI the most capable assistant at our side, rather than our rival.

How do you become an assistant rather than a rival? It’s simple: hand over to AI whatever we can do and AI can also do. What we need to do is learn to command AI, the same way one learns to drive a car. We need to let go of the cognitive framework we previously held, reset to zero, and learn a new skill model to adapt to the arriving AI era. Our past ability to acquire and filter knowledge must be upgraded into the ability to discover and solve problems. Humans are responsible for accurately posing the right questions; AI is responsible for gathering, summarizing, and predicting information, then handing it to humans, so that humans can make correct judgments and decisions from the knowledge and information presented. As for how to judge — for example, the ability to gain insight from limited information — that has become exceptionally important today. AI is responsible for observation; humans are responsible for insight.

How do you build the capacity for insight? You must learn to distinguish signal from noise in the information AI gives you. You must distinguish the relationship between factual judgments and value judgments. You must have framework thinking, and must know where the boundaries of a problem lie. You must know how to ask AI questions — the efficiency of your communication with AI determines the efficiency of your work. You must have critical thinking, the ability to build a whole-picture framework around a topic, and you need to know how beliefs, positions, and emotions will influence your decisions. You must stay curious, able to keep asking why about the world. You must acknowledge your own objective ignorance, and stay open to different viewpoints. You must relearn new survival skills. Honestly, the arrival of AI is a rather cruel thing — Yuval Noah Harari said that in the future, the vast majority of people will fall into a group with no value, and only a small minority of humans will be able to transition from Homo sapiens to Homo deus. The question is: which kind of person do you want to become?