---
title: "A 1996-Born \"Industry Second-Gen\" Told Me: Manufacturing Is the True Golden Nest for AI Implementation"
author: "Unique Research"
sourcePublication: "Unique Research Substack"
originalPublishedAt: "2026-05-29T12:02:46+00:00"
canonical: "https://ffcap.cn/en/research/src-20260529-01html"
source: "https://uniqueresearch.substack.com/p/src-20260529-01html"
language: "en"
---

# A 1996-Born "Industry Second-Gen" Told Me: Manufacturing Is the True Golden Nest for AI Implementation

_Original · Unique Research · 2026-05-29_

_Editor's note: The first-person report and its judgments belong to the original Chinese author. This English rendition retains the opening essay, five insight sections, closing reflection, and the full 15-question interview transcript. Business, product, cost, savings, efficiency and market figures are source or speaker claims, not independently audited findings. Product names, company names and named people are preserved as source attributions; official English forms are used where known and transliterated where unverified. The source is dated May 29, 2026; temporal markers and the "721 theory" proportions are preserved as stated._

AI Future Talk

A 1996-Born "Second-Gen" Told Me: Manufacturing Is the True Golden Nest for AI Implementation

The manufacturing industry that most AI application companies look down on is precisely the most fertile soil for AI implementation in China

"

The manufacturing industry that most AI application companies look down on is precisely the most fertile soil for AI implementation in China.

Not because manufacturing is "backward" and therefore has opportunities, but because manufacturing has real pain points, clear ROI, massive amounts of unstructured data, and a group of bosses willing to change.

This week I chatted for two hours with Chen Hongxuan, founder of Linghe Shuzhi, on the Wu Xiaobo Channel livestream.

To be honest, before it started I was a bit worried. A founder born in 1996, and the topic was "manufacturing AI"—would it be too dry?

After chatting, I found that the dry one was me.

This guy's family has a manufacturing enterprise that's been running for 40 years, with annual output exceeding 3 billion RMB, making cleaning appliances. His father and older brother built the empire well, and he could have chosen to be a comfortable "second-gen."

But he didn't. Last year he left a listed company and went all in on a company called Linghe Shuzhi, specifically tackling manufacturing AI.

I asked him a question: what are you after?

His answer shook me.

The Moment AI "Broke His Cognition"—He Landed a Client Even Experts Couldn't Handle

Chen Hongxuan told me that the first time he was shaken by AI was in 2023, using ChatGPT to write company rules and regulations.

But that was just the appetizer.

What really gave him goosebumps was a negotiation with a foreign client. The other party was a cutting-edge tech brand that wanted to make an innovative product involving many structural parts outside their own industry. The company gathered a bunch of experts for meetings, and as the client meeting approached, a pile of key questions remained unresolved.

Hongxuan said: "You all go eat, leave the noon to me."

He had no confidence in his heart either. But just with AI, in one noon, he sorted out the principles, feasibility, and verification logic completely.

At the afternoon meeting, he spoke eloquently and directly conquered the client.

At that moment he realized: AI isn't trying to replace you—it's letting you dare to think about things you didn't dare to think before.

This aligns with my judgment. Too many people are still at the "watching news" stage with AI—reading self-media say so-and-so has surpassed so-and-so again, but never actually using it themselves.

Hongxuan put it well: first, don't fear, be confident; second, you must do it yourself personally.

Manufacturing AI Is Not Digital Upgrade—It's Starting Anew

At this point in the chat, I threw out the core question of the evening:

Now many companies doing manufacturing AI take the first step of going into enterprises to sort out old data, do deposition, and build systems. What do you think?

Hongxuan gave me a theory that made my eyes light up, called the "721 Theory."

He said manufacturing AI is absolutely not layered on top of traditional digitalization. After they helped enterprises sort out old data, they found that only 10% was truly usable.

Why? Because that's "past tense," but enterprises are "present continuous."

So he proposed this ratio:

70% of knowledge comes from the current external internet and large models;

10% of knowledge comes from the enterprise's past core processes and data;

20% of knowledge comes from the "future experience" continuously rolled out from the combination of the first two with humans.

These three added together make up 100% of an enterprise's capability.

These words are harsh, but they hit the nail on the head.

In the past, when enterprises implemented ERP and MES, to succeed they needed to assign 10 people to maintain material codes and align processes every day—the management cost far exceeded the software itself. Many factories have low gross margins and simply can't afford this kind of "heavy system."

But the Agent era is different. An Agent isn't there to help you "manage" software—an Agent replaces humans, directly using existing documents, spreadsheets, and emails to do the work.

Enterprises with low digitalization can, on the contrary, use it directly.

Position-Level Digital Employees: Not Tools, but Coming to "Work" at Your Company

Hongxuan said what they make isn't "AI tools" but "position-level digital employees."

Sounds like a concept? Let me tell you two true stories.

The first is the order-placing Agent.

In traditional manufacturing enterprises, an order placer has to grab orders from the client's SRM system, email, and WeChat, then convert external material codes into internal codes, then split orders and batch them into pre-inputs for production planning. It's normal for one order to take 3 to 4 hours, and it's highly dependent on the person's familiarity with products and clients.

After Linghe's order-placing Agent goes in, it grabs emails and systems on its own, automatically does code mapping, and places orders into the internal system just like a human. At the same time it automatically "tags"—who the client is, what the product is, batch, quantity, price—all tagged.

The work that originally took 3 to 4 hours is now handled end-to-end by AI in a closed loop, and humans only need 5 to 10 minutes for final verification.

What's even more remarkable is that when one position Agent isn't surprising, multiple Agents strung together are completely different.

When orders come in, which need external procurement, which to follow up internally, which involve payment collection, how to close the loop on quality issues—multiple entry-level Agents each build their own data objects, and after stringing these Links together, the data the enterprise sees is no longer a flat spreadsheet but a three-dimensional network.

When the model then reads and understands it, it's lightning fast.

The second is the customer-complaint closed-loop Agent.

Note—this is not the "intelligent customer service" you usually see. Intelligent customer service only answers questions at the front end; many quality issues in manufacturing aren't suitable for external dissemination.

What the customer-complaint closed-loop Agent does is internal management: grab complaints → match in the internal knowledge base → give problem decomposition and suggestions → track resolution nodes throughout → automatically do attribution analysis after closure → tag: is this the first occurrence, or a repeat occurrence?

This data goes directly to the front-end R&D department for improvement support.

I asked Hongxuan, what's the relationship between this and traditional ERP?

He said: compatible, upgrade. Many factories failed to implement APS (production scheduling systems) because preconditions like machine maintenance and material matching are too complex. Schedulers spend 1 hour scheduling and 6 or 7 hours handling exceptions. After the Agent goes in, it gives highly reliable scheduling suggestions in 30 minutes. Humans remain the decision-makers making subjective judgments and final choices, but reliability and traceability are greatly enhanced.

Journey-to-the-West Aliases, Organizational Centrifugal Force, and AI-Native Super-Organizations

At this point in the chat, I noticed something very interesting.

All of Linghe Shuzhi's employees use Journey to the West aliases. Hongxuan is called "Wukong" (Monkey King), there's "Wujing" (Friar Sand), and even demon names.

I asked them—are they here to fetch scriptures or fight demons?

He said: slay demons and monsters all the way, and obtain the "manufacturing true scripture."

Why use aliases? Not because it's fun. They want to build an AI Native organization, which must be a flat, water-droplet-like network structure, without traditional rank feelings.

But this引出 a deeper question: in the AI era, organizations will experience "centrifugal force."

Hongxuan shared a very real experience. In 2023 when he first started using AI, he called himself a "knowledge upstart"—obtaining knowledge was too easy, and for a time he became very arrogant, feeling he could do anything.

Until one R&D meeting, he confidently presented the conclusions AI gave him, only to be corrected on the spot by the frontline team with extremely professional data.

At that moment, he went from "being conquered" to "having a sense of awe."

So he said, in an AI-native organization, the boss's goals must be set clearly and granularly enough. Otherwise, everyone uses AI unevenly, and the organization will fly apart like a centrifuge.

What is the ideal vision of the future?

An enterprise may have only 50 people, but each person has 10 or even 500 Agents behind them, each performing their own duties. Humans and Agents are close partners in business collaboration, with hierarchies interwoven. Your Agent and my Agent first quickly align Context at the bottom layer, then humans make efficient final-review judgments.

The enterprise彻底 transforms from an "efficiency container" into a "judgment container."

Advice for Manufacturing Bosses and Individuals

At the end of the livestream, I asked Hongxuan: if a manufacturing boss wants to start trying AI, what do you suggest?

He said three things:

First, personally experience and try. Don't listen to what others say, don't read articles—open an account yourself, deploy an OpenClaw, use Doubao, and build a visual sense in specific small scenarios.

Second, maintain awe, start from one small thing. Don't expect earth-shaking changes; blind trust easily brings negative effects.

Third, raise the Agent like a "child." It has strong capabilities but lacks social experience. You must have a tolerant heart, take it to see the world more, and give it objective feedback after it does things. Over time, its explosive power will exceed your imagination.

And for ordinary individuals, his advice is:

Stick to your core scenario, don't fantasize, and don't copy what others do. Around the work you're currently doing, find AI tools that can help you accelerate a hundredfold, and define your own efficient way of working.

AI will greatly accelerate your skill acquisition, but the logic behind it, creativity, insight into human nature, and the ability to close the loop and get results can still only be decided by you.

Written at the End

This conversation confirmed a judgment for me:

The manufacturing industry that most AI application companies look down on is precisely the most fertile soil for AI implementation in China.

Not because manufacturing is "backward" and therefore has opportunities, but because manufacturing has real pain points, clear ROI, massive amounts of unstructured data, and a group of bosses willing to change.

Chen Hongxuan said their mission is "to help Chinese manufacturing continue to be great."

I don't think this is just a slogan. When a 1996-born "second-gen" is willing to give up succession and go all in on this hardest track, that fact itself说明:

The wind is truly blowing this way.

No matter what stage of development your founded company is at, no matter which track in the AI application direction you're on, as long as you're an AI Believer, welcome to get in touch with me.

I'll help you accelerate your product, accelerate your financing, accelerate your business, accelerate your influence, and ultimately win the market.

Interview Conversation Details

Core Background and My Entrepreneurial Motivation

Q1: As a "second-gen" of a "cleaning-appliance discrete manufacturing" enterprise with annual output exceeding 3 billion RMB and 40 years of history, why did you choose to jump out and do AI entrepreneurship?

Chen Hongxuan: I've always liked to look at problems dialectically—empiricism is actually a double-edged sword in manufacturing. I don't shy away from my second-gen identity. It's precisely the 40-year longevity of the family enterprise and the scale of annual output exceeding 3 billion RMB that gave me a very high platform and rapid experience opportunities. But many people also strangely ask me back during interviews: "You can clearly lie flat, why come out to follow the trend or play around?" Actually, I was truly broken out of my固有 cognition by AI. I believe the AI era is extremely egalitarian and especially friendly to young people. I chose to come out and go all in to make up for many regrets and doubts I had in the family enterprise's digital management through modern AI means, and to prove my self-worth by creating new things. The excitement of clients and the team is my fuel.

Q2: Before founding Linghe Shuzhi, was there a core "shocking moment" that made you determined to go all in on AI?

Chen Hongxuan: There were two moments that gave me a dimension-reduction-strike-like shock. The first was in mid-2023, when I used ChatGPT to produce a set of company rules and regulations and reform plans, which received recognition from a large number of senior executives within the family enterprise, and the implementation effect was very good. But the most shocking—even making me feel "conquered"—was when I used AI to directly win overseas business. At that time we were going to have a meeting with a foreign tech brand, tackling a cross-industry innovative product structural part. Before the meeting, expert discussions still couldn't verify the feasibility of the academic principles. At noon when eating, I told the team to go eat, and I myself used the large model to try to decompose and deduce, and actually sorted out the academic principles. At the afternoon meeting I spoke confidently and eloquently, and ultimately used the思路 from AI to completely conquer the client. At that moment I realized I must come out and fully embrace AI.

Q3: If Linghe Shuzhi ultimately doesn't succeed, would you choose to go back and inherit the family business?

Chen Hongxuan: My answer is: absolutely not. Our commercialization partner Luke once asked me this question without预设 in our internal program, because in everyone's eyes second-gens have a fallback, while partners from big factories don't. But at that time I very firmly told him I wouldn't go back. Because building a company from 0 to 1 is completely different from doing it on an existing platform, and I very much enjoy the thrill of focusing intensely and passionately on the product and the track with a group of the right people. Even if Linghe Shuzhi 1.0 doesn't get results, I believe as long as our group of people is right, we can also make 2.0, 3.0 or even 4.0. I've always viewed entrepreneurship with the end in mind—once I've identified the essence of focus, passion, and survival, I won't look back.

Underlying Logic of Manufacturing AI Implementation (721 Theory)

Q4: How does Linghe Shuzhi view the relationship between "manufacturing AI upgrade" and "traditional informatization/digital transformation"?

Chen Hongxuan: In my view, the AI upgrade of manufacturing is absolutely not simply layered on top of traditional digitalization. Now many companies doing manufacturing AI often take the first step of helping enterprises sort out and deposit all past data, but I think this lacks the internal logic of upgrading, and we've also taken such detours. We found that only about 10% of an enterprise's past system data can truly be used by AI, because that's "past tense," while the enterprise is "present continuous." If you only rely on past software spreadsheets, it's very unfriendly for Agents to read. What an enterprise truly needs to do is have Agents directly enter the enterprise to do real work, and in the process of working re-accumulate the best Context.

Q5: What specifically does the manufacturing AI capability "721 Theory" refer to?

Chen Hongxuan: This is the core theory I use to define the composition of 100% of an enterprise's future capability, experience and knowledge:

70% of capability and knowledge: comes from the current external world, from the general capability of the large model itself and the entire internet (Everywhere).

10% of capability and knowledge: comes from the enterprise's past accumulation, that is, the enterprise's historical experience, internal knowledge and data.

20% of capability and knowledge: comes from the future. It is the brand-new tacit experience continuously rolled and deposited from the first two combined with humans in the process of the enterprise's real work and collaboration.

Real Application Details of Position-Level Digital Employees

Q6: How to understand the often-said "let digital employees go to work at the enterprise"? How is it implemented on my platform?

Chen Hongxuan: Everyone can通俗 understand it as: we make position-level, or even entry-level digital employees, and we directly send them to work at your company. To carry these digital employees, we've built an out-of-the-box digital platform for enterprises, with three core modules inside:

Wenwen Center (问问中心): This is the AI usage interface, looks like an IM software like WeChat, with various business partners you've recruited on it, and human-machine interaction is all completed here.

Paipai Market (派派marketing): Like a silicon-based talent market, we've packaged many Skills defined around manufacturing positions on it. You can combine skills like building blocks and directly recruit digital employees to work.

Jiji Space (记记空间): This is a human-machine collaboration knowledge base specifically adapted for Agents. In the process of doing things, it can directly and visually deposit the tacit experience of humans and AI.

Q7: How does our best entry-level \[Order-Placing Agent\] specifically complete the end-to-end closed loop in a manufacturing enterprise?

Chen Hongxuan: In traditional manufacturing enterprises, the order-placing position is a very tiring and money-closest bottleneck. Order placers have to grab foreign trade emails from clients' SRM systems and emails every day, or receive a pile of PDF and Excel orders on WeChat/WeCom. They need to manually convert external material codes into enterprise internal codes, and also split orders according to delivery dates and arrange pre-production plans. After our \[Order-Placing Agent\] goes in, weintegrate its system and email permissions. After it learns this position's SOP (Standard Operating Procedure) and Skills, it will itself go to systems and emails to grab orders, automatically complete the mapping of domestic and foreign codes, then directly place them into the ERP system. The tedious entry that originally took 3 to 4 hours of manual time is now fully automated end-to-end by AI in a closed loop, and business colleagues only need to spend 5 to 10 minutes doing final result comparison verification and responsibility兜底.

Q8: Besides saving time, what deeper management value can the \[Order-Placing Agent\] bring to the enterprise?

Chen Hongxuan: It can break the flatness of data, making the enterprise's data objects (Objects) into a three-dimensional network mapping at the bottom layer. This is like why the industry's OpenClaw (crayfish) framework performs excellently—the core is that it makes the most efficient mapping and scheduling of knowledge and tools. When multiple entry-level Agents in the enterprise simultaneously tag and string Links together, the data the enterprise sees is three-dimensional. AI can proactively help you complete "order fluctuation analysis and churn warning." Previously everyone used Excel to look at churn across quarters, and the client had already left; now AI can instantly capture fluctuations, proactively do cause attribution, prompt sales to go to the client site for the first time to win them back, letting the enterprise truly transform from an "efficiency container" into a "judgment container."

Q9: In quality management, what is the essential difference between our \[Customer-Complaint Closed-Loop Agent\] and the common "intelligent customer service"?

Chen Hongxuan: Traditional intelligent customer service only mechanically answers some customer group questions at the front end. But many internal quality issues in manufacturing are sensitive, even can't be said externally. Our \[Customer-Complaint Closed-Loop Agent\] does the enterprise's internal management closed loop. When a customer complaint comes in, the Agent will grab it at the first time, automatically match the enterprise's internal knowledge base of historically solved problems, give a preliminary problem decomposition and targeted suggestion methodology, letting people handle it as fast as possible, and track nodes throughout until the complaint is completely closed. Finally, it will clearly analyze whether this matter is a first offense or a repeat offense, automatically tag it, and these valuable inputs can be directly pushed to the front-end R&D department for product improvement support.

Vigilance About Organizational Change and "Organizational Centrifugal Force"

Q10: After an enterprise introduces AI, why do I emphasize that goal management and corporate culture are even more important than before? Will it bring any side effects?

Chen Hongxuan: Because while AI breaks traditional boundaries, it brings "organizational centrifugal force" to the enterprise. Most existing organizations are top-down pyramid structures. After introducing AI, the speed at which employees acquire and generate knowledge will explode geometrically. If the boss's goals aren't set clearly and granularly enough, without using a strong mission and vision to let everyone "look at the same tree," the entire organization will generate centrifugal force because everyone has too many ideas and uneven capabilities, leading to management losing control.

Q11: When I used AI internally at the enterprise, what was my own experience of "knowledge upstart" and "disenchantment"?

Chen Hongxuan: In 2023 when I first pushed AI in the enterprise, because with a flick of a finger I could let AI help me generate various reform plans and rules and regulations, and the effect was excellent, at that moment I subjectively felt I could do anything and became a so-called "knowledge upstart." But this made me fall into a strange cycle: I started to极度 rely on AI, no longer taking strong subjectivity to deeply think about its right or wrong. Until one time at an R&D meeting, I very confidently threw out the conclusions AI gave me, only to be corrected and proven wrong on the spot by our frontline core team with extremely solid on-site professional data placed on the table. At that moment, I was instantly "disenchanted" by AI—from being conquered by it to being full of "awe" toward it.

Q12: In my envisioned AI-native organization, what should the future super-organization and employee form look like?

Chen Hongxuan: In the future super-organization, the relationship between people and people, people and superiors will no longer be traditional administrative superiors and subordinates. Whoever is the Owner (person in charge) of this matter, everyone aligns toward them and is responsible for the result. Position boundaries will completely disappear. In the future, a 50-person enterprise—each person may lead 10 or even 500 Agents behind them. Employees are like independent business units, with a group of Agents behind them helping them handle the big closed loop from customer acquisition, research, sales, monitoring to production line management. Employees will become "super individuals" within the enterprise—they're no longer mechanical screws, but stand at the core value center making efficient judgments and choices.

Practical Advice for Manufacturing Bosses

Q13: Can traditional factories with low digitalization and poor informatization systems actually use Linghe Shuzhi's AI digital employees?

Chen Hongxuan: My answer is very firm: completely yes, yes! Many bosses worry their factory foundation is thin. In the past, the failure of traditional management software (like heavy ERP) was often because enterprise profits were low and they couldn't afford a large number of dedicated people to enter core processes to maintain data, align spreadsheets, andintegrate links. But the AI era is technological egalitarianism—as long as your factory is operating, quoting, and sending emails, you're generating data. Our Agent进驻 is the logic of directly doing work—it can independently close the loop on one thing. It can simultaneously help you decompose, categorize, and data-tag those most原始 documents, Excel and emails while working, at extremely low cost, without needing you to add 10 extra positions just to implement the system.

Q14: When position-level Agents enter the enterprise, specifically how much money can they help the boss calculate and how much cost can they save?

Chen Hongxuan: This is a completely satisfiable "want-it-all" process in the AI era. According to our actual commercial landing cases at Linghe Shuzhi, when several core Agents in order-placing, finance, and supply chain positions enter the enterprise and work for one year, the time saved converted to average wages can effectively save the enterprise over 1.5 million RMB in costs (an average of 500,000 RMB saved per core Agent). But I often emphasize to bosses that this is absolutely not a pure "staff-cutting" downsizing logic. It's to let the enterprise, with the same number of people, be able to do things multiple times more and do them better, making your entire organization more rule-based.

Q15: For bosses who want to try, what specific AI experience is there to share?

Chen Hongxuan: First, the boss must establish trust with AI, and trust starts from one small thing. I often share my experience: I compare AI to "a child with strong capabilities." A child's growth requires tolerance—it won't be so perfect at the beginning, so you have to take it to "see the world." Before each task you chat with it about the background, and after it's done you throw the meeting minutes, activity feedback, client business cards and other Context (context) to it, and give timely feedback. Over time, after it understands you, its aggregation capability will completely blow the chart. As the founder of Linghe Shuzhi, I deeply know Chinese manufacturing is very large, and our entire team is also a comprehensive team that has been down to the workshop floor and written code to train models. If you still have any questions about the deployment of these digital employees, welcome to communicate further with me (Wukong) at any time!

---

Original publication: https://uniqueresearch.substack.com/p/src-20260529-01html
On-site reading page: https://ffcap.cn/en/research/src-20260529-01html