Original · Unique Research · 2026-06-09
Editor's note: The first-person report and its judgments belong to the original Chinese author. This English rendition retains the three-founder roundtable (Sico Education / Hou Hao, Benmo Technology / Yang Kai, Milu Interactive Wapitee / Ken, hosted by CGL's Xenia Wang), the five themed sections, and the complete five-part transcript. All named companies, products, and people are preserved. Founder statements are source attributions, not independently verified findings.
AI Industry Observation
The Cruel Organizational Truth of the AI Era: The Grassroots Should Never Have Been the Anxious Ones
A three-founder roundtable surfaces the core judgment on today's organizational change
"
This is a story about how power is redistributed inside an organization.
A 3D designer had been at the company three years; the boss barely knew he existed.
Then last week, Sico Education (思珂特教育) ran an internal hackathon — over 100 employees split into 20-plus teams, each using AI to solve a real business problem. This designer's submission was a platform that builds an AI profile and skills map for every employee in the company — like an internal "smart dispatch system" for talent.
Sico CEO Hou Hao (侯昊) was stunned on the spot: this person had never even been on his radar.
Hierarchical organizations systematically filter out people who aren't good at speaking up. AI breaks that filter.
This isn't a story about "will AI replace people." It's a story about how power is redistributed inside an organization.
01 | What AI Is Eliminating Is the Information Middle Layer
A traditional company's org chart is essentially an information-transmission and processing mechanism. The front line executes and collects data; middle management aggregates, filters, and reports upward; senior leadership makes the call. Middle management's value comes, to a large extent, from "information exclusivity" — it holds information that neither the top nor the bottom fully possesses.
But AI has broken that logic.
Benmo Technology (本末科技), which makes core components for smart hardware, grew from 200 people to 700-800 last year, and senior leadership's sense of the front line grew blurrier. Their fix: use AI to "kanban-ize" frontline data and middle-management information, letting leaders read the numbers directly, bypassing long reporting chains.
Hou Hao put it directly: "In the future it may be the CEO connecting directly to every node; each person carries an Agent, transparently interfacing through dashboard data. No more CEO-1, CEO-2 layers."
In plain terms: middle management's information-processing function is being progressively replaced by AI. The remaining question is whether middle managers can quickly turn into "value creators" rather than "information movers."
This is why Liu Run (刘润) says: in the AI era, the ones most anxious are never the front line — it's middle management. Three founders, each through their own company's practice, independently verified this judgment.
The information moat is gone; the layer that existed only by controlling information is losing its reason to exist.
02 | The New Talent Filter Has Switched Coordinates Entirely
Talent standards usually lag technology shifts. This time, the lag is short.
Milu Interactive (眯鹿互动 / Wapitee) does full-service AI hardware brand going-global work. Co-founder Ken defines the hiring bar in one word: curiosity. "We ask what products they've used lately, what fun things they've done. Degrees can do a first pass, but whether they truly join depends on whether they stay curious about the world."
This isn't a soft requirement. Curiosity is a proxy for ability — people curious about the world actively accumulate cross-domain understanding, and cross-domain understanding is the core raw material for doing creative work well in the AI era.
Benmo partner Yang Kai (杨凯, Gary) screens people with three words: wants to do it, can do it, knows his lane. "Wants" is subjective drive; "can" is execution; "lane" is a sense of boundaries. In a manufacturing-plus-AI hard-tech company, employees must self-drive to fit a fast-changing market while clearly holding their role without overstepping. The logic: AI-era talent must be a person who can close the loop, not a person waiting for orders.
Hou Hao's framework is the most systematic. He proposes the concept of "M-shaped talent": not T-shaped (one horizontal bar of general literacy plus one vertical bar of specialty), but someone with multiple professional pillars. The M-shaped talent's core trait is learning transfer — in the past 2-5 years, did they enter a brand-new field and become an expert in it?
But learning transfer alone isn't enough. Hou Hao stresses a second dimension: process and systems thinking. "To wield AI for Agent orchestration you first need systems thinking; the tool is only the third thing. Many people are point-skilled specialists; they can't orchestrate their own workflow."
Three people, three standards, one direction: can this person, in a highly uncertain environment, quickly identify the problem, self-organize resources, and close the loop?
The core of ability isn't which tool you use; it's whether you can define the problem itself.
03 | Culture Isn't a Slogan — It's Determined by Org Structure
Many companies treating AI transformation as a publicity exercise treat culture-building as promotional material. But culture is a function of org structure; if the structure doesn't change, culture is just waste paper on the wall.
Hou Hao's hackathon let that 3D designer surface not because the company chanted "we encourage innovation," but because it created a showcase mechanism that bypassed the hierarchical reporting chain — everyone on the same stage. That's a structural change, not a slogan change.
Ken's team is only 40-plus people; he deliberately won't expand to 100 or 200. His judgment: once creative work is assembly-lined, quality starts to drop. So they keep a lean team, renting a converted old factory in Nanshan, Shenzhen, regularly hosting gatherings of Chinese and foreign founders so employees touch the latest overseas products and teams directly. Creativity comes from real contact, not from PowerPoint market analyses.
Yang Kai's company grew 3-4x and faces cultural dilution. Their fix: give the whole company a unified anchor of belief. Benmo's mission is "to drive the world": their smart motor modules sit inside high-end consumer electronics and robots; once these products hit the market, Benmo's technology literally drives the world's operation. There's an industry saying: "No Benmo, no high-end." This mission is concrete and tangible, not an abstract value sticker. It lets 700-plus people, through the pain of organizational-effectiveness change, know clearly what their work means.
Culture works only if it grows out of org structure, not if it's written into a press release.
04 | Knowing AI Tools Is a Floor, Not a Moat
This is the most important unstated premise in the conversation; no one says it outright, but all three founders' practice points to it.
Hou Hao says AI proficiency matters for non-technical education-industry staff, but it ranks only third — behind learning transfer and systems thinking.
Ken says the spread of no-code/low-code tools lets non-technical people do what once needed a front-end engineer. The "knows the tool" barrier is rapidly approaching zero.
Yang Kai says AI's role in hard tech is to accelerate people's understanding of product and market and remove repetitive work, not to replace people who truly grasp manufacturing's underlying logic.
Three industries, three angles, one conclusion:
When the barrier to using AI tools approaches zero, the tools themselves stop being a competitive advantage. The deciding factor shifts to another dimension — are the questions you ask more accurate, deeper, and more worth solving than others'?
05 | The Direction of the Organizational Revolution Is to Let the People Who Deserve to Be Seen Be Seen
From this roundtable you can piece together a roadmap for AI-era organizational evolution:
The information middle layer's value is falling; the distance between true value-creating nodes (executors and decision-makers) is shrinking; people previously drowned out by "not being good at reporting" now have a chance to be seen; the dimensions for measuring talent shift from "background + experience + degree" toward "curiosity + learning transfer + systems thinking."
But this direction holds under one condition — the company's top leader must truly be willing to dismantle hierarchy, not just draw a mesh on the org chart.
Hou Hao is most candid: "The CEO has to hold the big picture in his head, then kill off what doesn't fit; only then can the new AI organization form." Then he adds:
"Killing it off is definitely not easy; I'm in the middle of doing it."
This is what AI-era organizational change really looks like: not a one-time technical deployment, but a continuous, painful dismantle-and-rebuild that may never finish.
That 3D designer silent for three years was pushed onto the stage by a hackathon. The question is: does your company have that hackathon?
When AI lays everyone's true ability out on the table, is your organization ready to receive the people who were never seen?
More Conversation Detail
Panelists:
Milu Interactive / Wapitee Co-Founder — Ken
Benmo Technology Partner — Yang Kai (Gary)
Sico Education CEO — Hou Hao (侯昊)
Host: CGL Senior Partner — Xenia Wang
I. Panelist introductions and current AI business
Xenia Wang: For AI-native companies and those exploring organizational transformation in the AI era, this is a very valuable topic. Today we have three founders from very different AI tracks to talk about 2026 AI-era organization building, culture, and talent. Let's start with brief introductions and your concrete AI business. Ken first.
Ken: Hello, I'm Ken of Milu Interactive (Wapitee). We mainly do full-service brand going-global work for consumer electronics — mainly AI hardware — helping them build brands and grow in the European and American markets.
Yang Kai: Hello, I'm Yang Kai (Gary), sales director of Benmo Technology. Benmo mainly does underlying service technology and key core components for smart hardware. This direction mainly targets large B2B clients and some high-end consumer electronics, providing foundational support technology. Thank you.
Hou Hao: Hello, I'm Hou Hao; our company is called Sico Education. We're an international-education institution that organizes some of the world's top competitions, offering high schoolers heading abroad competitions in economics, business, and technology, including AI competitions.
We're now building a second curve — a credential and talent-data system. The traditional model evaluates people by diploma; we hope in future to use talent data to connect companies, universities, and high schools, building a brand-new talent ecosystem.
II. How traditional orgs balance the old business with AI transformation
Xenia Wang: The three panelists are leading in their own tracks with distinct styles. Let's dig in, starting with Hou Hao. Your company is over a decade old and is now doing AI organizational and business transformation. How do you balance the old business with the new AI business?
Hou Hao: Our company has some years on it, so a traditional org doing AI transformation does face challenges. But last week we just ran an internal AI hackathon, about 100-plus employees split into 20-plus groups. It was fiercely competitive; many employees previously resistant to or less eager about AI joined in.
The winning team's project basically achieved one-click smart completion from sales-side mapping to closing follow-up — essentially a fully intelligent CRM. Through this hackathon we found many young talents are very strong at business-process identification, creation, and AI ideation.
This gave us a very felt talent standard: what kind of talent matches AI-native, and what kind may not keep up. So we're now doing fairly major organizational change, picking out such talent and putting them to enable business in the first or second curve.
Xenia Wang: As a 10-year-old company, doing AI transformation bottom-up and even discovering AI-native young force in the process is quite inspiring for many enterprises.
Let me ask Gary. You're in hard tech, your track is right on the hot wave. Your org surely has many R&D and marketing people. Facing different functions, how do you pick who fits or can do AI transformation?
Yang Kai: Our company is five years old. During this time our production organization moved from manual work five years ago to importing automated equipment now; in the process pure manual labor has been disappearing. That doesn't mean no human labor, but using the process to raise efficiency and optimize the org. Underneath, its AI-ification is also a strong intelligent-automation transformation.
Meanwhile, in large business pushes like sales or market breakthrough, you need to quickly understand customer needs and fit the market. In this step, AI accelerates your understanding of the product; fast front-end guidance speeds R&D and mass production, accelerating the whole product iteration. AI's enablement across the full product chain is extremely efficient.
Also in R&D, AI-based algorithms and models all have a certain foundational logic architecture; we need more people to adapt the foundational architecture to meet different functions or needs. Overall, AI's future hard-tech use will penetrate manufacturing, R&D, sales, and every step, reducing repetitive work and redundancy and raising everyone's ability to quickly fit the market.
Xenia Wang: Gary mentioned that in an AI hard-tech org, workflows keep iterating from early manual to automation to now AI algorithm matching. The corresponding talent must also cross from automation to AI algorithms; the bar on people is indeed high.
Yang Kai: Yes, exactly.
III. How AI-native orgs define architecture and screen talent
Xenia Wang: Finally, Ken's company. To me you're an AI-native org, doing both going-global and brand marketing. From the start, how did you define your org architecture and talent profile?
Ken: Our company is relatively new, founded at the end of 2023, spun out of another company to focus specifically on full-service brand work. We saw 2023 as the so-called "AI year," OpenAI and ChatGPT giving us ability to do more and replace lots of previously very repetitive work. Seeing the opportunity, we decided to all-in the track and later focused on serving mainly AI-driven hardware brands.
Our talent needs differ a lot. As AI tools enable more, what people do differs greatly from before. When helping brands build and grow independent sites in Europe and the US, we used to need many front-end engineers to handle tech. Now no-code/low-code tools are very convenient, letting many non-technical colleagues do this work.
Building the team, we found we now more need all-rounder PMs. Front-end tech issues can be quickly solved with AI, but they must have good aesthetics and understand global culture (mainly European and American). What people need to know is actually more comprehensive and broad. So the bar on people rose, but in return we can use a leaner, smaller team to deliver more services.
Xenia Wang: Everyone must be curious — in the AI era everyone wants talent to have AI ability. Hearing you, the three panelists' companies, whether cultivating internally or searching externally, have seen some excellent people. I want to know how you screen and evaluate these AI talents; what dimensions?
Ken: I'll start. Our hiring may be unusual; we have a hard requirement: we want someone "curious about the world." Because we mainly do creative and marketing work, how well you tell a story depends on this marketer's cognition of the world — how much they've seen, how many products they've used, to make the story rich.
So in interviews we ask: "What products have you used lately?" "What fun things have you seen or done?" Degrees and basic background help us do a first pass, but later, to truly join and do the work well, we still value whether they stay curious about the world, willing to keep learning and trying new products.
Xenia Wang: So curiosity is your keyword for recruiting core AI talent. Gary?
Yang Kai: On hard tech, our "curiosity" requirement may be less heavy, but we need talent who truly lands the whole industry chain. Our guidance to employees is three words: wants to do it, can do it, knows his lane. Through self-drive and boundary sense, we want to build a talent pipeline that lets hard tech iterate fast and quickly adapt to market needs in the AI era.
Xenia Wang: Agreed. This is what high-end recruiting often says: AI-era talent must truly learn to discover and define problems and close the loop. In hard tech this may validate more directly.
Hou Hao? You doing AI education now also need many young people; how do you define or measure?
Hou Hao: I can talk about the shift in recruiting trends. In the past we valued professional ability, the so-called "T-shaped talent" (horizontal axis general literacy, vertical axis a field's professional ability). But from this transformation onward, we especially feel the org needs "one specialty plus broad ability." We used to talk "π-shaped talent"; now there's also "M-shaped talent."
The M-shaped talent's pillar is that he can cross from one professional field to a second. First he needs this confidence; second he can truly learn fast. Such talent is relatively rare. If we screen people with the old industrial-era lens, we only look at which field they specialize in and may miss whether in recent years they entered a new field and became an expert. So in future we'll value people's "learning transfer" more, whether they have success cases.
The second dimension: when reworking business processes we found people with process and systems thinking can well wield AI for Agent orchestration. Mastering tools is the third step; many point-skilled specialists may lack this training and won't orchestrate their own workflow. If someone has better engineering thinking and systems thinking (i.e., workflow orchestration), they're more competitive in the AI era.
Third is AI proficiency. We in education are mostly non-technical; they must proactively embrace AI's technology, software, and tools. In short: learning transfer, process thinking, and AI proficiency.
IV. AI-era org architecture and culture building
Xenia Wang: The panelists from different angles mentioned the underlying logic for selecting AI talent. The past 10 years everyone looked at experience, background, language advantage; today more at underlying curiosity, learning transfer, and loop-closing.
After attracting people, how to build culture to keep them and create value is what founders must think about. In your org, what core cultural genes and elements do you value most?
Hou Hao: I think culture is largely tied to org structure; the organizational form determines culture's genes. We went from early functional forms to, around the pandemic, learning Huawei to do process transformation, opening up into a matrix model. This brought new challenges in cross-department collaboration.
In the AI era, the future org may be a "node-network system." We're still transforming. In a networked org, my feeling is: smart, capable people want to play with equally capable people. Can we design a working mechanism where the CEO connects directly to each node, not through traditional hierarchy (CEO -> -1 -> -2)? Everyone carries an Agent, transparently interfacing data through dashboards. This is the culture we've longed for many years, fully combat-capable and free. The CEO needs a big picture in his head, kills off the old modes that don't fit the new system, and the new AI org forms.
Xenia Wang: But this "killing off" process — do you think it's easy?
Hou Hao: Definitely not easy; I'm in the middle of it. For example, this hackathon company-wide showcase — without this opportunity I'd never have seen the excellent people below. I heard a podcast saying the AI era isn't just about layoffs; many who were introverted and bad at reporting to the boss have come forward. This time we had a 3D designer who alone built a 3D-modeling employee Agent platform, putting every employee's profile and skills up. Under traditional hierarchy such talent might never be seen. If new ways break the limits and let excellent people show, the org's vitality will be enormous.
Xenia Wang: This reminds me of Liu Run's recent sharing: in the AI era the anxious ones shouldn't be frontline employees but middle management. Per Hou Hao's mesh structure, CEO connecting directly to nodes does fade the traditional middle layer. Gary, please share your culture.
Yang Kai: Benmo has developed five years and headcount is growing fast. From about 200 in 2023 to six to seven hundred, even near 800 today. As people change, senior leadership can't finely attend to every frontline employee. So our organizational-effectiveness change is moving from a flat org to one pursuing organizational efficiency.
For us the biggest pain point is raising organizational efficiency and straightening the top-down management path. Two core points run through. First is organizational-efficiency change — modularizing org functions and capability building, achieving top-down "responsibility empowerment." Leaders can use AI dashboards to visually output frontline data and middle-management data, enabling efficient reporting.
Second is org-culture building. As people grow, culture becomes every organization's faith. Only by embedding culture into the org can efficiency come out; otherwise the org develops serious departmental barriers, hurting internal coordination. Culture must empower the change, achieve fast connection, and ensure the company doesn't swing wildly as it scales.
Xenia Wang: Gary, from 200 people fast to near 800, how do you ensure culture isn't diluted?
Yang Kai: At different stages, what an enterprise pursues in culture differs. Every year, everyone heads to the same belief — the company's strategic goals. Through breaking down strategic goals, each department clearly knows its responsibilities.
Meanwhile we build the "Benmo person" belief. Benmo's mission is "to drive the world." We make robots, have the world's first dual-wheel foot robot, and apply smart motor modules in consumer electronics. The market saying is "No Benmo, no high-end"; Benmo brings higher added value to high-end products. That's our reason to exist. Once people build this belief, pushing organizational effectiveness raises efficiency faster.
Xenia Wang: Future high-end robots (including home robots) reaching the consumer side will be driven by your motors and components; very much looking forward. Ken, what's the underlying cultural element in your AI org?
Ken: Our team is small, just 40-plus people, relatively simpler and flatter. The partners each own brand site-building, influencer ads, or outward business development, managing directly and meeting regularly.
Our culture pursues relaxation. To spark colleagues' creativity, at the end of last year we rented an old factory in central Nanshan, converted into a Silicon Valley-vibe office.
Since last year we've participated in several head AI projects; many US teams proactively reached out wanting us to build their brands. A pure-American team asking a pure-Chinese company to build an American brand is quite special. On the hardware side, these overseas projects ultimately must come to Shenzhen. So we regularly host many Chinese-foreign founder gatherings, encouraging employees to join and discuss. Back to what I said — only by knowing more about the world, understanding overseas culture and the latest products, can they tell stories well. Our culture is driving everyone to join fun things, meet fun people, and keep producing good ideas.
Xenia Wang: Those fun young people inside — any labels?
Ken: Everyone has their own character. Today's young people gladly invest huge time in hobbies; on our team there are deep 2D/anime fans and people into different niches. We strongly encourage them to dig into their interests.
Everyone says "AI makes every product worth redoing." AI is relatively universal; it can redefine and empower many hardware products. You don't know what products will come in, or what teams we'll work with. This more niche know-how greatly helps us tell brand stories.
V. Summary: underlying logic held, and mindsets shed
Xenia Wang: You've built an open, diverse environment letting young people exercise curiosity, which may extend new business possibilities. Finally, please summarize: in developing your AI business, what underlying logic did you firmly hold? What did you experience and feel should be shed? Gary first.
Yang Kai: We mainly face To B clients; everyone must have clear cognition: face the market, face demand. We need to quickly connect with the market; the market drives Benmo's internal organic development. If we purely position as a manufacturing foundation, we must keep looking at what the future needs and should do, so we truly apply our products into the coming AI wave.
Hou Hao: On what doesn't change, I think people in education keep their mission and feeling for the education cause; that's the core reason we're together. Education isn't a concept of chasing a hot wave.
But after the AI era arrived, education hit a new turning point. Traditional education makes people anxious: even with a good diploma, does a child's future guarantee a good job? Can they adapt to the times? This is a great opportunity for us doing education innovation and quality education. So employee mission is the unchanging base color.
Second, employees must show the literacy the AI era needs: curiosity, transfer learning, and the ability to keep breaking boundaries. If employees can do these, the company can make good AI education products.
Ken: On talent, we still value curiosity most. On culture and development strategy, we now take a leaner route, not blindly expanding to 100 or 200 people. Because once creative services become an assembly line, the core thing may be lost in the process. Keeping a lean force lets us do more high-value work.
This also gives us our own customer-selection standard. First we ourselves must understand the product and be its users (this is why we encourage colleagues to dig personal interests); second, we measure whether it's truly fun and a real market need. AI is relatively universal now, many things seem empowerable, but we assess what's actually needed, not blindly developing just to chase AI. Only people and things we understand and find fun will we partner on.
Xenia Wang: Thank you three. In today's AI era, the capital narrative (measuring a company's health by A/B/C rounds) is gradually failing. Beyond discussing new business possibilities, everyone more often looks from the organization, talent, and culture dimensions at whether a company can go further.