Does the Company Say AI Is the Top Strategy? Organizational Structure Doesn't Lie

Deep thoughts on AI and aspirations —— ByteDance Deep Thinking Circle

In 2015, Daniel Zhang proposed the “big middle platform, small front platform” at Alibaba. In 2018, he appointed Zhang Jianfeng to lead the big middle platform and placed Alibaba’s most important project, the Alibaba Business Operating System, into this department. Looking at the launch event, it was just a slogan plus a personnel appointment; looking at the organization, a new direction was given resources, given position, given a trusted person—all three boxes checked.

To judge whether a company’s strategy is real or fake, the most reliable evidence isn’t the financial report or the launch event—it’s the organizational structure. Founder Securities’ research team reviewed the organizational evolution of four companies—Alibaba, Xiaomi, JD.Com, and Meituan—and the first lesson they summarized was exactly this: organizational structure serves actual operations; it’s a mirror of the company’s business and doesn’t lie.

Structure Is More Credible Than Words Because It’s Collateral for Commitment

Saying you value a business is cheap. Truly valuing it requires putting up collateral, three things: money, people, power. A new business gets an independent budget, key positions are appointed with trusted core figures who are fully empowered, and resources tilt toward it. Conversely, if a company announces a major transformation but department setup, personnel arrangements, and evaluation mechanisms remain unchanged, then the signal it’s sending isn’t worth full trust. Strategy can be performed; reporting lines, headcount, and budgets are hard to fake.

The histories of these four companies are full of such evidence. JD.Com had 180,000 employees at the end of 2018, compared to Alibaba’s 70,000, Meituan’s 50,000, and Xiaomi’s 16,000. Managing such a massive front-line team relied on military-style governance. That same year, Richard Liu proposed a building-block organization, breaking the three major groups—mall, logistics, and finance—into front platform, middle platform, and back platform: the front platform was cut into agile small teams by scenario and customer type, while the middle and back platforms created standardized components that the front platform could assemble like building blocks. How 180,000 people fight is written clearly in the architecture diagram, more concrete than any speech about organizational agility.

Alibaba’s decentralization culture is equally visible. During the Big Taobao strategy period, internal uncertainty about whether e-commerce would be B2C or C2C led them to simply appoint three presidents to each explore one model, running Taobao, Tmall, and Etao simultaneously. Daring to let three people pursue competing directions—that determination is written into the organizational chart. The management committee system was the same: cultivating middle management by having a group of people nurture a group of successors within mature departments, with the intent of power succession directly embodied in the structure.

Turning This Mirror Toward AI Transformation

This methodology comes from a 2019 retrospective, but I believe its most valuable use today is to hold it up to the “AI-first” claims flooding the streets in 2026.

When a company claims AI is its top strategy, four structural signals can immediately verify its authenticity. First, where does the AI leader report: directly to the CEO, or tucked under some business unit as a support department? The retrospective already established patterns—companies with empowerment cultures have stable positions for innovation business executives; centralized companies rotate innovation business executives frequently. An AI leader being replaced every six months is basically a performance position. Second, what authority do they have: can they mobilize products and personnel, or do they only have the right to make suggestions? Third, has evaluation changed: do front-line department KPIs include AI-related output requirements? Transformation without changing evaluation is like not paying for new behavior. Fourth, which layer is being moved: cutting execution layers is just cost reduction; redrawing reporting lines, merging departments, cutting middle layers—that’s real restructuring.

But this mirror needs an honest qualification: it reflects determination, not success or failure. Alibaba’s big middle platform was genuine determination in 2015, yet it was later dismantled and restructured. An organizational chart can prove a company is serious, not that it’s right. Using it as a lie detector works well; using it as a crystal ball doesn’t. After seeing through the performance, you still need to judge the quality of the direction itself.

Which Old Lessons Still Shine

That retrospective had two other conclusions that fit surprisingly well in the context of AI transformation.

One is capital independence and cash flow. Companies that can execute ten-year long strategies are first and foremost companies not held hostage by cash flow. Xiaomi used a pre-sale model to strictly control supply chain capital usage, maintaining healthy cash flow even during high-growth periods, with capacity left over to invest in the ecosystem; in contrast, hardware companies of the same period applying internet tactics—LeEco, Gionee—were directly crushed by capital. In 2007, Jack Ma proposed that Big Taobao’s GMV should exceed Walmart within ten years. By 2017, Alibaba hit 4.6 trillion RMB versus Walmart’s 3.3 trillion—goal delivered, sustained by the patience supported by main-line cash flow. AI transformation is likewise a money-burning marathon; companies supported by operating cash flow dare to push forward through several quarters of losses; those supported by financing narratives cut when the market tightens. To see the authenticity of a company’s AI investment, first look at where the money comes from.

Two is empowerment culture. The retrospective’s comparison: highly empowering organizations like Alibaba and Xiaomi have continuous front-line innovation, with product ideas largely coming from the front lines; JD.Com is highly centralized, with frequent turnover of innovation business executives and new businesses often attached under core businesses. AI transformation happens to be the type of change that most requires empowerment: decision-making authority must be delegated to the front lines that work with models and customers daily. In centralized organizations, every new tool requires headquarters approval—the speed gap isn’t something employee effort can make up. This is also why centralized companies that talk agility most easily spin their wheels on AI transformation—the bottleneck isn’t tools, it’s authorization.

Next time you read about a company announcing AI as its top strategy, don’t rush to believe or refute. Pull up its organizational chart and ask three questions: Who does the AI team report to, what authority do they have, how is evaluation calculated? After these three answers, you’ll have a pretty good sense of whether the talk is real or fake. Strategy can lie, reporting lines don’t—this is the most durable tool that can be extracted from a decade-long retrospective of four giants’ organizational histories.