The Wall-Breaker: From BYD to Haier — How Chinese Enterprises Are Tearing Off AI Agent's "Tech Demo" Label
In May 2026, two industrial announcements arrived almost simultaneously: on May 12, Haier Smart Home broke ground on the world's first AI Agent factory for central air conditioning at Thailand's Rojana Industrial Park, with over 1 billion RMB investment; on May 8, at the "AI Strategy Entrepreneur Forum" in Hong Kong, BYD's AI chief Huang Hai revealed that thousands of AI agents are already serving R&D, auditing, and customer service across the company.
One is China's white goods leader; the other, China's EV giant. Different industries, but pointing to the same core question: can AI agents create genuine value in manufacturing enterprises' core operations?
Haier's Answer: Embed "AI Decision-Making" into the Production Line
Three dimensions worth examining:
Intelligence level: AI deeply integrated from R&D to manufacturing — real-time data monitoring, analysis, and autonomous decision-making. Management paradigm shifts from "human decision" to "AI decision." Factory automation rate exceeds 65%, with a 5-level smart logistics system supporting multi-category, small-batch flexible production.
Scale effect: 180 mu (12 hectares), full-line commercial HVAC products. Combined with the earlier Chonburi residential AC base (6 million units/year capacity), Haier's AI manufacturing matrix in Thailand is taking shape.
Strategic depth: This isn't simple capacity transfer. Haier is using Thailand as a pivot, deploying "AI factories" as a competitive differentiator for the Southeast Asian HVAC market.
BYD's Answer: Thousands of Agents, Not One
Huang Hai's original words deserve attention: "We're not using AI for AI's sake. We want AI to help the enterprise achieve cost reduction and efficiency improvement."
Not a grand narrative of "one AI platform ruling everything," but thousands of scenario-specific agents: some doing code reviews, some generating test cases, some analyzing customer complaint trends, some optimizing logistics routes.
Three key details: (1) Organizational transformation leads — AI projects must be driven by business units, not IT departments. (2) Complete self-built system — "AI middle platform + knowledge middle platform + application development platform" since 2022. (3) Utilitarian philosophy — no AGI or superintelligence rhetoric, just "cost reduction, efficiency improvement."
Two Paths, One Trend
| Dimension | Haier Model | BYD Model |
|---|---|---|
| Entry point | Build AI-native from greenfield | Layer AI onto existing systems |
| Agent scale | One factory-level AI system | Thousands of scenario-level small agents |
| Core goal | Quality + efficiency + flexible production | Cost reduction + efficiency across the board |
| Approach | Physical + AI synchronous construction | Business-unit-driven, IT-supported |
Three Takeaways for Followers
-
The first person accountable for AI deployment isn't the CTO — it's the business unit GM. Only when a business unit claims the AI project's KPI does it escape the curse of "IT built it, nobody used it."
-
"Thousands of agents" beats "one superintelligence." Manufacturing complexity means universal AI can't dominate. The right strategy: break AI into pieces — each piece solves one clear problem, each problem has quantifiable ROI.
-
AI-native factories are achievable, but timing matters. Haier could do it because it had a strategic window for new capacity. For most enterprises, "find AI entry points on existing production lines" is the most pragmatic path.
The biggest change in 2026 isn't AI technology advancement — it's the qualitative shift in how manufacturing enterprises perceive AI: no longer a "tech demo" topic, but an engineering metric that belongs in monthly business operations reviews.