In-Depth Analysis of the 2026 Bund Conference: AI Moving from “Showoff” to “Practical Application” – A Coexistence of Bubble and Realization
Key Points Summary
The theme of the 2026 Shanghai Bund Conference was “Co-creating a New AI Economy.” Unlike previous years, the focus has shifted from “how powerful the technology is” to “whether the business models will succeed.”
The most striking detail was that among the Ant Group CEO, a MasterCard executive, an OPPO executive, and Alibaba’s chief scientist, only Ant Group CEO Han Xinyi had actually used an AI agent to complete a real transaction, accounting for just 25%. This reveals a harsh reality: although the concept of “Agentic Commerce” (business conducted through AI agents) is popular, its implementation is much slower than expected, and the necessary infrastructure and trust mechanisms are not yet in place.
Meanwhile, sales of AI hardware (such as smart glasses) have declined, and the robotics industry is starting to experience a “bubble burst.” The industry is no longer blindly optimistic; instead, it is calmly questioning whether users truly need these technologies, whether the business models are viable, and whether the reliability is sufficient. This marks a period of harsh selection, transitioning from “PPT-driven dreams” to practical, real-world applications.
---
Detailed Analysis: Understanding the “Cold Realities” of the New AI Economy from Five Dimensions
1. The “Last Mile” in Payment: Why hasn’t AI Yet Helped You Spend Money?
Phenomenon:
Han Xinyi used AI to purchase snacks for over 40 yuan, but this is a very rare case. Jorn Lambert, an executive at MasterCard, admitted that although he used AI to find products, he didn’t dare to let AI make the payment directly.
In-Depth Analysis:
This is not just a technical issue; it’s also a matter of trust and rules:
- Old systems are not designed for AI: Current payment systems, banking rules, and risk control processes are all tailored for humans. For example, humans hesitate and verify before making purchases, while AI makes decisions in milliseconds. Existing systems cannot distinguish between “AI errors” and “hacker attacks” and cannot handle the vast number of small transactions generated by AI.
- Trust gap: Consumers are reluctant to hand over all their financial information to AI because it’s unclear whose responsibility would fall if something goes wrong (e.g., if AI makes a wrong purchase or induces unnecessary spending). There is a lack of clear accountability standards.
- Conclusion: AI payments are still in the “demonstration phase” rather than the “popularization phase.” To truly let AI manage finances, a new set of financial infrastructure and security standards for machine behavior must be established.
2. The “Cooling Down” of Agentic Commerce: From “Everything Can Be AI” to “Difficult Scene Adaptation”
Phenomenon:
Last year, the industry was extremely optimistic about agentic commerce, but this year, its implementation has slowed significantly. Although Alipay’s “Abao” can order coffee and hail a taxi, it’s still far from widespread use.
In-Depth Analysis:
- Being functional doesn’t mean being user-friendly: While AI ordering coffee may seem cool at conferences, in real life, users care more about speed and accuracy. If AI is slower or frequently misunderstands their requests, they will abandon it.
- Fragmented supply ecosystem: Only a few leading platforms (like Alipay and Taobao) have fully integrated AI. Many small and medium-sized businesses and offline services have not yet adopted AI interfaces. AI wants to help with tasks, but the systems are not ready for it.
- Long-term challenge: Manufacturers need to provide AI with a complete context (e.g., understanding user preferences, budget, location) and establish secure, reliable transaction processes. This requires the cooperation of the entire industry chain.
3. The Decline in AI Hardware Sales: The Truth Behind the Drop
Phenomenon:
In July, sales of smart glasses on major e-commerce platforms decreased by 15.3% year-over-year. Investor Yan Qianhang pointed out that AI hardware has not created new ways of interaction; visual interactions remain limited.
In-Depth Analysis:
- The novelty effect has worn off: Early buyers of smart glasses were mostly tech enthusiasts. When the novelty fades, sales will drop if the products don’t solve a common, essential problem (like a smartphone).
- Lack of breakthrough interactions: Current smart glasses are merely extensions of phone screens, without providing revolutionary interactions. Users still look at screens rather than interacting naturally with the world.
- Return to consumer electronics logic: Investors are evaluating AI hardware based on traditional standards: does it solve real problems? Does it have a large audience? Can it build brand loyalty? If the answers are no, the product will fail, regardless of how popular the concept is.
- Implication: The future of AI hardware lies in creating new demands and new interaction paradigms rather than simply enhancing existing functions.
4. The “Bubble Burst” in Embodied Intelligence (Robots): From “Dancing” to “Working”
Phenomenon:
More than 40 robotics companies were showcased, but the industry is discussing a potential bubble. Han Zheng, CEO of Sudu Technology, noted that consumers demand near-perfect reliability (99.9%) from robots, which robots currently only achieve in 5%-10% of tasks.
In-Depth Analysis:
- Reliability is critical: Robots performing impressive tasks in labs (like dancing or playing music) are attractive, but in real-world scenarios (e.g., pharmacy sorting or factory assembly), a single mistake can cause significant losses. Users won’t accept robots that are only 80% reliable.
- Market overestimation: There was a huge hype that robots would replace much human labor, but in reality, they can only handle very specific, repetitive tasks.
- Capital bubble: The market expected many leading robotics companies in China, but only a few will survive. This means many startups will face elimination.
- Practical approach: Smart entrepreneurs are shifting from the vision of “general-purpose robots” to focusing on high reliability for specific tasks. For example, they might develop a robot that can efficiently sort deliveries rather than one that can do everything but poorly.
5. Industry Consensus: Long-Term Optimism, Short-Term Pragmatism
Phenomenon: Despite declining sales, slow implementation, and controversy over bubbles, many investors and entrepreneurs still see value in embodied intelligence and the new AI economy.
In-Depth Analysis:
- A dark period before dawn: The industry is in the “early stages of dawn.” With the accumulation of data (e.g., human-centered data, world models) and the application of scaling laws, technological breakthroughs are imminent.
- Undryed capital: Despite the bubble, overall investment remains low, but it’s being reallocated from companies that just make exciting claims to those with real data and practical applications.
- Fiercer differentiation: In the next few years, the industry will see a sharp split:
- Winners: Companies that solve specific reliability issues, have unique data, and clear business models.
- Losers: Those that rely on hype, have homogeneous technology, and cannot prove their commercial value.
- Core logic: “Lowering expectations” is not pessimism; it’s a rational adjustment. The industry is distinguishing between long-term visions and short-term realities. In the long run, AI will transform the economy; in the short term, we must focus on solving practical, reliable problems.
---
Lessons for the General Public:
1. Don’t rush to hand over your wallet to AI: AI payments are not yet mature; be cautious and verify large transactions personally.
2. Be cautious when buying AI hardware: Only buy if it clearly solves a specific problem for you (e.g., real-time translation, assisted driving).
3. Focus on reliability, not just coolness: When evaluating AI products or services, ask yourself: Can it work reliably? Is it better than what you use now? Who is responsible if something goes wrong?
4. The industry is undergoing a reshuffle: If you’re interested in investing or starting a business, now is not the time to chase hype; instead, observe which companies can actually implement their ideas and generate revenue.
In one sentence: The 2026 AI economy is no longer about showing off technology; it’s about testing its practical value. Those who can turn AI from a “toy” into a “tool” will shape the future.