---
title: "On the Day of the Hong Kong AI Summit, I Saw AI Truly Begin to \"Take Over the Real World\""
author: "Unique Research"
sourcePublication: "Unique Research Substack"
originalPublishedAt: "2026-04-18T12:00:09+00:00"
canonical: "https://ffcap.cn/en/research/src-20260418-01html"
source: "https://uniqueresearch.substack.com/p/src-20260418-01html"
language: "en"
---

# On the Day of the Hong Kong AI Summit, I Saw AI Truly Begin to "Take Over the Real World"

_Original · Unique Research · 2026-04-18_

_Editor's note: The first-person report and its judgments belong to the original Chinese author. This English rendition retains the opening essay, all six panel summaries, two open-mic sessions, and closing consensus. Product, market, technical and company figures are source or speaker claims, not independently audited findings. Company, personal and product names are transliterated where official English forms remain unverified. The event was held on April 17-18, 2026 (source dated April 18, 2026). The next event preview (May 21, 2026 Shenzhen) is preserved as source content._

Global Unique Awards

2026 Global Unique Awards · Hong Kong AI Summit Concluded Successfully

AI Is Restructuring Enterprises and Roles

At HALL 2 of the AsiaWorld-Expo in Hong Kong, the "2026 Global Unique Awards · Hong Kong AI Summit," co-hosted by Unique Research and Global Sources, concluded successfully.

Listening through the full-day agenda, you will find a very distinct signal: the AI discussion in 2026 has switched to "how to become enterprise productivity, product form and commercial results." At the same time, it has also landed on several realistic questions: who will be replaced first, what products truly hold up, which business models can work, and how enterprises should organize the relationship between humans and AI.

Starting from an Exhibition Hall, First Build the Global AI Ecosystem Plate

At the beginning of the event, the opening remarks by John Kao, VP of Global Sources Hong Kong and Overseas Exhibitions, set the tone for this conference.

John Kao has been at Global Sources for 37 years and, at 61, is still deeply researching AI. He mentioned that he personally built this venue 20 years ago, and it has witnessed brands such as ASUS, Anker and BYD going global all the way. In this round, what he is optimistic about is no longer the traditional consumer electronics upgrade, but the opportunity window in which AI is redefining the entire industry chain.

In his judgment, TSMC's 2-nanometer technology, CPO optical communication, copper-clad laminates and PCBs will all become new foundationalfulcrum in the AI era. He also clearly pointed out that the global AI leading forces are still China and the US, but when it comes to truly large-scale mass production and truly putting AI into consumer electronics, China will be the most important landing ground. Precisely because of this, Global Sources will next pour more resources into the AI ecosystem—from exhibition hall layout to buyer demand matching, the entire exhibition is converging toward "AI industrialization entry point."

This opening is very important. He is reminding everyone: AI has already begun to turn from a technology wave into an industrial entry point.

Digital Employees, Not Assistance but Replacement

The theme of the first Panel was "Token Economy: The Evolution of Digital Employees from Computing Power to the Productivity Era," hosted by Wu Wei, founder of Unique Research. Guests included Chen Peilin, CEO of Tutu Yangxia; Weilian, founder of ClawdChat.cn; Guo Zhen, founder of Shulex; and Wu Xiankun, founder of Kuse.ai.

This session was the most direct, because what the guests discussed was not an imagined future, but replacement that is already happening.

Chen Peilin represents the path of using AI to rebuild internal enterprise workflows. Tutu Yangxia itself mainly does multidimensional tables, but what is truly impressive is that they are already using AI to restructure team collaboration, accomplishing with fewer people what used to require more people. Weilian's ClawdChat goes in another direction—not making a single-point tool, but attempting to build a network of interconnection between Agents, so that AI capabilities can be called, traded and combined. Guo Zhen's Shulex is more pragmatic, targeting cross-border AI intelligent customer service, and daring to sign performance-commitment contracts—essentially telling clients: I am not selling models, I am selling results. Wu Xiankun's Kuse.ai focuses on enterprise fixed-process automation, launching AI employee products that directly take on repetitive, standardized and processable work.

The guests had a consensus, put very bluntly: what large modelspriority replace is not necessarily manual labor, but rather high-intelligence yet highly processual white-collar work. Positions without emotional value, without complex judgment, and purely execution-oriented are theoretically already standing at the edge of replacement.

But they did not take the issue too lightly. Application-layer startups currently indeed face the risk of large-model vendors "eating downward," so for small and medium companies to survive, they cannotremained at general capabilities—they can only dive into vertical industries,accumulated cases, workflows and industry memory. Digital employees are also not the more the better; they must be designed according to permission management, separating internal and external, to avoid context being scattered and the system becoming messy and uncontrollable. Overseas clients are more willing to pay for experience improvement, while domestic and Asian markets are more direct—whoever can reduce costs and increase efficiency is more likely to close deals.

By the end of this session, a new enterprise form had actually been outlined: in the future, enterprises may only have three types of roles—Builder, User and Service. Product R&D is responsible for building systems, reviewers are responsible for judging AI results, and offline service providers are responsible for the part of the physical world that AI cannot yet handle. Thea large number of remaining intermediate processes will increasingly be handed over to digital employees.

Not Stacking Features, but Finding a Real Scenario

The theme of the second Panel was "The Spirit in the Shell: Agents in Devices and New Business Opportunities for AI Hardware," hosted by Kang Zhengzhong, host of World Manager's "Tech Vision." Guests included Li Shuting, CCO of Dongfang Qiming; Hua Kun, founder and CEO of Wavenote; Jiang Dawei, General Manager of GYGES LABS Greater China; and An Shanzi, founder of Zhuijiao Technology.

If the previous session was about "how AI enters organizations," this session was about "how AI enters devices."

Jiang Dawei talked about AI glasses with a very clear viewpoint: the endgame of this type of product is not having many features, but making people willing to keep wearing them. There is an unavoidable "impossible triangle" between battery life, weight and capability—whoever first balances this well will truly have an opportunity. Hua Kun talked about recording and translation hardware, drawing the product boundary very clearly—it is not a phone substitute, but a professional supplement to the phone. What truly pays are heavy users who want efficiency, not gimmicks. As for privacy and security, it is indeed important, but from a commercialization perspective, it is still a high-cost, niche demand in the short term. Li Shuting's Dongfang Qiming leans more toward the health and wellness direction; she emphasized that the core of wellness AI hardware is not a feature list, but emotional value—whether it can provide "warm service" through vertical models in the Eastern cultural context. An Shanzi talked from the combination of XR and AI to another judgment: lightweight AR is more suitable for landing first, and MR headsets are still difficult to scale in the short term.

The most valuable part of this session is that it pulled AI hardware back from "technique-flaunting competition" to "product judgment."

In 2026, the core consensus on AI hardware is no longer making an all-around device, but deeply cultivatingsegmented scenarios, getting wearing feel, usage feel and frictionless experience right. The directions that truly have opportunities are instead those lightweight, long-battery-life products that users can naturally bring into their lives and work. Starting from the B-end first, then slowlypenetrate into the C-end, has almost become the industry's default path.

Orders and Closed Loops in the Open Mic

The next open mic felt like a set of condensed industry samples.

Zhou Yixu, founder and CEO of Cybertron, shared IP + AI embodied intelligent hardware, enabling figurines, plush toys and robots to have conversational and character-interaction capabilities, quickly helping traditional figurines complete AI upgrading. Wu Qilin, founder of Lingxiao Technology, makes high-cost-performance emotional companionship robots, benchmarking Japanese high-end products but bringing the price down to one-tenth, and has already secured government orders at the million-unit level. Chen Ming, founder of Nuomibao, focuses on high-responsibility scenarios such as law, tax and compliance, proposing moving from "tool intelligence" to "responsibility intelligence," emphasizing that AI output must be reviewable, accountable and embeddable in real workflows. Wang Ting, partner and CCO of Linghe Shuzhi, focuses on manufacturing, using AI digital employees to replace repetitive data-handling work in sales and planning positions. Bill Li, founder of PainHunt, brought another path: using PMF tools to capture real demand and payment signals from eight major North American platforms, helping overseas entrepreneurs avoid detours.

This set of sharing particularly illustrates one thing: AI-native projects that can still stand up and speak in 2026 can no longer rely solely on "vision." What everyone is more willing to talk about is orders, mass production, delivery, review mechanisms, client results and overseas payment signals.

Emotional Companionship Track, Also from "Buzz" to "Is It Worth Money"

The theme of the third Panel after the afternoon began was "Emotional Companionship: In Chatting, Accompanying, and Being Cute, the Heat Is High but the Value Sense Is Unstable."

But this discussion clearly showed that the industry is already moving to the next stage. The guests all agreed that there are three key trends for emotional companionship in 2026: first, moving from functional chatting to deep relationship companionship, with long-term memory and personality consistency becoming important; second, the combination of software and hardware will become mainstream, because only when AI capability has a perceivable physical carrier will the companionship feel more real; third, IP-ization and scenario segmentation will become core competitiveness.

At the same time, everyone did not avoid the risks of this industry. User psychological expectations are hard to control, privacy data security is always a minefield, and once the "novelty dividend" passes, how to maintain user loyalty is a problem no one can avoid.

But more important than these risks is that this session clarified "where high premiums come from." It is not simplyrely on a chatty AI, butrely on thesuperimposed of IP, technology and scenario, then through relationship networks and trust mechanisms, turning one-time interest into long-term payment. Japan being regarded by multiple guests as a priority verification market also shows that this track is not without money in the future—it is just that money will not flow tobroad companionship, but to products that truly build relationships, carry emotions and have clear scenarios.

In the Future, Much Software Will No Longer Be Clicked by Humans, but Run by Agents

The title of the fourth Panel was sharp: "Product Revolution: When Software Is No Longer Designed for Humans."

This session was hosted by Wang Chaochao, partner of Unique Capital. Guests included Li Jinglin, CEO of CometAPI; Song Jian, founder and CEO of NoDesk AI; Xu Jia, Product VP of Pollyreach.ai; and Alex Li, founder of AllyClaw.

These guests come from API aggregation, e-commerce AI Agent, overseas voice AI, and e-commerce growth AI respectively—sounding like different tracks, but by the end of the discussion, everyone was basically saying the same thing: the default user of much software is shifting from "humans" to "Agents."

The software logic of the past, designed around human clicking, human understanding and human operation, is being rewritten. Email marketing, phone customer service, e-commerce operations—work that originally requireda large number of continuous human execution can increasingly be taken over by AI in a full closed loop. Humans do not completely exit, but become goal-setters, decision-makers and correctors.

A deeper change is that team structure will also be rewritten together. Role boundaries will become increasingly blurred—frontend, backend, operations, execution, roles that used to be finely divided, may all compress into the capability of "can you manage AI well." Employees are no longer just executors, but managers and decision-makers of AI. At the same time, risks are also rapidlyamplify: after delegating power to AI, execution speed will far exceed human review; over-reliance on AI will cause human basic capabilities to退化; and models updating faster and faster means talent management and organizational adaptation must start over.

So the conclusion this session finally landed on is not "AI replaces humans," but "AI restructures roles, and humans must redefine their own value." What is truly irreplaceable is still decision-making, emotion and production relationships.

The Second Open Mic, Continuing to Press Down "AI + Scenario"

Four sharers, respectively Xue Shuqi, growth marketing lead of Kuangye Qunxing; Feng Chen, CEO of Guangyu Zhilian; Duan Ran, CEO of Xingqiong Fangzhou; and Chao Guang, co-founder of AceiiLab.

Xue Shuqi introduced Mulan AI, essentially a one-stop AI creative imaging fully automatic film studio, through self-developed Agent automatically generating video workflows, reducing creation costs by 99% and increasing efficiency by 90%—and it is not a laboratory product, but grew out of the real pain points of agency operations. Feng Chen further introduced how Guangyu Zhilian brings AI hardware into cultural tourism scenarios, using parallel robots and AI cultural creative products to solve old problems such as shallow experience and data silos, forming a ToB and ToC closed loop. Duan Ran's Xingqiong Fangzhou focuses on full-scenario wearable AI hardware, emphasizing frictionless随身 Agent, emotional perception, relationship memory and task execution, and特别 emphasizing data security and task confirmation, targeting workplace and financial scenarios with device-cloud collaboration. Chao Guang shared AceiiLab's smart tennis robot, putting AI into sports hardware, using high-speed movement, full-domain perception, AI coaching and gamified courses to solve the problem that traditional ball machines are not good to use and easily gather dust, and has already received good feedback in overseas crowdfunding.

If you put the two open mics together, you will find a very intuitive change: this generation of entrepreneurs is less addicted to the grand rhetoric of "redefining the world," and more focused on "for whom to solve what problem, how to charge, can it be mass-produced, can it sustain repurchase."

People Going Global No Longer Care About "Can Generate Content", but More About "How Not to Involution After Generation"

The theme of the fifth Panel was "Traffic Rules: The Dual-Wheel Drive of AI Content Industrialization and Marketing Automation," hosted by Wei Fangdan, founder and CEO of Baijing Chuhai. Guests included Chen Lingshan, investment and financing director of VidAU; Wang Ming, founder of K2 Lab; Chen Erhang, senior VP of Yixuan Technology; and Tony, co-founder of FOSHO.

This discussion was like giving a class to going-global companies.

Chen Lingshan is more concerned with how AI content can be produced at scale and truly enter thedeployment closed loop; Wang Ming emphasized that the future will be a world of automation and interfacing, and content industrialization is not unilaterally pushed, but the result of technology capability and market demand pulling together; Chen Erhang put the problem more directly—for going-global enterprises, content industrialization and marketing automation are already dual wheels, and neither can be missing; Tony regards content as means of production and automated marketing as productivity, andraised a very eye-catching judgment: in the AI era, content will inflate.

This sentence actually hits the nail on the head. Because AI makes content production efficiency soar, it also means that low-quality, homogeneous content will flood at an unprecedented speed. Everyone can do it, and in the end everyone will内involuted together until there is no differentiation. What to do? The answers given by the guests are also consistent: first, do technology and data differentiation, using the enterprise's own private small models, knowledge bases and data assets, not completely relying on general models; second, do user and scenario differentiation, truly understanding user needs, making more personalized and real-time content; third, master resource and channel differentiation, including scarce resources such as overseas media, influencers anddeployment networks.

So the core conclusion of this session is not "AI makes marketing cheaper," but "AI will make low-level marketing cheaper and also more worthless." What is truly valuable is the combination of differentiated content and precise automation.

Super Individuals Are Hot, but the Real Boundaries Are Also Clarified

The theme of the last Panel was "Super Individuals: OpenClaw Training and the Scientific Method of Raising Shrimp," hosted by Bruce of OpenClaw Asia. Guests included Sang Zhuohao, senior director and AI lead at Focus Media; Zhang Pinpin, founder of Yongbao Zhixu; and Huang Naiyuan, co-founder and COO of Jifan Liuxing.

The good thing about this session is that it did not deify OpenClaw.

Bruce first started from the value of AI going-global products and OpenClaw's breakout, then the three guests gave very specific usage experience. Sang Zhuohao is responsible for KA sales and AI marketing exploration at Focus Media, using OpenClaw as a personal assistant and sales intelligence system; Zhang Pinpin mainly does AI agents and digital human services, connecting OpenClaw to WeChat for personalized replies and community operations; Huang Naiyuan is more concerned about the large-model hallucination problem, using 3D models for constraints, and treating OpenClaw as a personal assistant for execution-type tasks.

The guests' overall view of OpenClaw is rational and even restrained. First, it cannot be simply equated with "generalized Agent." OpenClaw in the narrow sense, as an open-source Agent framework, does have value, but its upper limit is limited and its results are not stable enough. Second, its best application scenarios are still personal assistants, standardized execution tasks, packaged skill launching, and flexible-scenario customer service—efficiency-type work, rather than serious, stable, strong-delivery core business. Finally, regarding the "super individual" thing, everyone is actually not that optimistic. Ordinary people following the trend to "raise shrimp" are often because of anxiety, not because they truly have the ability to turn AI into results. Those who can truly leverage it are still those who understand both AI and their own business.

The most memorable sentence from this discussion is not "OpenClaw is very strong," but another layer of judgment: the market has overestimated its current autonomous long-term execution capability, but underestimated the future self-evolution potential of Agents. For enterprises, the most realistic entry point is still starting from assistant-type work, with the boss personally getting hands-on, first running through small scenarios, then slowly expanding outward.

Looking at the Whole Day, Consensus Is Forming

Looking back at this "2026 Global Unique Awards · Hong Kong AI Summit," you will find that its most valuable place is actually not how explosive a single point of view is, but that between different sessions, an industry consensus after cross-verification is constantly forming. Digital employees are no longer just concepts—they have begun to rewrite enterprise role structures and workflows; AI hardware no longer relies solely on technique-flaunting, but returns to wearing, battery life, scenario and commercialization capability; the emotional companionship track has entered the stage of relationship operation and value realization; Agent-oriented product design is rewriting software logic; content industrialization and marketing automation are becoming infrastructure for going-global growth; and narratives such as "super individuals" have finally begun to be put back into real boundaries for re-examination.

The biggest feeling this summit gives is that the AI industry is bidding farewell to a stage where "anything can be talked about," and entering a stage of "can you actually make it happen." This is of course not as romantic, but it is instead more exciting. Because a technology truly begins to change industry not when everyone is excitedly talking about the future, but when more and more companies start talking about orders, delivery, repurchase, permissions, costs, stability and organizational management.

Returning to the context John Kao mentioned at the opening, it is actually easier to understand why this conference is important. A venue that once helped brands such as ASUS, Anker and BYD go global has once again welcomed new industrial protagonists. It is just that this time, the protagonist is no longer just consumer electronics itself, but AI capability embedded in consumer electronics, enterprise services, cross-border trade and organizational systems.

The next stage of AI is not bigger slogans. It is deeper entry into reality.

If this Hong Kong AI Summit left one clearest conclusion, it is probably this: in 2026, the true competition of AI is no longer just competition of model capability, but competition of industry understanding, scenario landing, organizational collaboration and global connection capability. And this may be the moment when Chinese AI innovation, going to the global market, truly begins to form momentum.

Preview

2026 May 21 Unique Awards · Shenzhen Registration Opens

---

Original publication: https://uniqueresearch.substack.com/p/src-20260418-01html
On-site reading page: https://ffcap.cn/en/research/src-20260418-01html