AI Large Model Factory Hosts 2026 AI Industry Ecosystem Conference Today: Industry Leaders Share the Stage to Explore Intelligent Growth and Industrial Symbiosis

AI Large Model Factory Hosts 2026 AI Industry Ecosystem Conference Today: Industry Leaders Share the Stage to Explore Intelligent Growth and Industrial Symbiosis

On September 15, the "2026 AI Industry Ecosystem Conference," hosted by AI Large Model Factory, was held in Beijing.

On September 15, the "2026 AI Industry Ecosystem Conference," hosted by AI Large Model Factory, was held in Beijing. After a two-year hiatus, the AI industry gathered again to collectively face a new question: as intelligence becomes increasingly accessible, where will future competitiveness come from?

The answer to this question may lie in this year's conference theme. As China's vertical industry conference most attuned to AI implementation and industrial ecosystems, this year's event centered on the theme of "Intelligent Growth, Industrial Symbiosis," shifting the focus from technical capabilities to industrial delivery, discussing how AI moves from demos to business operations, and from single-point tools to industry chain collaboration.

This year, the industry's expectations for AI have clearly changed. Frontier AI capabilities continue to advance, Agents are accelerating into real business processes, and scenarios across industry, dining, content, branding, and security are beginning to demand that AI evolve from "being able to answer" to "being deliverable."

As Meng Haowei, co-founder of AI Large Model Factory, stated in his opening speech: "What is truly scarce in the AI industry is not just information, but the ability to turn information into judgment, judgment into collaboration, and collaboration into deliverable results."

At the conference, eight industry leaders shared their progress and unique insights, offering diverse perspectives on the development of the AI industry in 2026: Meng Haowei, co-founder of AI Large Model Factory; Jia Anya, VP of Large Model Products at SenseTime; Li Guang, co-founder of TDengine; Wang Yu, founder & CEO of iPollo and Chairman of the Beijing "Industry-Education-Assessment" AI Skills Ecosystem Lead Unit; Chen Xiaojun, AI Product Lead at Linggandao (Inspiration Island) under Tianxiaxiu; Zhou Qiuye, VP of Shumei (Topsec); Song Xuan, founder of Shaozi AI x Shaozi Classroom and former VP of Xibei; and Li Jie, founding partner of Weture.

Computing Power, Data, Security, and Scenarios: A Full-Chain Dialogue on AI Industrial Implementation

At the conference, eight guests engaged in a high-density, cross-industry discussion around the core propositions of AI industrial implementation.

In his opening speech, Meng Haowei, co-founder of AI Large Model Factory, first expanded the perspective to the scale of the global industry chain.

From a single AI API call in Silicon Valley, to electricity and data centers, to copper mines, railways, and ports in Africa, he argued that today's AI is no longer an isolated software industry. Models, computing power, energy, resources, and infrastructure are interlocking, with upstream and downstream players forming increasingly tight industrial linkages.

On the application side, the changes are equally pronounced: AI is moving from Copilot to "digital employee"—in the past, humans completed the work with AI assisting in one part; next, it will increasingly become the case that humans set goals, AI executes continuously, and humans ultimately verify results and bear responsibility.

He thus left the industry with three questions: What irreplaceable capabilities does my enterprise possess? Where is the next hardest problem to solve in my industry? Can I first turn it into a result that customers are willing to pay for?

Jia Anya, VP of Large Model Products at SenseTime, subsequently offered a sharp judgment: AI addresses "whether something can be done," while humans decide "whether it's worth doing."

In her view, over the past two years, AI has undergone three distinct shifts: from "talking" to "doing," from a single model to a full technology stack, and more importantly, the industry is shifting from "capability-driven" to "value-driven."

As models, products, and even open-source code gradually move toward democratization, enterprises no longer need to find scenarios for AI; instead, they must first identify genuinely high-value problems, then consider what measurable incremental value AI can create. For individuals, this may mean efficiency gains, income growth, and expanded capability boundaries; but for enterprises, individual efficiency gains do not naturally equate to enterprise efficiency gains—enterprises care more about customer acquisition, growth, operational optimization, and ultimately ROI.

Therefore, AI entering enterprises cannot stop at "providing a set of tools." Enterprise data, processes, and business needs are highly personalized, and the real challenge is whether AI's general capabilities can be transformed into a standardized delivery system, which then achieves personalized results through standardized capabilities.

SenseTime's "Raccoon" currently covers over 20 million individual users and serves more than 8,000 enterprises, gradually expanding from data analysis to complex task planning, tool invocation, and enterprise cloud-edge integrated workflows.

She particularly emphasized that enterprises don't want demos; they want genuine end-to-end, measurable value. Tools are only the first step; how to integrate with enterprise data, processes, and specific business scenarios, and truly achieve the precision and usability required for production environments, is the key to AI entering enterprises. At the same time, AI is not meant to replace humans, but to further amplify human experience, judgment, and core competitiveness.

When AI moves from the office to the factory, data becomes the first threshold to cross.

Li Guang, co-founder of TDengine, believes that industrial scenarios face three major disconnects: data, semantics, and intelligence. Addressing this, TDengine forms a complete data foundation from data collection, time-series storage, industrial ontology, and real-time computing to AI agent runtime, and attempts to further close the loop of device anomaly detection, data analysis, judgment, and action.

Meanwhile, through a Skills mechanism, industrial analysis algorithms, industry knowledge, and third-party system integration capabilities are packaged as pluggable modules. Li Guang believes that industrial AI cannot be accomplished by a single enterprise alone; in the future, it will require joint construction by platforms, industry experts, equipment vendors, and ecosystem partners.

Wang Yu, founder & CEO of iPollo and Chairman of the Beijing "Industry-Education-Assessment" AI Skills Ecosystem Lead Unit, shared his thoughts on how to reorganize work after AI acquires capabilities.

In his view, the most fundamental changes of the AI era can be summarized in two things: the replication of capabilities, and the parallelization of time.

In the past, if someone wanted to acquire the capabilities of a lawyer, engineer, or analyst, they either had to learn them themselves or hire professionals; after AI emerged, people can directly invoke encapsulated capabilities and have multiple agents work in parallel. What will truly matter in the future will shift from "what I can do" to "what I can organize and invoke."

Based on this judgment, iPolloWork attempts to build an AI Work platform for next-generation work, enabling new collaborative relationships between Agents and humans, and between Agents and Agents.

Wang Yu believes that while AI can already accomplish most basic work, the "last mile" still requires human judgment, editing, and calibration. The future is not about humans exiting work, but about humans moving further up from large-scale execution to goal definition, organizational capability, and final decision-making.

As AI enters enterprises, it is also changing the way brands connect with users.

Chen Xiaojun, AI Product Lead at Linggandao (Inspiration Island) under Tianxiaxiu, noted three clear changes in brand AI marketing in 2026: cognitive upgrades, budget restructuring, and data infrastructure.

As AI gradually becomes a new entry point for information and decision-making, brand competition is no longer just about "being seen," but about "whether it can be understood and cited by AI, and ultimately enter user decisions." Therefore, GEO, Agents, and AIGC should no longer be just experimental projects within traditional marketing budgets; enterprises need to establish independent budgets and new ROI evaluation standards.

Meanwhile, past marketing investments were largely one-time expenditures, whereas data can accumulate as a reusable enterprise asset. Tianxiaxiu is further AI-izing processes such as influencer selection, ad placement, monitoring, and campaign review through an influencer database, marketing Agents, and an enterprise data middle platform.

Chen Xiaojun believes that the significance of AI is not merely to reduce content production costs, but more importantly, whether it can improve placement effectiveness and ultimate ROI.

But when AI gains more data, tools, and execution permissions, another issue is pushed to the forefront—security.

Zhou Qiuye, VP of Shumei (Topsec), called 2026 the stage where AI security truly needs to reach the enterprise "desktop." As Agents begin to invoke software, operate accounts, access enterprise data, and even execute real actions, the boundaries of traditional cybersecurity are also changing.

He summarized the current problems facing enterprises as "can't see, can't explain, can't defend, can't manage," and proposed that enterprises need to move from "being able to use AI" to "daring to use AI": first, see the risks; second, manage behaviors; then, establish a governance mechanism covering the full lifecycle; and ultimately, implement mandatory standards and compliance systems.

In his view, future AI security is not just a technical issue, but a systems engineering effort combining governance, technology, and operations. As Agents penetrate deeper into business processes, prompt injection, agent poisoning, and trust and permission risks will all become foundational capability building that enterprises must face long-term.

If the preceding discussions focused more on enterprise digital scenarios, Song Xuan, founder of Shaozi AI x Shaozi Classroom and former VP of Xibei, brought the issue directly into a 5-trillion-yuan dining market that is extremely traditional and fragmented.

Shaozi AI has embedded a wealth of dining industry know-how into Agents, currently featuring 25 specialized agents covering scenarios such as strategy, branding, marketing, operations, menus, and delivery, aiming to enable dining practitioners lacking professional experience to achieve near-expert-level working capabilities.

But after truly entering the industry, Song Xuan's conclusion was instead: "In industry applications of AI, AI itself is the least important thing."

Dining enterprises don't care about what advanced models are used; what they truly care about is whether it can bring growth and whether the investment ROI can be calculated. Even if business owners understand AI, there is still significant resistance to truly driving frontline employees to use it. Therefore, the biggest barrier to vertical AI is not just the model, but industry cognition, know-how, key scenarios, and whether one can find people who are genuinely willing and able to put AI to use.

"People are the protagonists of application." Song Xuan believes that as technical capabilities increasingly converge, the true barrier to AI industry applications will return to an understanding of the industry itself.

Finally, Li Jie, founding partner of Weture, turned the focus to the content industry, which is being rapidly restructured by AI.

Compared to traditional industries like dining, AIGC content such as AI short dramas and comic dramas has been built on large model capabilities from the very beginning. AI has significantly reduced content production costs, shortened production cycles, and further standardized the process from creativity and production to distribution. Content supply has thus expanded rapidly, but the new problem that follows is: as "producing content" becomes cheaper and cheaper, where will the next stage of competitiveness lie?

Li Jie's judgment is—OPC and interactive content.

On one hand, AI will continuously amplify the productivity and creative capabilities of individual creators, spawning more "super individuals"; on the other hand, content will further shift from past one-way viewing to real-time interaction. Users will no longer just watch a fixed story, but will be able to influence the plot direction through their own feedback, truly achieving a thousand stories for a thousand people.

In this vision, AI changes not just content production efficiency, but the very form of content: from "I produce, you watch" to a two-way co-creation of "user participation, AI real-time generation."

The Next Journey of Intelligent Growth and Industrial Symbiosis

In 2026, AI capability evolution has reached the point where Agents enter business processes, embodied intelligence touches the physical world, and AI-native applications restructure organizational processes. Industry competition has also shifted from "who can generate better" to "whose ecosystem runs through first."

Looking across the entire conference, while the guests' sharing spanned different domains, they gradually converged into a clear set of core consensuses. AI value assessment is shifting from "whether it can generate" to "whether it can deliver."

Discussions across all parties generally landed on tasks, processes, costs, and results, rather than simply showcasing AI capabilities; the foundation and scenarios need two-way alignment—models, computing power, data, and security form one side of the infrastructure, while industry, branding, security, dining, and short dramas form the other side of real scenarios; without either end, large-scale implementation cannot succeed.

For AI Large Model Factory, what this conference hopes to connect is not just seven guests and seven speeches, but to place model vendors, infrastructure, AI platforms, security enterprises, and industry practitioners on the same industry map, to find where technology truly goes.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.