Qianfeng Interview with Focus Technology's Lu Yi: Foreign Trade B2B Moves from 'Page Browsing' to 'Organizing Actions Around Procurement Goals'

【Ebrun Original】Lu Yi is Vice President of Focus Technology and General Manager of its AI Center. Focus Technology continues to explore the deep integration of AI technology with foreign trade scenarios. Based on the global trade services of Made-in-China.com, it is building an intelligent service system for both suppliers and buyers, advancing the two-way intelligent layout of AI Maco and SourcingAI, and exploring a new model of AI-driven foreign trade growth. In November 2026, AI Maco will be upgraded into a dedicated AI workspace for foreign trade enterprises, while SourcingAI 3.0 is positioned as a global procurement AI partner, further improving intelligent services for both suppliers and buyers. In this Qianfeng Award interview, Lu Yi shared Focus Technology's practices and judgments in the foreign trade B2B direction, centered on the modal evolution brought by AI and going global.

Ebrun: This year's Qianfeng Award takes 'Modal Evolution Brought by AI and Going Global' as its core theme, focusing on the deep restructuring of industrial capabilities. In the context of the foreign trade B2B services industry, how do you understand this round of industrial modal evolution jointly driven by AI and globalization? What core strategic layouts has Focus Technology made in AI capability building and global expansion? Could you specifically introduce the company's positioning and business model?

Lu Yi: In foreign trade, we observe three interconnected changes.

First, procurement entry points are more diverse. Buyers search, compare, and verify suppliers across search engines, AI, B2B platforms, independent sites, social media, and offline trade shows. Suppliers need to present their professional capabilities accurately and consistently across different channels and enter the buyer's decision-making process.

Second, AI is beginning to participate in actual business. It can help buyers clarify needs, and it can also help sellers understand inquiries, combine enterprise knowledge, and form reply and follow-up suggestions. Enterprises' expectations of AI are shifting from completing a single piece of content production to continuously advancing a business task.

Third, the value of platforms is further extending toward trusted data, transaction capabilities, and fulfillment services. Even if buyers initiate procurement through an AI entry point, they still need real product information, reliable suppliers, trustworthy qualification information, and payment, delivery, and after-sales support.

Around these changes, Focus Technology is advancing a two-way intelligent layout. On the supplier side, AI Maco will soon be upgraded into a dedicated AI workspace for foreign trade enterprises, adhering to the principle of 'understanding foreign trade, and understanding you better,' enabling enterprises to use it out of the box and letting their operations grow on AI. On the buyer side, SourcingAI 3.0 was released on October 16, 2026, positioned as a global procurement AI partner that can help buyers start from original needs, from intelligent retrieval based on a paragraph of text or a drawing to business discovery and industry insights, and then to AI agents assisting business negotiations.

Our business is based on platform membership services, AI subscriptions, and related value-added services, continuously increasing the depth of service for customer operations. The core of foreign trade has always been the matching of excellent supply with precise demand, and AI is changing the way this matching is achieved.

Ebrun: You mentioned that AI Maco will soon be upgraded into a dedicated AI workspace for foreign trade enterprises, and SourcingAI 3.0 is positioned as a global procurement AI partner. Who are these two product lines respectively aimed at? Why is 'two-way intelligence' needed instead of serving only one end?

Lu Yi: On the supplier side, AI Maco helps foreign trade enterprises understand inquiries, combine enterprise knowledge, form reply and follow-up suggestions, and will be upgraded into a dedicated AI workspace so that enterprises can use it out of the box. On the buyer side, SourcingAI 3.0 is positioned as a global procurement AI partner, helping buyers start from original needs, from intelligent retrieval based on a paragraph of text or a drawing to business discovery and industry insights, and then to AI agents assisting business negotiations. The core of foreign trade has always been the matching of excellent supply with precise demand. If only one end is served, matching efficiency cannot improve.

Ebrun: At present, in which core business links has your company implemented AI? What distinctive AI or digital-intelligence competitiveness system has been formed?

Lu Yi: Our applications mainly revolve around marketing customer acquisition, business conversion, and business management, and serve both suppliers and buyers.

In the marketing customer acquisition link, AI Maco supports product publishing, graphic and video production, social media marketing, and other work. Made-in-China.com is also advancing generative engine optimization, improving content quality, information structure, and crawl conditions, and increasing the chances that high-quality supplier information is discovered, understood, and cited by AI.

In the business conversion link, AI Maco provides round-the-clock multilingual reception and, combined with buyer profiles, enterprise materials, and industry knowledge, helps business personnel analyze inquiries and form professional replies and quotation suggestions. We focus not only on response speed but also on whether replies are accurate and consistent with the enterprise's actual supply capabilities.

In the business management link, we integrate enterprise knowledge, buyer information, channels, and tools. The desktop end will carry this direction, allowing AI to be more deeply integrated into the daily work of business personnel.

On the buyer side, SourcingAI 3.0 covers product selection market analysis, demand parsing, sourcing matching, supplier insights, AI agent negotiations, and fulfillment coordination. Among them, the demand clarification capability helps novice buyers sort out vague or contradictory requirements, while vertical industry terminology support helps professional buyers conduct more refined searches.

In value chain coordination, we focus on connecting procurement demand, supply capabilities, and trade services. Payment, logistics, quality inspection, and other links are still handled by the corresponding service systems, and AI helps improve information processing and process connection efficiency. This forms Focus Technology's distinctive feature: based on real trade scenarios, AI capabilities and platform services work together in the business process.

Ebrun: You mentioned that Made-in-China.com is advancing generative engine optimization to increase the chances that high-quality supplier information is discovered, understood, and cited by AI. How exactly is GEO implemented in Focus Technology's business? How is it different from traditional SEO?

Lu Yi: We place more emphasis on improving the authenticity, professionalism, structural clarity, and accessibility of content, so that high-quality supply information is more easily discovered and understood by AI. Enterprises need to clearly explain their real advantages and enable information across different channels to corroborate one another. This is different from traditional SEO: traditional SEO focuses more on keywords and rankings, while GEO focuses more on whether AI can understand, cite, and accurately present an enterprise's professional capabilities.

Ebrun: Could you share one of the most representative 'AI value chain' innovation practice cases? What common industry pain points did it solve, and what perceivable and quantifiable value did it create?

Lu Yi: A relatively representative case is how AI Maco helped an enterprise solve the problem of capturing cross-time-zone business opportunities.

Wuxi Baosiyou International Trade Co., Ltd., which mainly deals in mini excavators, previously faced the problem of being unable to reply in time to nighttime inquiries from European customers. After enabling AI Maco's nighttime reception, an inquiry sent by a German buyer at 2 a.m. received a timely response, and after manual follow-up, factory visits, and other steps, a customized order was ultimately reached.

This case shows that the business value of AI can be reflected in a specific chain: promptly capturing demand, completing initial communication, leaving effective business opportunity information, and then having business personnel advance cooperation.

The case of Xiamen Weier Jinfu Industry and Trade Co., Ltd. reflects the importance of professional knowledge. The enterprise combined product materials, business scripts, and design terminology into the application, completed 38 rounds of automated professional dialogue, and increased the number of customers handled per person from 20 to 40. Only after professional knowledge enters the AI work process can enterprises more easily convert response efficiency into effective service capability.

In terms of product scale, AI Maco has accumulated more than 25,000 purchased memberships. Platform operating data also show that in August 2026, the buyer business opportunity reply rate increased by more than 50% year on year. This reflects improvements in the relevant business links, but customer individual cases and overall platform indicators need to be understood separately, and all changes cannot be simply attributed to a single AI function.

Ebrun: You mentioned that when evaluating input-output, you will also consider model invocation, R&D, knowledge maintenance, system connection, and service delivery costs, and that calculating value by token consumption and computing power cost scale will not be very high. How should this judgment be understood? How does Focus Technology control AI costs?

Lu Yi: For customers, the most meaningful indicator is how much comprehensive cost is required to complete an effective job, whether AI's business professional capabilities support commercial-grade needs, whether service quality is stable, and whether business personnel can therefore take on more and more valuable work. In addition to matching the capability, speed, and cost requirements of different tasks through model routing, invocation statistics, and quota management, when evaluating input-output we also consider model invocation, R&D, knowledge maintenance, system connection, and service delivery costs. Token consumption and computing power consumption will not be very high when value is calculated at scale. For customers of different scales and different depths of application, the gap in computing power will also be relatively large.

Ebrun: In advancing the large-scale implementation of AI, what supporting role do data elements play? Does Focus Technology focus more on deep cultivation of its own industry data, or on advancing cross-entity and cross-industry data collaboration? What are the core challenges, and how do you respond?

Lu Yi: We believe that the value of data is first reflected in enabling AI to understand the business, and second in supporting reliable action.

Focus Technology's data foundation comes from long-term foreign trade platform services, including product information, enterprise information, procurement demand, industry knowledge, and business behavior. For enterprises to actually apply AI, they also need to further add their own product materials, quotation logic, customer history, and business experience.

In the latest AI Maco product approach, there are three important types of knowledge assets: the buyer profile library, the industry expert library, and the enterprise knowledge library. Buyer profiles help understand what this buyer has communicated about before and what needs remain unresolved; industry expert knowledge supports professional judgment; the enterprise knowledge library clarifies the enterprise's own product characteristics, capability boundaries, and expression standards.

For example, the same set of enterprise knowledge can support product publishing on B2B platforms, selection guides on independent sites, social media content, and trade show invitation emails. In this way, enterprise information across different channels can remain consistent, and the work of repeatedly organizing materials is reduced.

Therefore, we simultaneously advance deep cultivation of our own industry data and authorized cross-system collaboration. The main challenges lie in scattered materials, content updates, permission boundaries, and connections between different systems. The response focus is to give data a clear source, update mechanism, and usage permissions, and to test through specific tasks whether the data is truly usable.

The value of enterprise context is to let AI know what kind of judgment this enterprise should make under what circumstances.

Ebrun: You summarized AI Maco's knowledge assets into three categories: 'buyer profile library, industry expert library, and enterprise knowledge library.' What problems do these three types of knowledge respectively solve? Why does the enterprise knowledge library need to clarify 'capability boundaries and expression standards'?

Lu Yi: Buyer profiles help understand what this buyer has communicated about before and what needs remain unresolved; industry expert knowledge supports professional judgment; the enterprise knowledge library clarifies the enterprise's own product characteristics, capability boundaries, and expression standards. The reason the enterprise knowledge library needs to clarify capability boundaries and expression standards is that AI needs to know what this enterprise can do, what it cannot do, and how it should express things; otherwise, no matter how fast the reply is, it may be inaccurate or inconsistent with the enterprise's actual supply capabilities.

Ebrun: How do you view industrial agents rewriting B2B? Is there a long-term development necessity for vertical large models? Where is the AI moat of industrial internet companies?

Lu Yi: Industrial agents will gradually move B2B from primarily page browsing and manual search to organizing information and actions around procurement goals.

After a buyer proposes a procurement goal, AI can help complete requirements, screen suppliers, compare solutions, and organize communication results; on the supplier side, AI can combine enterprise materials to answer professional questions, clarify supply capabilities, and assist follow-up. In the future, some standardized information interactions may be completed by agents on both sides, with people participating more in important judgments, negotiations, and responsibility-taking.

This change also increases the importance of trustworthy information. AI makes content production easier, but buyers still need to judge whether an enterprise is real, whether its capabilities match, and whether promises can be fulfilled. AI can generate information at scale, but business trust still needs to be built on the authenticity of information and reliable fulfillment.

The moat as we understand it includes high-quality industry data, task execution capabilities in real business, and long-accumulated trust and service systems. Supplier review, qualification verification, quality inspection, and transaction services are all important foundations for AI to connect with the real business world.

As for vertical large models, industry-specific capabilities have long-term value, but whether to train a model from scratch depends on the specific task and input-output. Industry knowledge, enterprise context, professional tools, and necessary model training and evaluation should be combined around business outcomes. The ultimate measure is whether industry tasks can be completed stably.

Ebrun: In terms of overseas layout, which core markets has Focus Technology currently laid out? How does it carry out localization according to the local environment?

Lu Yi: Made-in-China.com provides services to global buyers, and the layout in different markets is adjusted according to local procurement needs and industry characteristics.

In North America, the group has accumulated localized service capabilities through businesses such as inQbrands. The Jiangsu Premium Products North America Cross-border Exhibition and Sales Center, which opened this year in the Los Angeles Cross-border Park, combines product display, brand promotion, and local operations to provide Chinese enterprises with opportunities for continuous display and contact with local customers.

In more markets around the world, Made-in-China.com helps suppliers understand local needs through professional exhibitions, market inspections, and procurement matchmaking. For example, during the Malaysia Food and Beverage Exhibition, the platform organized relevant equipment enterprises to participate and coordinated logistics, booth setup, and business matchmaking.

AI applications also need to adapt to different buyers. Buyers new to a category need help understanding products and usage conditions; professional buyers place more emphasis on accurate industry terminology, technical parameters, and supply capabilities. SourcingAI's demand clarification and refined search are precisely built around these differences.

Our understanding of localization is to make supply capabilities fit the actual needs of specific markets. Language, content, and communication efficiency can be improved with AI assistance, but local professional services, customer relationships, and delivery capabilities still need continuous building.

Ebrun: You mentioned that localization means 'making supply capabilities fit the actual needs of specific markets,' and that language and content can be assisted by AI, but local professional services, customer relationships, and delivery capabilities still need continuous building. What does this judgment mean for Focus Technology's overseas investment?

Lu Yi: Language, content, and communication efficiency can be improved with AI assistance, but local professional services, customer relationships, and delivery capabilities still need continuous building. This means our overseas investment cannot be limited to language translation and content adaptation; we also need to build local service capabilities, customer relationships, and delivery networks in key markets. AI can reduce information processing costs, but it cannot replace localized professional services and trust building.

Ebrun: In the process of expanding overseas, what core support do AI and digital capabilities provide for cross-border business?

Lu Yi: We are paying increasing attention to multi-channel coordination, because buyers often verify the same supplier through multiple sources.

Platforms can provide supplier information and trust support, independent sites help enterprises fully display their brand and professional capabilities, social media continuously presents enterprise developments, and trade shows provide opportunities for face-to-face communication and physical verification. These channels can jointly serve one procurement decision path.

AI undertakes two aspects of work in this. On the one hand, it helps enterprises form content suitable for different channels based on the same set of product and business knowledge. For example, platforms need clear product parameters, independent sites need selection guides and FAQs, social media need application scenarios and updates, and trade show invitations need to be combined with specific customers. Channel expressions can differ, but facts and capability standards should remain consistent. On the other hand, it helps enterprises continuously operate customers who enter these channels. Through demand identification, customer profiles, and follow-up suggestions, it reduces information omissions when business personnel switch between different tools.

The upgrade of AI Maco's desktop end is precisely centered on this need. We hope to gradually connect platforms, independent sites, social media, email, and the enterprise's existing business tools, so that enterprise knowledge and business processes can be used continuously.

Ebrun: To coordinate the implementation of AI capabilities, the upgrading of the domestic value chain, and overseas business expansion, what key adjustments has Focus Technology made in organizational structure or governance mechanisms?

Lu Yi: As AI applications deepen into business, organizational coordination will become an important component of product effectiveness.

At Focus Technology, we are advancing coordination around the public capability building of the AI Center and the scenario practices of business teams. Public capabilities need to be reused, while specific applications need to be close to customers, keeping technology iteration and frontline feedback connected. The upgrade of AI Maco is also integrating capabilities in platform operations and order conversion to form more continuous services for customers.

In the product, we can also see organizational management needs. For example, AI reception has extended from individual use to unified team configuration: the main account centrally sets reception rules, and sub-accounts flexibly enable them according to actual conditions, helping teams form relatively consistent service standards.

The same approach also applies to ourselves. Product, R&D, business, and operations need to jointly clarify task goals, knowledge sources, execution permissions, and acceptance standards. Evaluation should look not only at answer quality but also at whether tasks are completed, whether business rules are followed, and whether cost and stability are acceptable.

In the future, excellent enterprise experience can also be gradually organized into knowledge and reusable skills so that team members can use them together. The role of the organization is to turn individual experience into continuously effective work capability.

Ebrun: You mentioned that AI reception has extended from individual use to unified team configuration, with the main account centrally setting reception rules and sub-accounts flexibly enabling them. What does this change mean for organizational management?

Lu Yi: This change means that as AI applications deepen into business, organizational coordination will become an important component of product effectiveness. AI reception has extended from individual use to unified team configuration, with the main account centrally setting reception rules and sub-accounts flexibly enabling them according to actual conditions, helping teams form relatively consistent service standards. The same approach also applies to ourselves: product, R&D, business, and operations need to jointly clarify task goals, knowledge sources, execution permissions, and acceptance standards. The role of the organization is to turn individual experience into continuously effective work capability.

Ebrun: If you were to give one core piece of advice to enterprises hoping to build both AI competitiveness and globalization capabilities, what would it be?

Lu Yi: My advice is to focus on a real customer problem, accumulate the enterprise's professional capabilities, and then let AI help you deliver stably.

First clarify whom you serve, what problem you solve, and why customers are willing to pay. Then organize product materials, service experience, customer requirements, and judgment standards so that AI can understand the enterprise and work within clear permissions and processes.

Start from a scenario where results can be verified, such as improving inquiry response quality, reducing communication costs, or improving customer follow-up. Through actual results, gradually form replicable capabilities.

For enterprises going global, I have always emphasized: in this era, enterprises must make products better, deepen services, and strengthen professional capabilities. Platforms, independent sites, social media, and offline trade shows can all serve this goal.

Technology and entry points will continue to change, but the matching of excellent supply with precise demand will always be the core of foreign trade. We build AI capabilities always around this goal.

As an annual bellwether in the TO B field, the '2026 Eighth Ebrun Industrial Internet Annual Conference and First Industrial Agent Conference' will be held in Beijing on November 20, 2026. The theme is set as 'Modal Evolution Brought by AI and Going Global,' focusing on in-depth discussion of frontier topics such as the implementation of industrial agents, value reconstruction of AI supply chains, data elements and the token economy, and the new ecosystem of industrial going global. The conference will bring together academic authorities, benchmark central state-owned enterprises, industrial leaders, investment institutions, and service ecosystem partners, gather core industrial forces, break down the paths of industrial digital-intelligent transformation, and share frontier practical experience. At the same time, the annual industrial internet report and benchmark cases will be released, and a decision-maker closed-door meeting and the 'Qianfeng Award' ceremony will be held.

This article was first published on the official website of Ebrun.