Beijing Dongcheng, November 2025 — Ting Yang, CEO of SOIN AI and Vice President of Sointech (stock code: 871539), was recently invited to the tech interview program "G-Talk" for an in-depth dialogue with host Garfi, sharing her transformation from tech executive to AI entrepreneur and her forward-looking thinking on the commercialization path of Agentic AI. The interview video has been released on major platforms, attracting wide industry attention.
About G-Talk: A Deep Intellectual Arena Focused on Technology Innovation
"G-Talk" is an in-depth dialogue program on technology and innovation hosted by veteran tech media professional and investor Garfi, with the core philosophy of "talking with the people who change the world about the things that change the world." The program insists on the style of "all substance, no filler," dedicated to tracing the sources of disruptive technology, listening to the most authentic journeys of top entrepreneurs, and locking in the next technological future for viewers. Program slogan: Talk with the people who change the world about the things that change the world. This interview represents G-Talk's important attention to an enterprise-grade application explorer in China's Agentic AI field.
Technology Commercialization Philosophy: Returning Technology to Business First
In the interview, Ting Yang summarized her core understanding formed across the Microsoft MRPP ecosystem, Kunlun Tech and other tech unicorns as "returning technology to its business role." She pointed out that technical capability is the foundation, but what truly converts into sustained value is a deep understanding of industry logic and processes. "The market is not a passive testing ground for technology, but a guiding force that continuously shapes technology direction," Ting Yang emphasized. In leading Tiangong AI's closed loop from technology to commercialization, she deeply experienced that the product R&D paradigm has fundamentally changed in the AI era — market feedback must flow directly back into product and model design decisions. What enterprises really need are agents that can read industry documents, automatically complete processes, and work within existing PLM/ERP systems — these needs directly determine technology priorities.
Entrepreneurial Transformation: The Role Evolution from Promoter to Owner
Speaking of the shift from top-tier company executive to running all of SOIN AI's business, Ting Yang defined it as a switch from "promoter" to "owner." The trigger was the sense of mission to turn AIGC into enterprise-grade capability, and the desire to bring the ecosystem and governance experience accumulated at large companies into an environment where fast trial and error is possible.
She admits the biggest challenge is switching identity and rhythm: both advancing product deployment faster and more pragmatically, and balancing corporate governance, team culture and resource allocation. To this end, SOIN AI especially emphasizes the "productization + ecosystem" path, converting strategy into replicable engineering capability — both quickly serving customers and accumulating into long-term company assets.
SOIN AI's advantage lies in combining agility with a global view: continuously running in fast small steps driven by market value, agilely combining market value with rapid deployment — an efficient management model for new tech companies in the AI era.
The AWE Engine: Building a Self-Evolving Agent Operating System
Ting Yang highlighted SOIN AI's proprietary Agentic Workflow Engine (AWE). Through four technical characteristics, it realizes the vision of an "open, self-evolving agent ecosystem":
Programmable workflows and modular agents: decomposing enterprise processes into composable modules, with each agent having clear inputs, outputs and role definitions — combining business scenarios like building blocks;
Multimodal heterogeneous data access: non-intrusively processing PDF, CAD, XML, drawings and other multilingual content, converting dormant data into structured knowledge assets. Especially noteworthy is its CUA-mode web retrieval capability, which simulates human browser operation to access the internet — for example, completing cross-border e-commerce product selection on 1688 without any API;
Knowledge graph and reinforcement learning loop: combining RAG retrieval optimization with RLHF for continuous agent improvement, forming a feedback optimization loop of "data — knowledge — decisions";
Cloud-native and edge-cloud collaboration: supporting cloud deployment and proprietary enterprise deployment, balancing scalability with data autonomy.
"AWE is not a mere model scheduler, but an enterprise-grade agent operating system," Ting Yang concluded. "It converts model capability into orchestratable, evolvable business capability."
Industry Cultivation: Dual Drive of Advanced Manufacturing and AI Video Generation
SOIN AI chose the "AI to B" path, focusing on traditional industries such as advanced manufacturing, education publishing and marketing. Ting Yang explained that these fields have clear, stable processes constrained by inefficiency — precisely fertile ground for Agentic AI's structured value.
In advanced manufacturing, SOIN AI, as an AI partner in multiple global leading industrial software ecosystems, provides intelligent R&D assistants with component knowledge Q&A, intelligent classification, multi-format document parsing and more, improving modeling efficiency by 80%, query speed by 75%, and reducing engineering error risk by 40%.
In AI video generation, building on the successful practice of the domestic toB product NarrativeVision Pro, SOIN AI launched the one-stop commercial platform CineVerse for the global market. Through its proprietary intelligent scheduling engine, integrating the world's top multimodal model services, it achieves full-process automation from text to final cut — and has shortened PLM agent projects that traditionally require 10 person-months to delivery by a 3-person team in 3 weeks.
International Layout: Balancing Technology Localization and a Trust Ecosystem
On the company's "international expansion" strategy, Ting Yang said that when expanding in Singapore, the Middle East and other markets, trust and cooperation ecosystems often determine success earlier than technical details. Technology localization (language, regulations, data compliance) is necessary work, but the first threshold for entering a new market is understanding local customers' procurement decision chains, establishing cooperation with local partners, and gaining industry endorsement.
SOIN AI adopts parallel strategies of "localization + joint delivery": technically ensuring a configurable industry adaptation layer; commercially first co-building showcase projects and industry validation with local partners, gradually converting localized cases into replicable productized capabilities. Its business now covers China and multiple overseas markets.
Women's Leadership: Curiosity, Product Thinking and Responsibility
As a woman leader in AI, Ting Yang values three qualities most in team building: curiosity, product thinking and responsibility. She believes curiosity drives the team to actively learn the latest technologies and convert them into product value; product thinking ensures technical work points to user value; responsibility drives the team to be accountable for customer results, forming a culture of continuous delivery.
Inside SOIN AI, every member is expected to have a PMM profile (Product, Marketing, Manager). This composite talent development strategy reduces team friction and improves execution. 80% of the company's requirement documents are AI-generated, product managers have all transformed into agent architects, 80% of R&D coding is done by AI, and each developer leads multiple AI engineers.
The AI-Native Enterprise Vision: A New Organizational Form of Human-Agent Collaboration
Looking ahead 5–10 years, the "AI-native enterprise" Ting Yang describes is not machines replacing people, but a new organizational collaboration model of "humans + agents." Business is driven by capability-based modules; people handle judgment-based, higher-level decisions; organizational structure shifts from functions to capability centers with blurring department boundaries; workflows are threaded by agents, forming a self-driving closed loop.
"Everyone can innovate with AI," Ting Yang envisioned. "Work shifts from 'doing things' to 'collaborating with agents and supervising agents'; management metrics shift from hours and output to business capability and human-agent collaboration efficiency."
This G-Talk interview fully displayed Ting Yang's strategic thinking and practical achievements as an AI company leader, and showed the world the technical strength and international ambitions of China's enterprise-grade Agentic AI applications. With the AWE engine's continuous iteration and the deepening expansion of its industry ecosystem, SOIN AI is steadily advancing toward its vision — becoming the value integrator of the AI video era and a reliable foundation for industrial intelligent upgrading.