On December 5-7, the "2025 (23rd) China Entrepreneurs Influence Annual Conference" (formerly known as China Enterprise Leaders Annual Conference) was held in Beijing under the theme "Emergence·Infinite—Co-creating New Forms of Intelligent Business." Fang Han, Chairman and CEO of Kunlun Tech Co.,Ltd., attended and delivered a speech.
When discussing the evaluation criteria for whether enterprises can apply AI agents, Fang Han emphasized that the most critical factor is whether process data can be learned. Once process data can be learned, agents can rapidly improve efficiency.
However, this also raises another issue. He noted that it is widely believed the only universal agent application is office software, as all enterprises use office tools. Beyond office software, no universal agent can address cross-industry challenges. For various industries, vertical agents tailored to specific sectors remain essential.
What do vertical agents rely on? Fang Han pointed out that they depend on industry-specific process data. "If your industry lacks clear process data, agents cannot automate workflows."
Using Kunlun Tech as an example, he explained that the company operates short-video drama businesses overseas, where translation is a key process. A single short drama may need translation into over a dozen languages simultaneously. "First, an agent extracts subtitles, another handles translation, and a third manages dubbing—which must also synchronize with characters' lip movements, requiring yet another AI agent."
Fang Han highlighted that in the entire workflow, all data must be collected before large models can learn how to automate operations efficiently at high speed—a principle applicable to all enterprises.
"You must evaluate whether process data in your business flow can be clearly learned by large models. If not, the agent's workflow won't function. I believe this is the most critical factor for successful agent implementation in industries."
Comments