Robin Li: Future jobs will converge into two categories—one for creation, and one for building trust.

Beating AI News, November 12 — Robin Li made a judgment at the Baidu World Conference: future jobs will not disappear, but will gradually converge into roughly two categories.

One category is called Builder, responsible for creating new things; the other is called Seller, responsible for person-to-person communication, building emotional connections and trust.

In the past, companies hired on the principle of "one radish, one hole" — everyone had a fixed position. In the AI era, this will gradually shift to task-based division of labor, where one person can take on multiple types of work and collaborate with different AI agents, and the boundaries between traditional positions will become increasingly blurred.

Robin Li also addressed a real problem: many enterprises are spending more and more on AI, but revenue and profit have not grown accordingly.

He drew an analogy to the Industrial Revolution. Companies simply equipping employees with AI tools is like replacing old factories with new electric motors without changing the original production methods. To truly improve productivity, the entire business process must be redesigned.

Baidu itself has stumbled along the way. In the early days of large models, many engineers were keen on fine-tuning open-source models, and every department wanted to train its own model. Initial results were good, but once the foundation model was upgraded, the models that had taken time to train quickly fell behind. The company simply mandated that business departments could no longer train models on their own — this was centralized under a dedicated model team.

After agents emerged, something similar happened again. Whichever Agent became popular, every department wanted to build one too. As a result, projects proliferated, resources became increasingly scattered, and the success rate for any single direction actually declined. To address this, Baidu began holding monthly meetings requiring departments to reduce the number of projects pursued simultaneously and concentrate resources on a few key directions. Robin Li said that the less you do, the higher the efficiency and success rate.

He also released the latest data on Baidu's internal use of AI. Over the past year, daily active agents increased from 5,000 to 37,000, and effective tasks completed per week grew from 84,000 to over 1 million. The company has also established 4,300 Workspaces and accumulated over 20,000 Skills that employees and AI can reuse.

These Skills are mainly used to codify employees' work experience so that others and AI agents can directly call upon them. Robin Li believes that only by turning individual experience into company-wide capability can a moat be built that others cannot dig away or buy.

During a conversation with Lenovo CEO Yang Yuanqing that day, they also discussed that AI currently cannot replace a CEO's business trips or building relationships with clients face-to-face. Robin Li agreed, acknowledging that emotional connections between people remain important.