Zhiyuan Proposes Worken Effective Intelligent Work Hours, Robot Billing Model Still Awaits Validation

"Thought in the digital world is measured by Tokens, so what should labor in the physical world be measured by?" On October 29, at the AgiBot Innovation Day, Peng Zhihui, co-founder, president and CTO of AgiBot, shared their thinking on this question and proposed the concept of "Worken," namely effective intelligent work hours, to measure the actual output of autonomous robot operations.

In an interview with media including Yicai, Peng Zhihui said that in the past the market focused on robots' performance parameters and flashy demos, but as the industry develops further, robots' "timesheets" will be increasingly taken into account by customers.

At present, AgiBot has already calculated "Worken" for robots in multiple scenarios and has even applied for a trademark for this new concept. But to redefine the value of robots, embodied intelligence companies must first answer whether "customers are willing to pay."

"Worken" billing is still in the exploratory stage

AgiBot's revenue grew from 300,000 yuan in 2023 to 60 million yuan in 2024, exceeded 1 billion yuan in 2025, and reached nearly 2 billion yuan in the first half of this year.

On the product side, according to statistics from third-party institutions, in the first half of this year AgiBot ranked first globally with shipments of 9,700 units and a market share of more than 43%. Peng Zhihui disclosed that in September this year, AgiBot's 20,000th robot officially rolled off the production line.

At the main forum, AgiBot launched the new-generation embodied foundation model GO-3 Preview, the embodied intelligence native multimodal interaction large model WITA Omni1.0, as well as a new-generation dexterous hand product family, the safety-standard heavy-duty embodied robot Genie G2 Max, and more.

Beyond technology, AgiBot also explained the concept of "Worken" and progress in deployment.

According to Peng Zhihui, the formula for Worken is "Worken = number of active robots × autonomous working hours × output efficiency ratio." Specifically, the number of active robots refers to the scale of robots actually working on the job, not robots sitting in warehouses; autonomous working hours require robots to be autonomous agents that do not depend on humans; and the output efficiency ratio is "how much human value the robot's labor can create."

"What Worken wants to answer is how much sustainable and effective intelligent labor robots can actually contribute in real environments, including the robots' capabilities, reliability, and actual output for customers," Peng Zhihui said, defining Worken as "the unit of productivity in the physical AI era."

Worken calculations have already been applied in some of AgiBot's deployment scenarios, such as loading and unloading at the Nanchang Longcheer production line, security guidance at Guangzhou Metro, and entertainment performances at Zhuhai Chimelong. Among them, the 300 robots in the Chimelong entertainment performances had the largest Worken number, reaching 33,480 Worken.

Wang Chuang, partner and senior vice president of AgiBot, said in an interview that the concept of Worken involves the future supply method of productivity. He introduced that AgiBot currently mainly sells full-stack capabilities, including the robot body and subsequent operation and maintenance services, as well as leasing. As technology continues to develop and robots truly become capable of doing work, future payment methods may change. For example, having robots complete fixed tasks could be achieved by purchasing package plans and billing according to Worken consumption. He mentioned, "This is our vision; at present it has just started."

Worken, this calculation and measurement system, is currently carried out internally at AgiBot and has not yet formed a business model. Whether it can ultimately be recognized by the industry and customers and develop into something that can be clearly priced like Tokens remains unknown.

The reporter noted that the on-site large screen showed a small "TM (trademark)" note in the upper right corner of "Worken." AgiBot said the team is currently applying for the "Worken" trademark. A lawyer analyzed that this is a business strategy; once approved, it could block peers from using it.

Production volume signal is not equal to productivity

AgiBot is not the only participant in defining the value of robots.

In October this year, at an industry summit forum, Tsinghua University professor Lu Jiwen mentioned the physical AI economy, introduced the concept of "Woten," and said that "Woten gives the physical world, for the first time, a measurable commercial unit similar to Token."

Whichever concept it is, the common reality they face remains that robots have not yet been deployed on a large scale.

According to data from the Ministry of Industry and Information Technology, China's full-machine output of humanoid robots is expected to exceed 100,000 units this year. Peng Zhihui said this is a signal of industrial development, but at the same time it must be recognized that how many robots are produced does not equal how much productivity is added in the real world. He believes that to solve the problem of application scenarios, it is necessary to make embodied large model capabilities sufficiently generalizable, make the robot body sufficiently reliable, and enable the comprehensive software-hardware system capabilities to reach the threshold for entering the deployment state.

Wang Chuang mentioned that humanoid robots moving from the laboratory to real deployment will face six major challenges: battery life and system stability, large-scale production and delivery, network and device-cloud collaborative systems, model evolution and content updates, active and passive safety and ease of use, and operations and maintenance.

Take the application of AgiBot robots at the Chimelong Spaceship park as an example. Around late September this year, more than 300 humanoid robots and quadruped robot dogs went to work at the park, providing visitors with entertainment performances, science popularization and study tours, guide and shopping guidance, service retail stations, intelligent companion tours, and other services at more than 100 locations. According to the introduction, these deployed robots brought growth in visitor traffic to the park.

"The first challenge to face is how to manage so many robots, such as solving issues like running out of power, damage, and program updates," Wang Chuang also said. He added that in actual operations they also encountered problems outside expectations; when children interact with robots they may hug them, which requires robots to respond to emergencies and thus ensure safety.

Large-scale robot deployment also requires a considerable amount of manpower. According to Wang Chuang, at the beginning of deployment one staff member managed one robot, and later it gradually developed to one staff member managing three robots. Their goal is to reach a robot-to-human ratio of 10:1 by the end of this year.

Deployment costs are clearly a key factor limiting the large-scale rollout of robots.

"We have the ability to manufacture tens of thousands of robots a year, but it depends on whether customers are willing to pay and when they will pay," Peng Zhihui believes. Robot deployment should not be just a POC; it also involves long-term equipment management, software iteration, and so on. Only when operating costs continue to decline as scale rises and finally reach the ROI break-even point will customers be willing to choose AgiBot's products, and only then will there be an opportunity to expand the scale of robot deployment. "If selling a robot also requires providing a 'nanny,' then that day will never come."

He also said that what AgiBot cares more about at present is whether it can form a reusable technology foundation and whether customers can reduce integration and maintenance costs after using its products and solutions. "Rolling 10,000 units off the production line is only a victory for manufacturing; 10,000 units being able to take up posts is the victory of physical AI."