On July 28, 360 Group founder Zhou Hongyi officially launched the new-generation enterprise AI agent work platform NanoWork in Beijing. This product is designed for business owners, entrepreneurs, one-person companies, and corporate employees, positioning itself as an AI agent work platform "built for enterprise AI transformation."

At the launch event, Zhou Hongyi announced that the first batch of users would each receive a 100 million token trial quota. Through professional training and coaching services, he aims to help 1,000 small businesses kickstart their AI deployment and boost operational efficiency.

Zhou Hongyi revealed that before the official release of NanoWork, 360 first used itself as a "testing ground." Various AI agents, including OpenClaw, were deployed across over 1,000 real business scenarios within the company. More than 56,000 user feedback entries were collected, and 166 versions were optimized and iterated over five months.

NanoWork had previously been unveiled at the ISC conference in June 2026. At that time, Zhou Hongyi stated that 360 had deeply optimized the product, aiming to upgrade it from a "geek tool" to an "AI expert platform capable of entering real workflows."

With this official launch, Zhou Hongyi said NanoWork is "not a simple wrapper around a single AI agent." Instead, the goal is to conceal complex technical capabilities, turning them into a productivity tool that users can directly employ, while also embedding native security features.

According to the launch presentation, NanoWork builds a five-pillar capability system: a multi-agent engine, a multi-model foundation, a cloud-based office, multiple working modes, and multiple experts. Users do not need to study models or agent frameworks. They only need to set a goal, and the platform will organize different AI experts to collaborate on tasks, matching the appropriate model and execution method.

In terms of application scenarios, business owners can use NanoWork to assist with business analysis, growth tracking, and progress management. Entrepreneurs and one-person companies can use it for opportunity discovery, product validation, and marketing lead generation. Corporate employees can improve efficiency in research, planning, and project delivery.

Security was a key topic at the event. Zhou Hongyi stated during the launch: "When a large language model makes a mistake, it says the wrong thing. When an AI agent makes a mistake, it does the wrong thing."

To address the new security challenges posed by AI agents executing tasks, 360 has adopted "native security, built-in from the factory" as a design principle for NanoWork. This approach reduces the risks enterprises face when using AI through mechanisms such as cloud isolation, permission control, and data protection.

The launch of NanoWork comes at a time of rapid expansion in the office AI agent sector.

An analysis report from Analysys shows that in June 2026, the total visits to 17 mainstream desktop-based AI-native office agent platforms in China exceeded 60 million, with Tencent's WorkBuddy leading at 20.97 million visits. Data from iiMedia Research indicates that the Chinese AI agent market size reached 80.4 billion yuan in 2025, a year-on-year increase of 123.2%.

At the same time, the competitive landscape is shifting. Since 2026, major internet companies like Tencent, Alibaba, and ByteDance have been consolidating their AI office products. Amidst this intense competition, customers are no longer solely focused on model parameter size, but are more concerned about whether a product can solve real-world problems.

Zhou Hongyi believes that AI is entering a phase of "dual-track evolution for models and agents." He notes, "Models determine how smart AI is, and agents determine whether AI can actually do work." Enterprise competition now depends not only on model capabilities but also on the ability to combine models, knowledge, tools, and business processes to deliver stable task completion.

Whether NanoWork can carve out a place in this fiercely competitive market remains to be seen. The leap from conversational agents to agents that can deliver results tests not only technical capability but also the understanding and adaptability of real business scenarios.