360 Ties Every Manager's Year-End Review to Three AI Deliverables

Zhou Hongyi's 8/17 all-hands memo to 360 reads on first pass like a slogan. "AI Native." "All in Agent." "Rebuild the company with AI." The mechanism is one line deeper: every manager's year-end review now carries three AI deliverables, and the manager's job description shifts from "approval machine" to "AI team coach."

360 is a major Chinese internet-security firm. The memo, titled "Driving Business Innovation and Organizational Change with Leadership," was published on August 17, 2026 and first reported by Chinese tech outlet Leiphone (雷峰网). Zhou frames the moment as "not a typical leadership upgrade, but 360 using AI to restart entrepreneurship." The line is meant to position the move as more than a refresh of management training.

The three deliverables are concrete. Each manager must, in the current review cycle, lead their team to deliver at least one AI product or business innovation, run through at least one core process that has been rebuilt around AI, and produce at least one substantive piece of work they personally completed with AI. All three are tied to the annual review. Without that link the memo would be a manifesto; with it, the memo is a performance contract.

The role reframe is the second mechanism. Zhou tells 360's managers to stop acting as "approval machines" (审批机器) and become the "coaches and commanders of super individuals" (超级个体的教练和指挥官), meaning the AI-fluent individual contributors on their teams. The memo also tells managers to "learn, iterate, and get hands-on" so that they themselves become "super individuals" proficient in AI. Zhou's claim: an AI-native org cannot be run from the top of an approval chain.

The organizational moves that travel with the memo: flatten the structure, push grassroots managers back toward individual innovation, accelerate the R&D team's "AI-native" transformation, and intensify campus hiring to lower the median age of the talent base. 360 frames the move as an upgrade of its year-old "All in Agent" strategy, in which AI has already been deployed in product R&D, business operations, and internal collaboration. The company says R&D efficiency and innovation capability have "markedly improved." That is a self-reported metric, not independently verified.

The industry-insider reading, as paraphrased by Chinese coverage, is that enterprise competition has moved from "do you know how to use AI" to "does the org have AI-native capability." The variable that separates winners is no longer access to models or compute, but organizational learning speed and the human-machine collaboration layer above the model. Sina Finance's same-day coverage and Sina Tech's parallel report carry the same framing; NetEase's repost confirms the three-assignment list. The view is paraphrased, not attributed to a named source, so it should be read as mood, not as analysis.

The falsifier is straightforward. Do the year-end reviews actually fire? The mechanism only works if a manager who fails to land any of the three assignments visibly pays a price in compensation, promotion, or scope. If the assignments are absorbed into existing targets and the review process continues as before, the memo is a rebrand, not a change. The next two review cycles will tell.

The second watch item is the manager's lived experience. A current or former 360 manager asked whether the memo reads as "real empowerment" or "restructuring-by-memo" would settle the question. The memo gives the company the option to do either.

360's move is not isolated. Baidu, Alibaba, and ByteDance have all made AI-driven org changes in the same quarter, and the slogans travel between them. 360's version is distinguished by operational specificity: three named assignments, year-end-review enforcement, and an explicit reframe of the manager's job from gatekeeper to coach. Whether that specificity survives contact with the review process is the test the wire cannot run.