Why Middle Managers Block AI‑Driven Change at 360: Insights from Zhou Hongyi’s Letter

On August 17, Zhou Hongyi sent a letter to all 360 managers assigning three AI‑related tasks that will be evaluated annually:

Drive at least one AI product or business innovation.

Run at least one core process re‑engineered with AI.

Deliver one tangible result personally completed with AI.

The third task stresses that managers must do the work themselves, not delegate or outsource.

01 Change failures first blame the front line

Post‑mortems of failed transformations often attribute failure to employee non‑cooperation, weak execution, or attachment to old habits. For frontline staff, AI‑enabled process redesign means learning new tools, gaining efficiency, and moving closer to decision‑making—essentially an opportunity without loss of existing assets.

A landlord‑renovation analogy illustrates that tenants (frontline workers) have little reason to resist because the old furniture is not theirs.

02 Middle managers are the motivated blockers

Middle managers control three levers: process authority, information interpretation, and team direction. AI‑driven flattening and process re‑engineering directly threaten these levers.

Harvard Business School research on AI‑native firms finds a 15% lower manager‑to‑employee ratio than traditional firms. ByteDance has announced the removal of “relay‑type” managers, and 360 is simultaneously launching a flat structure, AI‑native product‑research transformation, and increased campus hiring.

03 “Seemingly compliant” is the safest resistance

Middle managers rarely oppose change overtly. Instead they adopt high‑compliance passive execution: publicly endorse the initiative, selectively filter information, exaggerate risks, and deliver only 70‑80% of the work, leaving critical steps blocked.

This creates a mismatch—executives see enthusiastic, orderly progress, while frontline staff encounter loud promises but little actual change, leading to the blame being placed on employees.

04 Executives can’t see because information is filtered

Middle managers’ assessment sheets lack transformation metrics, while their personal cost‑benefit tables are filled with change‑related expenses. Success benefits the boss; failure penalizes the manager, making a hands‑off stance rational.

Information filtering turns middle managers into toll stations: every piece of feedback must pay a “fee” (their perspective) before reaching the top.

Therefore Zhou requires managers to “personally” complete an AI task and ties it to performance: only hands‑on experience can reveal AI’s boundaries, what to delegate, and what to guard—knowledge that reports alone cannot convey.

05 Turning middle managers into the engine of change

Recognizing middle managers as a source of resistance does not mean treating them as enemies; instead, make them the primary beneficiaries of transformation.

Three concrete practices:

Embed transformation metrics into performance evaluations. Unmeasured change remains a campaign; measured change becomes behavior.

Define a new role for middle managers in the redesigned organization, showing them a forward path rather than only what they lose. Fear‑driven resistance can be replaced by incentive‑driven participation.

Allow frontline feedback to bypass middle managers and reach executives directly. This prevents information decay at the “middle station”.

Conclusion

The sequence of organizational change is always: tools, processes, structure, then people. The “people” gate is first passed by the middle layer.

Frontline staff are not the obstacle; they reflect the attitude of middle managers. When change stalls, examine whether the middle layer is genuinely pushing forward or merely appearing compliant.

Code example

AI