Big tech’s “workhorses” pushed into AI: implicit mandates, token quotas and the risk of deskilling

"Use AI to boost efficiency" — or else?

It has been reported that China’s largest internet firms and their foreign counterparts have moved from encouraging AI use to effectively mandating it for ordinary employees. TMTPost spoke with six staffers across roles — from junior engineers to CIOs — who describe daily token quotas, internal tracking dashboards and explicit expectations to produce “AI outputs” before human work. What began as a productivity boost for early adopters is morphing into a new layer of managerial control: prompt-tweakers are rewarded, steady coders are labelled “inactive.”

Productivity gains, repetitive toil and new metrics

The on‑the‑ground stories are mixed. Some workers say AI doubled their throughput; others recount grinding through dozens of prompt iterations — one operator reported reworking a data dashboard 80 times — or deliberately deleting code so an internal assistant would generate activity. It has been reported that some teams are asked to convert institutional knowledge into reusable “Skills” for agents, and that one company set a target to have 50% of development tasks generated by agents now and to scale toward full automation by 2026. Internal tools — from company assistants to proprietary LLMs — are being pushed in the name of data security and cost control; parallel usage of Western services like ChatGPT (OpenAI), Claude (Anthropic) and Gemini (Google/谷歌) continues where allowed.

Culture shifts and geopolitical context

Workplace dynamics are shifting fast. Employees who obsess over prompts and token consumption are lauded as “embracing innovation,” while those who focus on deep engineering work can be sidelined. Reportedly, companies track token use centrally and some managers use those metrics in performance conversations, creating new anxieties about replaceability. This push coincides with broader geopolitical pressures — including U.S. export controls on advanced chips and AI tools — that have accelerated Chinese firms such as Baidu (百度) and Alibaba (阿里巴巴) to develop and prefer in‑house models, often citing data‑security and regulatory reasons.

A second industrial revolution — for whom?

The promise of higher efficiency is real. But employees warn of an equally real cost: the slow erosion of hard‑won technical judgment as routine work is siphoned into SOPs and Skills. Are workers being asked to wield AI, or to become its fuel? For now, the answer varies by team and talent. What’s clear is that the era of raw human hours as the primary productivity metric is ending — and people are scrambling to work out what comes next.