Convenient for Whom?

At Qidian Interactive, an AI-video studio in Wuxi, faster production changes different jobs differently. In Sixth Tone’s August 25 report, lighting artist Ma Shuan says some cleanup takes half a day instead of two or three days, although conventional tools remain faster for other fixes. Project director Chen Jie describes a different burden: reviewing dozens, sometimes hundreds, of shots daily, compared with roughly 20 a week before.

Those accounts put time saved and review work in the same workplace. They do not measure the studio’s net productivity, pay or staffing changes. The August reporting also predates October’s employment policy; it cannot show that policy working. The useful question is what happens between generating a shot and accepting finished work.

The ILO brief broadens the inquiry through interviews across 21 firms and a survey of 1,591 professionals. Its heterogeneous measures and uneven evaluation do not yield a single national effect. Faster task completion is worth examining, but it leaves quality, review time and the distribution of gains to be established.

Can a worker keep the saved time, does management count review as work, and does the customer receive a resolved request? These remain questions, not measured benefits. The figure separates workflow changes, reported job loss, a dismissal dispute and policy.

Reuters’ June 10 report describes microdrama producer Ayase losing a job in February as the workplace adopted AI. Reuters interviewed nine workers across several industries; the inspected text is its Investing.com syndication. This supplies an attributed job-loss account, not a national displacement estimate or an independently verified explanation of the dismissal.

A different outcome appears in China Youth Daily’s April 28 coverage of a Hangzhou court press conference. An unnamed technology employer proposed moving AI quality checker Zhou to another role and reducing monthly pay from RMB 25,000 to RMB 15,000. Zhou refused and was dismissed. The report says arbitration and subsequent court proceedings supported compensation for unlawful dismissal, including on appeal. The lower salary was proposed, not shown to have been paid.

Southern Weekly’s July 10 follow-up narrows that gap. In an attributed repost inspected beyond the publisher’s restricted preview, it reports more than RMB 300,000 in N+1 severance already paid, distinct from the court-ordered compensation. The lawyer-mediated account describes a brief later job followed by unstable employment; the worker declined interview. No pay or hours from that job are given.

Reported prior severance does not confirm receipt of the court-ordered compensation. Bank receipts and the full judgment remain uninspected, and July’s employment account does not establish November’s position. None of these earlier cases demonstrates October policy implementation.

The State Council General Office opinion is dated October 16; the inspected official republication appeared on October 24. Section 6 encourages enterprises to explore a buffer period, give affected workers advance notice during large-scale AI adoption and support job transitions. Section 7 calls for employment-risk assessments during reviews of market-access projects that could substantially affect employment, with assessments in other fields according to circumstances.

These provisions identify responsibilities to examine. They do not show that an employer has provided notice, training or a new role, and encouragement to explore a buffer period is not a universal new entitlement. Local implementing measures and worker accounts are still needed to connect the opinion to workplace practice.

Follow completed work through production, checking and acceptance; record hours, quality, targets and compensation before and after adoption. Then follow affected workers beyond dismissal or an award into actual payment and new work. These are different outcomes, and none follows automatically from faster generation.

A nationally representative productivity gain for Chinese workers has not been established by the reviewed sample.