Beijing’s Quiet Lesson for Western CEOs

Every Western CEO announcing AI-led layoffs is publicly confessing they cannot lead the transformation.

I want to sit with that sentence for a moment, because it is the one most boards refuse to say out loud. When Marc Benioff tells investors he cut Salesforce support “from 9,000 heads to about 5,000, because I need less heads,” that is not a strategy. It is a leadership failure dressed in the language of efficiency. When Andy Jassy frames Amazon’s 14,000 corporate cuts as running “the world’s biggest startup,” that is not transformation. It is the Pilot Trap at scale.

The Pilot Trap (https://us.amazon.com/Pilot-Trap-Category-Enterprise-Failure-ebook/dp/B0DF5YVR4H) is the failure mode my friend and colleague Vijayan Jagannathan diagnoses in his book of the same name, and which I had the privilege of writing the foreword for. It is what happens when leaders mistake a technology project for a transformation. They run pilots, they measure tokens and cost-per-ticket, they announce headcount savings on the next earnings call, and they wonder six quarters later why the franchise feels hollow and the customers are leaving. They confused activity for capability. They confused cost-out for value. They confused doing AI for leading through it.

Now here is what makes this week’s writing different.

I have spent the past month watching China do the opposite, and the comparison is brutal.

Read the State Council’s “AI+” Opinions from August 2025 and you will not find a cost-takeout document. You will find a leadership document. AI must “create new job openings.” Companies must “explore new organisational models based on human-machine collaboration.” Provincial plans are instructed to “reduce impacts on employment.” Then read the Hangzhou Intermediate People’s Court ruling, published deliberately on the eve of International Workers’ Day this year. A tech company tried to cut an employee’s salary from 25,000 yuan to 15,000 because AI now did part of his job. He refused. They fired him. Three levels of the Chinese legal system ruled the dismissal unlawful. The principle, stated plainly: AI cost savings are not legal grounds for termination. A Beijing arbitration panel went further in December. Adopting AI is “a proactive technological innovation,” and a company that uses it as a pretext for dismissal is “shifting the risk of normal technological iteration onto the employee.”

You can read this as authoritarianism. Plenty will. I read it differently.

China has not solved enterprise AI because it is virtuous. It has solved it because regime stability requires what the data already requires, and Beijing has the unsentimental clarity to mandate it. MIT’s NANDA report this year found 95 percent of enterprise generative AI pilots delivered no measurable P&L impact. The 5 percent that worked produced their savings, two to ten million dollars annually, “without material workforce reduction.” McKinsey surveyed 1,993 executives across 105 countries and found the high performers, the small minority capturing real value, share one behavioural signature. They pursue growth and innovation, not cost reduction. They are 3.6 times more likely to use AI for transformative redesign. Many expect headcount to rise. BCG found leaders allocate ten percent of effort to algorithms, twenty percent to data, and seventy percent to people and process.

So here is the position I will defend in any boardroom that wants to test it.

The Western CEO announcing AI-led layoffs is not running a transformation. They are running the Pilot Trap with a louder microphone. Sebastian Siemiatkowski admitted this in May, after Klarna’s reversal, when he conceded that “cost unfortunately seems to have been a too predominant evaluation factor.” That is the most expensive sentence in enterprise AI in 2025. Read it twice. The CEO who built his brand on AI replacing 700 humans is now publicly admitting that cost-led AI strategy degrades quality, hollows out capability, and forces a rehire. He is the canary. The rest of the Fortune 500 cost-cutters have not yet finished walking into the same wall.

Three things this means for any leader currently signing off an AI business case.

First, if your board paper leads with cost-out, your board paper is laggard behaviour. The empirical ROI evidence is unambiguous. Cost-led framing produces lower returns and higher reputational tail risk. The fact that activist shareholders reward the announcement does not change the fact that the underlying strategy is destroying long-run value. Your job as CEO is to know the difference and to defend it.

Second, the cultural tax is brutal and underpriced. When AI is framed as worker replacement, your best people stop teaching it the workflow context it needs to actually work. They hoard knowledge. They build shadow tools on personal phones. MIT documented 90 percent of employees using personal AI at firms where only 40 percent had official subscriptions. That is not a productivity statistic. That is a vote of no confidence in leadership.

Third, the legal envelope is closing. The EU AI Act treats workplace AI as high-risk from August 2026. California’s automated-decision rules took effect in October. Colorado, Texas, and New York City are already in motion. The deregulatory rhetoric in Washington will not save your HR director from the state-level patchwork they are already trying to comply with. Beijing’s framing will arrive in Brussels and Sacramento with a different accent, but it will arrive.

The country I am not supposed to praise just out-strategised the Fortune 500 on AI. Not because it is wiser. Because its leaders had no choice but to do the thinking Western CEOs have been allowed to skip.

The only real choice now is to stop optimising the past and start creating the future. Cost-cutting is optimising the past. Headcount takeout is optimising the past. Announcing AI as efficiency is optimising the past. Real leadership in this moment is the harder work of redesigning the business around what becomes possible when human capability and machine capability compound. That is the future Beijing is mandating, the data is validating, and most Western boards are still avoiding.

The question is whether yours catches up before your customers do.