Work is a Verb
A newsletter by
Cut First, Measure Later
In early June, Meta's chief technology officer shared some surprising numbers on the company's internal processes.
Changes to the software platforms and infrastructure employees use inside Meta were up 220% year over year. Changes that actually resulted in new or upgraded features reaching Facebook and Instagram users were up only 36%.
Meanwhile, major technical and security incidents—including service disruptions and possible data leaks—were up 40% from the previous year. Time spent firefighting those incidents was up 70%.
Meta designed the cut before it had the measurement and proceeded with the first wave even as the measurements turned against it. In May, about 10% of employees were laid off, thousands more were moved into AI training work, and 6,000 open roles were closed. Hours before the layoffs began, Zuckerberg called off the second wave planned for November. By July he was telling employees that agent technology hadn't "accelerated" as quickly as he expected.
A Reuters investigation published this week detailed Project OT, a plan sketched in January at Mark Zuckerberg's Hawaii compound. It called for AI agents to do much of the daily work, while smaller "talent-dense" human teams supervise them. In the most aggressive scenarios, some teams would shrink by as much as 60%.
(In its response, Meta said those were scenarios, not an approved plan.)
Whether you believe that or not, the fact that the conversation happened at all demonstrates a level of cynicism about the company's talent that is hard to comprehend.
Meta is the best-documented case, not the only one.
Klarna presented its AI assistant as doing the work of 700 agents in early 2024; by May 2025 its CEO was telling Bloomberg the company had focused too much on cost, quality had suffered, and it was recruiting humans again. Commonwealth Bank of Australia cut 45 customer-service roles in July 2025, citing its new voice bot; within a month it called the redundancies an error, apologized, and let the affected employees choose to stay, redeploy, or leave. These are just a handful of examples of a broad problem.
The mistake underneath all three is defining a job by its most visible artifact. Engineers produce code, support reps close tickets, product managers produce plans.
But a functioning organization is not merely a collection of artifacts.
People decide what should be built. They understand dependencies that were never written down. They recognize when a technically plausible change will create a security problem somewhere else. They negotiate competing priorities, question flawed assumptions, review one another's work, and take responsibility when something breaks.
They also prevent bad work from happening.
That contribution is difficult to measure because its value often appears as an absence. Good leaders know it's there, even if it doesn't appear on any dashboards.
Meta believed too much in AI but, more importantly, too little in its people. It's a cautionary tale, but it's also cause for optimism.
Every day we are reading headlines about how soon we're going to be replaced, but the leaders who are buying into that narrative are showing that it doesn't survive contact with the work.
The full account, with Meta's own numbers, the Klarna and CBA reversals, and the questions to ask before you sign, is in Cut First, Measure Later: What Meta's AI Layoffs Actually Showed.
