Meta and Klarna’s AI Workforce Retreat: Why Automation Still Needs Humans

AI is changing how companies staff teams, but Meta and Klarna illustrate why replacing whole functions with AI remains far harder than automating pieces of work. Meta halted planning for a second Project OT restructuring wave after an initial workforce reduction, while Klarna resumed hiring human support workers after its AI-first service model produced lower-quality outcomes. 218

The shared lesson is practical rather than ideological: AI can absorb structured, repeatable tasks at scale, but organizations still need people to handle ambiguity, review outputs, resolve exceptions, and protect the customer experience.

Meta: a smaller-team vision met real-world limits

Reuters reported that Meta’s Project OT, short for Organization Transformation, was designed as a two-wave restructuring. Meta carried out the first reduction in May 2026, cutting roughly 10% of its workforce, then called off planning for the later November wave. Meta confirmed Project OT existed, describing it as a year-long effort involving cost cutting, team redesign, and shifts into priority areas. 2

The reversal matters because the plan was not simply a conventional cost-cutting exercise. It envisioned much smaller groups of employees supported by AI agents. The reported outcome suggests an important distinction: producing more drafts, code, or task outputs does not automatically translate into dependable product delivery. In complex workflows, people still have to define goals, connect work across systems, test results, spot failures, and take responsibility when something goes wrong.

Meta has not abandoned AI investment or AI-enabled work. What changed was the confidence that agent capabilities could justify a second broad workforce reduction on the original timetable. 2

Klarna: automation volume did not guarantee service quality

Klarna’s experience shows the customer-service version of the same problem. Its AI assistant handled a substantial share of standard support conversations, including routine questions about payments, refunds, and returns. But Klarna later moved away from an AI-only posture and resumed hiring human support staff. 1720

According to reporting on the company’s shift, CEO Sebastian Siemiatkowski acknowledged that Klarna had reduced human capacity too aggressively. The problem was not that AI had no value; it was that automated service struggled with complex, ambiguous, and emotionally sensitive cases, where experienced people bring investigation, discretion, and empathy. Klarna began rebuilding human support capacity alongside automation. 1819

That distinction is especially important in support operations. A system can resolve common requests quickly, yet still damage trust if customers with disputed payments, unusual circumstances, or escalating problems cannot reach a capable human promptly. Automation is valuable only when escalation paths and human accountability are designed into the service.

The emerging model: AI as leverage, not a substitute for every role

Together, the two cases point to a more durable operating model:

• Automate predictable tasks. AI is most useful where inputs are standardized, outcomes are easy to verify, and mistakes are inexpensive to correct.
• Keep humans at decision points. People remain essential when a task requires context across systems, trade-offs, judgment, accountability, or relationship management.
• Build escalation into the workflow. Customer support and other high-stakes processes need a clear handoff to qualified people rather than a dead end in an automated loop.
• Measure outcomes, not activity. More generated code, faster first responses, or higher task volume are incomplete metrics if quality, resolution, safety, and customer trust deteriorate.

The implication is not that jobs are untouched. It is that job change is likely to arrive through team redesign, altered skill requirements, and fewer roles dedicated to routine work—not through an immediate end to human work itself.

Hiring data show a reallocation toward technical capability

The broader technology labor market is consistent with that interpretation. SignalFire reported that hiring at large tech companies was 25% below its 2019 baseline, while software-engineering hiring was down 11%. Software engineers consequently represented 55% of hiring at those companies, compared with 46% in 2019. 49

This does not prove AI caused every hiring shift; tech hiring is also influenced by business cycles and post-pandemic workforce adjustments. But it does show that leaner organizations have not eliminated the need for technical talent. Instead, the mix of hiring has shifted toward people who can build, deploy, evaluate, secure, and supervise increasingly automated systems.

The biggest near-term risk is the entry-level career ladder

Stanford Digital Economy Lab researchers found no evidence of widespread, economy-wide job displacement in administrative payroll data through June 2026. However, employment among workers aged 22–25 in highly AI-exposed occupations was 19% below the level it would have reached had it kept pace with less-exposed peers. Experienced workers showed no comparable gap, and the adjustment appeared primarily through reduced hiring rather than increased separations. 35

That 19% figure is a relative “kept-pace” shortfall, not a count of jobs that AI directly eliminated. Still, it highlights a serious issue: when AI takes on portions of routine starter work, firms may offer fewer opportunities for people to acquire the experience that leads to more senior roles.

What companies should take from Meta and Klarna

The useful question is no longer whether to use AI. It is where AI can improve a process without removing the human capabilities that make the process reliable.

Meta’s halted second wave and Klarna’s support rehiring both suggest that companies should test automation against end results: successful product delivery, accurate decisions, resolved cases, customer confidence, and manageable recovery costs. Where AI performs well, it can increase the capacity of a smaller team. Where uncertainty or consequences are high, the right design is usually human-plus-AI—not human-free.

AI is therefore reshaping the composition and leverage of teams. The companies most likely to benefit will be those that treat automation as a tool for better human work, rather than a shortcut around the human work that complex products and customer relationships still require.