Why Companies Are Hiring Humans Again in the Age of AI

The story everyone got wrong and what the data actually shows

In 2023, Klarna’s CEO Sebastian Siemiatkowski stood next to Sam Altman and told the world he wanted his company to be OpenAI’s “favourite guinea pig.” Klarna froze all non-engineer hiring, built an AI chatbot in partnership with OpenAI, and within months claimed it was handling two-thirds of all customer service interactions doing the equivalent work of 700 full-time agents. It was held up as the definitive proof that AI was coming for jobs at scale.

Six months later, customer satisfaction had fallen sharply and service quality was inconsistent. Klarna was asking software engineers, designers, and marketing staff to answer customer inquiries. The same CEO who once minimized the role of people was acknowledging that the company had prioritized cost over experience.

By early 2025, internal reviews confirmed that the AI systems lacked empathy and could not handle the nuanced problem-solving required for customer support. Siemiatkowski publicly admitted: We went too far. Klarna began rehiring. The company that had made global headlines for replacing humans with AI was now making global headlines for bringing them back.

This is not an isolated story.

The “AI Boomerang” is real and it’s spreading

Industry analysts now call it the “AI boomerang.” According to talent consulting data from Robert Half, nearly 32% of companies that cut staff due to AI integration have already reopened and rehired for those exact same positions. Survey data from Orgvue and Forrester reveals that 55% of executives now openly regret their decisions to replace human workers with AI.

The companies involved are not small experiments. Commonwealth Bank of Australia cut 45 customer service roles in July 2025, claiming a new AI “voice bot” had cut call volumes by 2,000 a week. Within weeks, under pressure from the Finance Sector Union and a workplace tribunal, CBA admitted the claim was false, call volumes had actually risen, and reversed the cuts entirely, calling the layoffs an “error.” Salesforce tells a messier version of the same story: it cut roughly 4,000 customer support roles in 2025 as AI agents absorbed more service volume, then announced further layoffs in early 2026 even as it redeployed several hundred employees into sales and professional-services roles. Headcount didn’t bounce back so much as get reshuffled, and the reshuffling is still going on.

The pattern is consistent: aggressive AI replacement, followed by quality deterioration, followed by a quiet but expensive reversal.

What went wrong

The metrics that AI performed well on—resolution rate, time to first response, tickets handled per hour—masked quality deterioration on specific interaction types. The real damage showed up in CSAT scores and repeat contact rates: customers who had to contact support multiple times for the same issue because the AI could not actually resolve it.

There’s a reason even the most capable AI models fail in complex corporate environments. These models may possess vast data but lack understanding of a specific company’s culture, unwritten norms, and client history. AI tools that look flawless on paper break down when dealing with the messy, irregular workflows of actual business operations.

Emily Potosky, senior director of research at Gartner, put it plainly: “What we have figured out is that most of the layoffs in 2025 were unrelated to AI adoption entirely. They were related to things like the federal government, general business right-sizing decisions after having hired heavily in the 2021–2022 era.” The AI-caused-mass-unemployment story was, in large part, a narrative the data does not support.

The actual numbers on jobs created vs. destroyed

In 2024, AI generated approximately 119,900 direct jobs in the US. In contrast, outplacement firm Challenger, Gray & Christmas estimates that around 12,700 jobs were lost due to AI just 0.1% of all layoffs that year. The math is not subtle. AI created roughly ten times the jobs it eliminated in 2024.

AI-related job postings surged 25.2% in Q1 2025 alone, with median salaries hitting $157,000. Mentions of AI in job listings had already grown 114.8% in 2023 and 120.6% in 2024. These are not replacement roles being renamed. They are net new categories of work that did not exist before.

Harvard Business School analysed job postings from 2019 to 2025. Job postings for roles with high automation scores—structured, repetitive work—decreased 13% after ChatGPT’s release. But roles with high augmentation scores—analytical, technical, and creative work that AI complements—increased 20%. The direction of the market is toward people who can work with AI, not away from people entirely.

The specific roles companies are now racing to fill

AI Engineer postings grew 143%, Prompt Engineer 136%, AI Content Creator 135%. What is notable is that design has overtaken technical expertise as the most in-demand skill in AI-related job postings. Human judgment, communication, and applied insight—not raw coding ability—are what companies are now actively struggling to hire.

The demand for roles that combine domain-specific expertise with AI literacy including AI system architects, ethics and governance specialists, human-AI collaboration designers, and physical AI specialists will significantly increase through 2030.

BCG reported that 25% of its $14.4 billion in 2025 revenue—roughly $3.6 billion—came directly from AI consulting work. Clients are renting AI expertise faster than they can build it internally, which is funding thousands of new consulting roles in AI strategy, data engineering, MLOps, and change management.

The business rationale is straightforward: companies that bought AI tools now need people who know how to make those tools produce results. That is an entirely different skill set from the roles that were automated away.

The hiring shift nobody expected

ManpowerGroup’s global research found that companies investing most heavily in AI are also investing most in human capital. Their 2025 Global Talent Barometer showed worker confidence rising, not falling. The companies farthest into AI are not the ones cutting the most people. They are the ones hiring more carefully, for more specific capabilities.

The World Economic Forum’s Future of Jobs Report 2025, drawing on data from over 1,000 companies, projects 170 million new roles will be created by 2030 while 92 million will be displaced a net gain of 78 million jobs. The skills gap, not headcount, is identified as the most significant barrier to business transformation today.

The real story

The AI-replaces-humans headline was clean, dramatic, and largely wrong or at least, deeply incomplete. Rehiring costs at companies like Klarna exceeded the original savings estimate. Reversing AI-driven layoffs required recruiting, onboarding, and training new staff an expensive process that companies rarely model in their AI replacement business cases. The true cost of full replacement includes the cost of unwinding it when it fails.

Many leaders assumed automation would create a straight line from cost reduction to efficiency. Instead, they discovered that removing people without rebuilding the work around them often breaks the system that customers rely on.

What the data shows across job postings, layoff reports, consulting revenues, and company reversals is that the age of AI is producing more demand for human talent, not less. Different talent, yes. Differently structured, certainly. But the story of AI and employment is turning out to be one of transformation, not elimination. The companies learning that fastest are the ones already ahead.

So Are Companies Actually Hiring Humans Again?

Short answer: yes. Longer answer: the smart ones never actually stopped needing them, they just took the scenic route to figuring that out.

That’s the bet we’re making at Incresco too: AI that makes the people on your team dangerous in the best way. Engineering that augments judgment instead of replacing it, so it amplifies your team’s productivity.

If you’d rather skip the boomerang and get it right the first time, let’s talk.