What Happened When AI Replaced 80% of a Company's Workforce

The Klarna Experiment

In early 2024, Klarna's CEO Sebastian Siemiatkowski made a statement that sent shockwaves through the tech industry. The Swedish fintech giant had stopped hiring. Not slowed down. Stopped. And through natural attrition and AI adoption, the company had gone from roughly 5,000 employees to about 3,800.

But the headline number that really got people's attention was this: Klarna's AI assistant was doing the work of 700 full-time customer service agents. Klarna announced that within the first month, their AI was handling two-thirds of all customer service chats. Resolution times dropped from 11 minutes to 2 minutes. Customer satisfaction scores remained essentially unchanged.

Seven hundred people's worth of work. Handled by software. In its first month.

Let me say something that might sound obvious but needs to be said anyway: those 700 jobs weren't abstractions. They were people with rent payments, kids in daycare, car loans, and grocery bills. People who went to work every day thinking they were building a career, only to discover they were training their replacement without knowing it.

The Dukaan Story Is Even More Jarring

If Klarna was a gradual shift, what happened at Dukaan was a guillotine drop.

Suumit Shah, CEO of the Indian e-commerce platform Dukaan, publicly announced on social media that he had replaced 90% of his customer support staff with an AI chatbot. Not over months. Almost overnight.

His tone was triumphant. Resolution time went from 2 hours and 13 minutes to 3 minutes and 12 seconds. First response time dropped from 1 minute 44 seconds to instant. Customer support costs were slashed by 85%.

Shah called it "tough but necessary." The internet's reaction was more divided. Some praised the efficiency. Others were horrified by the casualness of it — a CEO essentially celebrating the displacement of his own team on Twitter.

What nobody talked enough about was the workers themselves. In India, where Dukaan is based, customer service jobs are often pathways to the middle class. They're the jobs that let people afford education for their children, build small savings, and dream a little bigger. When those jobs disappear overnight, the impact ripples through families and communities in ways that don't show up on a CEO's efficiency dashboard.

The Human Cost Nobody Wants to Quantify

I spoke with a former customer service team lead at a mid-sized tech company — not Klarna or Dukaan, but a similar story. She asked me not to use her name. Here's what she said:

"They brought in the AI system in March and told us it was to 'help with overflow.' By June, half my team was gone. By September, I was gone too. The worst part wasn't losing the job. It was the three months of watching it happen and being told everything was fine."

Her experience isn't unique. Across the industry, the pattern repeats: AI tools are introduced as "assistants" or "augmentation." They handle the easy cases first. Then the medium cases. Then most cases. And somewhere along the way, the math changes and the humans become the expensive redundancy.

Wired documented multiple cases of customer service teams being gradually hollowed out while being assured their jobs were safe. The psychological toll is significant — months of anxiety, watching your workload shrink, knowing what it means but having no power to change it.

The Numbers Keep Piling Up

Klarna and Dukaan are the most publicized cases, but they're far from the only ones. Consider:

• IBM announced plans to pause hiring for roles that AI could handle, potentially affecting 7,800 positions
• Dropbox laid off 16% of its workforce, with CEO Drew Houston explicitly citing AI as part of the rationale
• Chegg, the education company, saw its stock crater after acknowledging that ChatGPT was devastating its business model, leading to significant layoffs
• UPS cut 12,000 jobs in 2024, with AI-driven logistics optimization as a key factor

Each of these announcements represents thousands of real people receiving the worst email of their professional lives. And for every company that makes headlines, dozens of smaller firms are making the same moves quietly.

What the Companies Won't Tell You

Here's an uncomfortable detail that gets glossed over in the corporate press releases: the quality gap isn't as big as displaced workers want to believe.

Klarna's customer satisfaction scores stayed flat after the AI transition. In some metrics, they actually improved. The AI doesn't have bad days. It doesn't get frustrated with difficult customers. It doesn't put people on hold while it checks with a supervisor.

Does it handle every situation perfectly? No. Complex, emotionally sensitive issues still benefit from human touch. But those represent maybe 15-20% of interactions. The other 80% — password resets, billing questions, order tracking, return policies — the AI handles just fine.

And that 80/20 split is the real story. Companies don't need to replace all their human workers. They just need to replace enough of them to transform the economics. Keep a small, highly skilled team for the tough cases. Let the AI handle everything else. The result: dramatically lower costs with marginally different outcomes.

The Acceleration Problem

What makes this particularly alarming is the pace. Klarna's AI went from launch to handling two-thirds of all chats in 30 days. Not 30 months. Thirty days.

TechCrunch has tracked the deployment timelines and they're compressing rapidly. What took early adopters 6 months to implement in 2023 now takes 6 weeks. The tools are better, the playbooks exist, and every successful case study gives the next company more confidence to pull the trigger.

This means the window for workers to adapt is shorter than most people realize. If your company hasn't started its AI transition yet, it's not because they've decided against it. It's because they're watching the early movers and building their plan. The question isn't if. It's when — and "when" keeps getting sooner.

So What Happens to the Displaced Workers?

This is the question that should be driving our national conversation, and it's the one we're most aggressively avoiding.

Some displaced workers pivot successfully. They move into AI management, quality assurance, or training roles. But these positions are far fewer in number than the ones they replace. You don't need 700 people to manage an AI system that replaced 700 people. You might need 30.

Others end up taking lower-paying jobs in sectors that haven't been automated yet. Some leave the workforce entirely. The data on workers displaced by previous waves of automation — manufacturing, offshoring — shows that many never recovered their previous earning level.

It doesn't have to be this way for everyone. But pretending it's not happening, or assuming you'll figure it out when the time comes, is a recipe for being blindsided.

The first step is understanding your actual risk level — not based on vague anxiety, but on real data about your specific role and industry. That's exactly what our free AI career risk assessment at jobsaiwillreplace.com is designed to do. Three minutes of your time now could save you from being the last person in the room to realize the floor is moving.