Klarna just raised $1.37B — and proved why AI still needs humans
Big moment for European tech: Klarna finally went public in the US. Shares jumped 30% on the first day, giving the company a $17B market cap. Investors are clearly betting big on its future.
But here’s the irony: while the IPO is a success story, Klarna’s AI experiment tells a different one. Not long ago, the company shifted hundreds of engineers, marketers, and even lawyers out of their roles, betting that AI could replace 700 customer support agents.
On paper, it looked efficient.
In practice? Customers complained about wrong answers, irrelevant replies, and unresolved issues. Service quality dropped fast. Even Klarna’s CEO later admitted: “We took cost-cutting too far.”
The problem wasn’t the technology. It was the way leadership chose to use it. Instead of rebuilding the system, they reached for a “magic pill” — slash costs and hope AI would fill the gap. The KPIs were wrong from the start: efficiency over quality. The data likely wasn’t strong enough — incomplete or outdated customer interactions can’t train a system to handle complex issues. And they ignored something just as important: psychology. Customers want empathy. A chatbot that says “I understand your frustration” while clearly not solving the problem doesn’t cut it. This isn’t a new idea — management expert W. Edwards Deming wrote about it back in the 1980s: cut costs without rethinking the system, and quality will always suffer.
So what actually works in practice?
- Automate the routine. Document handling, draft responses, data validation — these deliver real gains in speed and cost. And this isn’t just our opinion. McKinsey’s report on the economic potential of generative AI shows that automation of repetitive tasks is where companies see the fastest ROI. Read the report here.
- Leave critical points to humans. Decision-making in edge cases, quality control, final validation. This reduces risk and improves customer satisfaction.
That’s how we approach it at CODECAVE:
- Start with the goal — which KPI are we improving (response time, cost per case, processing time)?
- Automate specific steps, not the entire cycle — integrate data, add AI where it speeds things up, leave oversight to humans.
- Test, adjust, scale. This approach let us get one of our products to testing earlier than planned— saving $100K in development costs.
It’s low-risk, high-impact, and sustainable. That’s why our first client from 2019 still works with us and recommends our team to partners. We’ve helped companies replace expensive tools and design custom solutions across logistics, e-commerce, healthcare, and agriculture.
If you’d like to see how this approach works step by step, take a look at our case study — we share the process and results in detail.
Let’s explore how AI can cut your costs without cutting quality — book a free call with our team.
