Companies that went AI-only or AI-first in customer service are quietly bringing humans back because the AI cut costs but lowered quality on the contacts that mattered most. The rebalance is usually a hybrid: AI handles the simple repetitive volume, humans handle the harder cases.
Klarna's AI assistant was held up in 2024 as the case study for AI-first support, said to handle the work of hundreds of human agents. By 2025, Klarna was hiring human agents again. The CEO said the AI cost less and delivered lower-quality customer experience. The shift is not unique to Klarna.
What people in the field are saying
CX Decoded has a piece, "The companies quietly rehiring humans," tracking exactly this pattern. CX Dive covered the Klarna reversal directly, with Klarna's CEO acknowledging that AI cost less but delivered lower-quality output.
What does the all-AI bet usually look like?
A company replaces a large share of its support workforce with an AI agent or chatbot. Cost falls quickly. Containment numbers look healthy. Average handle times drop. The early months of metrics show the bet paying off.
Why does it stop working?
Two things break. The first is quality on the cases that need judgement: returns gone wrong, account problems, complaints about a complaint. The AI handles those badly, the customers feel it, and the brand pays the bill in trust and churn that does not show on the support dashboard for months.
The second is that the AI itself drifts. A model that worked on month one is producing wrong answers on month four because the underlying knowledge changed and nobody maintained it. Without a human team in the loop, there is no one whose job it is to notice.
What does the rebalance look like?
A hybrid where AI is given the contacts it does well (routine FAQs, status lookups, simple actions) and humans get the rest. The team is smaller than the pre-AI team and the people in it spend their time on harder cases. Pay tends to go up, because the work is harder. Hiring criteria shift toward judgement and writing skills rather than throughput.
What is the takeaway for a CX leader?
Treat the first cost cut from AI as a one-off windfall, not the new baseline. Plan for human agents to handle the residual hard work at the same time you deploy the AI for the easy work. The companies that did this from the start did not need to make headlines about rehiring.
Related: Will AI replace customer service jobs, the business case for bringing humans back, and the glossary explainer on human-in-the-loop.