In its first month, one AI assistant ran 2.3 million customer conversations – the work of 700 full-time agents. Resolution time fell from 11 minutes to 2, at the same satisfaction scores. Estimated impact: ~$40M in profit.
Conversations handled by the AI
Full-time agent equivalent
82% faster, same satisfaction
Estimated, for 2024
Klarna serves 150 million consumers across 23 markets. That scale generates an enormous, relentless volume of support: refunds, returns, billing questions, disputes, order status – the same kinds of requests, in dozens of languages, around the clock.
Handled the traditional way, that means armies of agents, long queues, and an average resolution time of 11 minutes per request – much of it spent waiting and re-explaining. Every new market and every spike in volume meant more hiring, more training, more overhead. Customer service was scaling linearly with headcount, and headcount is expensive.
Working with OpenAI, Klarna built an assistant that doesn't just answer questions – it acts. Wired into the same backend systems agents use, it reads an incoming message, looks up the account, and resolves the request end to end: processing refunds and returns, answering billing questions, handling disputes – in 35+ languages, 24/7. The two-thirds it can fully handle never touch a human; the rest route to agents.
Illustrative – representative of how the assistant resolves and routes requests.
conversations in a single month – two-thirds of all of Klarna's customer service chats, handled by the AI.month one, not year one
full-time agents' worth of work absorbed by one deployment – reportedly built for a few million dollars.
average resolution time – an 82% drop – with customer satisfaction on par with human agents.faster, not worse
estimated profit improvement for 2024 from a single automated workflow – plus 25% fewer repeat inquiries.
The AI took the two-thirds that's repetitive – so the humans could keep the third that actually needs a human.The pattern behind every good automation
You don't have Klarna's volume – and you don't need it. Every business has its own repetitive two-thirds: the support tickets, the FAQs, the refunds and status checks, the same questions answered by hand, all day long.
That's the work we put a machine on. We find the requests that don't need a human, build the assistant that handles them end to end, and run it – and we prove the number on your real data in 30 days, or you pay nothing.
Klarna's number was $40M. Yours is smaller. It's also a lot closer than you think.
A 30-minute call with our senior team. You'll leave knowing what's automatable in your business, the number we'd go and hit, and how the money-back pilot works – whether you hire us or not.
30 minutes of pure value – no slide deck, just your numbersOr reach us directly · Telegram · info@beawhale.io
Figures from Klarna's February 2024 press release and OpenAI, with reporting from AIBusiness.vc. The 2.3 million conversations, two-thirds of chats, 700-agent equivalent, 11-to-2-minute resolution, 25% drop in repeat inquiries and ~$40M projected 2024 profit improvement are drawn from these accounts. Klarna later rebalanced toward a hybrid model, reintroducing human agents for complex cases. The support window above is illustrative. BeAWhale is not affiliated with Klarna or OpenAI – this story is shared as an industry reference of what support-automation AI makes possible.
