Klarna. Hiring the humans back.
Klarna replaced 700 customer service staff with AI, spent a year saying it was the future, and is now hiring humans back. The most expensive mistake is getting rid of people.
THE GIST ON KLARNA: "Klarna"
What did Klarna do?
Klarna replaced 700 customer service staff with AI. The CEO spent a year telling everyone it was the future, and now he's hiring them back. The most expensive thing in any business is a person, but the most expensive mistake is getting rid of them.
Klarna partnered with OpenAI and built a chatbot that handled 75% of customer queries, and it went straight to the CEO's head, as all cost-saving measures do. He said AI could do every job a human could, stopped hiring, and headcount fell by nearly a quarter in a year.
What went wrong?
The chatbot was fast, it was cheap, and it was infuriating. When you've been charged twice and you're fuming, you don't want a clanker telling you it understands how you feel. It doesn't.
AI can solve a problem, but it can't do what a customer wants: care about it.
What came next?
Six months later the CEO said it had focused too much on cost-saving, and he was hiring everybody back. The most famous AI workforce replacement in the world was reversed.
According to IBM, only one in four AI projects deliver what they promise. Klarna is a very expensive proof of it.
Is this new?
Every CEO is being told the same thing: cut the humans, the spreadsheet says it works. Klarna's spreadsheet said it worked too.
But this isn't even new. Replace AI with outsourcing, offshoring, automation. Every generation of management finds a shiny new reason to get rid of people, and every generation works out it isn't a good idea. You'd think someone would skip to the end.
Chatbot or person for a customer complaint?
Your take. One tap, then see how everyone else voted.
Quick answers
What did Klarna do with AI?
It replaced 700 customer service staff with a chatbot built with OpenAI that handled 75% of queries.
Why did Klarna reverse course?
The CEO said it had focused too much on cost-saving, and it began hiring people back about six months later.
How often do AI projects deliver?
According to IBM, only one in four deliver what they promise.
See you on the next one.
What should the next one be? Suggest a topic for Huw to explain
The conversation
Comments
Nothing here yet. Start the conversation: Huw reads every comment before it is published.
