How Klarna's AI Assistant Replaced 700 Customer Service Agents
This title is not clickbait (I swear).
When GPT launched in 2022, the team at Klarna saw a massive opportunity to boost their bottom line - so they jumped on the AI bandwagon by licensing OpenAI's model to build their AI assistant.
C-suite execs at Klarna be like: ↓
The CEO of Klarna, Sebastian Siemiatkowski, is a huge fan of AI. He actually reached out to Sam Altman, saying, "I want Klarna to be your favorite guinea pig."
No surprise that the two companies have worked closely together ever since.
If you're not familiar with Klarna, here's a quick overview:
- Huge Swedish fintech player
- 150 million+ active users
- About 2 million transactions per day
- Employs approximately 5,000 people
So... What Can Their AI Actually Do?
I can assure you it's not another chatbot that spits out generic responses (I promise).
Their AI assistant now handles two-thirds of all customer service interactions. It has participated in 2.3 million conversations.
They've designed their AI assistant to become your go-to for everything:
- Processing refunds
- Handling returns
- Invoice disputes
- Payment issues
- ...and more
It now works around the clock, in 35 languages, across 23 countries.
Let's put this into context:
Imagine you bought something, didn't like it, and now you want your money back.
Normally, you'd go through the whole rigmarole of contacting support, explaining your case, and waiting for someone to handle it.
Not anymore - this is exactly what their AI assistant can help with instantly.
Cool, isn't it?
Are They Really Replacing 700 Customer Reps?
Klarna says their AI assistant is doing the work of 700 customer service agents.
Let that sink in. Seven. Hundred. Agents.
Naturally, this raises questions. What happens to the actual humans?
Are they all getting replaced by a machine? 🥹
Um... sort of (but not entirely).
Klarna's AI handles Level 1 tickets - mostly mundane, repetitive questions that shouldn't need humans, as they're well documented in knowledge articles.
AI is amazing at learning from structured knowledge; you could train an AI on this data, and your team could ask questions, and the bot should easily fetch answers.
The complicated, more nuanced problems still go to real people (which isn't changing anytime in the near future).
In my opinion, Sebastian isn't trying to downplay the impact AI will have on workplaces, jobs, and industries. While some jobs are being replaced, he views this as an evolution in how companies manage customer support—making processes faster, more accurate, and scalable at any given moment.
He's committed to having continuous conversations about the aftermath of AI and its implications for the workforce.
Tangible Impact on Their Customer Support
Klarna reports that customer questions that used to take about 11 minutes to resolve now take just 2 minutes.
Asking for a refund is now faster than brewing coffee!
And it's not just about speed; this AI is accurate. They've seen a 25% drop in repeat inquiries because the AI gets it right the first time.
The assistant handles the job and ensures you don't need to come back asking the same question twice (win-win).
The result?
Happier customers, fewer headaches, and a smoother experience for everyone involved.
Cool, But What About the Bottom Line?
$40 Million in profit lift:
(Was hard to say this with a straight face)
Klarna estimates that their AI assistant will add $40 million in profit by 2024. That's not small change.
Think of it like this: while the AI assistant takes over mundane tasks, Klarna saves on costs related to staffing, all while improving customer satisfaction.
Efficiency and savings? Now, that's the kind of corporate math everyone can get behind.
Does the AI Measure Up to Human Agents?
According to Klarna, their AI assistant is on par with human agents for performance and CSAT (Customer Satisfaction).
That's a bold statement, but their numbers seem to back it up.
What's fascinating is that their AI doesn't just mimic human support - it's faster, handles multiple requests simultaneously, and never breaks a sweat.
Chatbot haters, where are you now? 👀
How Good Is Klarna's AI Support Really?
This article wouldn't be complete without searching for real feedback from users who've interacted with their AI assistants.
My mantra in life: Never take anything at face value
Let me share the experience of Gergely Orosz, who took their assistant for a spin.
If you don't want to read all the details, here's a small snippet of what he said:
"It's well-built, and will probably be a blueprint for how to put an L1 (first-level support) chatbot together."
So how did he test it?
He made a small purchase and initiated a chat.
Right off the bat, he noticed how fast the chatbot worked 😅
In less than 20 seconds, he had an answer to his question about his order status. He said the response was accurate, and it felt similar to talking to a human rep - just one that didn't waste any time typing up long-winded explanations.
It got straight to the point, provided the info he needed, and even threw in some order-specific context like the price he paid and the date of purchase.
Like any curious person, he tried to trip it up by asking some more off-the-wall questions, and when he did, the AI politely handed him off to a human agent.
The AI Assistant in Action
Here are a few screenshots of the assistant:
What Was the Net Impact?
Here's a snapshot of what Klarna has achieved with AI:
Have They Put Any Guardrails in Place?
An ongoing issue with all AI models is that they sometimes "hallucinate" - meaning they make up information without proper context.
To address this, Klarna has implemented serious guardrails to prevent their assistant from going rogue.
For starters, the AI is programmed to stick to strict guidelines. When it doesn't know the answer or isn't sure, it doesn't guess - it hands things over to a human.
That means less risk of errors, especially in the fintech world, where a simple mistake or wrong answer could cost someone money.
The team at Klarna has also established a strong verification system. The AI's responses are cross-checked against a whitelist of approved topics, ensuring it always answers what it knows versus what it doesn't.
Good job Klarna (not like they needed my validation)
Klarna's AI vs. Traditional Automation - What's Different?
"Wait, didn't companies automate customer support years ago?"
Good point, but no, not really
In the fintech/finance space, automation traditionally looked a lot more like those annoying automated phone systems that ask you to press 1 for this, 2 for that, and so on.
Klarna is trying its best to improve the customer experience. How?
Unlike traditional automation, Klarna's AI doesn't follow a rigid script or push people down a set path.
Instead:
- It actively engages with customers in conversation
- It adds nuance, understands context, and is always to the point
- It pulls real-time customer data to hyper-personalize support
The customer support we've always wanted?
What's Next?
Well, a lot of things, to be honest
Today, AI means faster replies, fewer mistakes, and a smoother way to meet customer needs.
Long hold times and poor customer support will no longer be acceptable; this level of service is becoming the norm.
But will only mammoths like Klarna benefit from this?
I doubt it, as this technology is getting more affordable by the minute, allowing smaller businesses to jump in too.
AI tools like Klarna's assistant will become cheaper and easier to use, helping businesses of all sizes offer fast, personalized customer service.
Klarna may be leading the way, but this is only the beginning for everyone else.
Interestingly, Sebastian (CEO of Klarna) doesn't just want to solve Level 1 (L1) support; his next step is to tackle more complex problems. He wants his AI to deal with tougher, more nuanced situations without making mistakes.
They're truly pushing the limits of what AI can do!
Exciting times ahead
Closing Thoughts
What Klarna has achieved will serve as a blueprint for other companies aiming to transform their customer support.
Companies need to think critically about how they implement AI and what that means for their employees. The key is striking the right balance between innovation and the human workforce.
Final Takeaways
Start thinking about how AI can improve your processes. Klarna's success story shows that even in industries where customer service is critical, AI can make things better, faster, and more cost-effective.
The sooner you adapt to this new reality, the better prepared you'll be for what's coming.
If you're evaluating AI for your customer support team, we can help you determine if Chat Aid is the right solution for your needs:
Book a demo with our team or sign up for a free trial today.