Klarna’s GenAI Journey: A Case Study Using the ‘AI Native’ Framework
Is Klarna Thinking Like An AI Native?
Narratives matter
These are the kind of headlines dominating the news:
Klarna chatbot doing work of 700 staff after AI-induced hiring freeze
What Klarna’s GenAI-Powered $40M Increase in Profits Means for Enterprises
Klarna to use AI to halve workforce
AI ‘efficiencies’ boost Klarna profits and marketing frequency
In addition to these headlines, social media posts are conveying the narrative that Klarna is aggressively deploying AI (umbrella term) through their tight collaboration with OpenAI to replace humans. Klarna`s PR aims to show the world the impact of this type of automation.
Facts matter
Facts reported by the private company which is in pre-IPO phase are:
- 73% increase in REVENUE per employee
- Revenues increased 27% year-on-year
- Headcount was cut from 5,000 to 3,800 year-on-year
- $40 million was saved (confusion on whether these are revenues or profits) due to AI chatbot implementation in Customer Support.
These are facts but what is not clear is to what we can attribute the efficiency leaps and the business strategy. Let’s also keep in mind that pre-IPO narratives should be taken with a grain of salt.
Perceptions matter
The verdict is still out there as to what kind of business is Klarna in and therefore, how public markets will value it. Is Klarna a Fintech — the BNPL early pioneer and disruptor of unsecured embedded credit — or is it an innovator in the shopping marketplace segment, or some hybrid platform business that keeps evolving? The business narrative that is currently being pushed through the media clearly aims for the Fintech crown. The proof is in the pudding with headlines like:
Klarna Takes on JPMorgan and BofA With Foray Into Bank Accounts
Klarna Expands Financial Services With New Checking Account, Cashback Offering
In reality, this new Klarna offering is a reward card similar to the wildly successful Starbucks (Starbucks has $1.77 Billion in unredeemed gift cards currently — see here ).
Asking the right questions, matters
Let’s make sure we distinguish revenues from profits. Most media articles allude to $40 million profit increase in the near future due to the deployment of AI chatbots.
I can only assume that the business goal is to save $40 million in Expenses.
From ALL the information disclosed, we can only infer that AI is deployed in Customer Support. I am sure if there were other areas beyond the Customer layer of Klarna`s business — for example within the enterprise or the ecosystem — they would have boasted about it. Even in the customer area, deployment seems very limited to contact centers.
According to the Pragmatic Engineer authored by Gergely Orosz (currently №1 newsletter on Substack with 500,000+ subscribers) Klarna`s AI deployment in customer support is really not that revolutionary.
- Was Klarna overspending on Customer support?
- Is the Klarna user interface inefficient resulting in higher customer support needs?
Gergely Orosz has tested it personally and reports that the current Klarna chatbot can only handle basic queries and straightforward questions (Level 1 support). Level 2 and Level 3 support are handled by trained staff. A user with a rare and complex issue would often go through L1 and L2 support to reach an agent who can help.
This means that Klarna was operating inefficiently and spending $60 Million per year on Customer Support. The $60 million inference comes from the reported $40 million savings from deploying an AI chatbot for Customer support (accounting for 2/3 of the staff).
More facts: Klarna`s CEO disclosed in February this year that Klarna has been using 3,000 FULL-TIME Customer agents via outsourcing.
So, the next question is `How much were these outsourced customer support agents costing? `
If Klarna `saved` 700 full-time agents per shift (as the media headlines report) accounting for 2/3 rds of their customer support and saved $40 million; then we can safely assume that per 24 hours (as Klarna is global and operates in 23 markets, 24/7) they saved 2,100 agents (three 8-hour shifts) which were paid c. $19k per year.
We can conclude that Klarna is now operating with 900 agents and probably paying them more than $19k per year. A reasonable estimate of the current total Customer Support spend would be $20 million (c. $22k per agent).
The Potential for Intelligence
Klarna’s AI Assistant is available in 23 markets, operates 24/7, and communicates in over 35 languages. Klarna`s emphasis on multi-language support does actually stand out. In comparison, Microsoft Copilot for Microsoft 365 AI chatbot customer support is in 28 languages and Amazon Lex in 27 languages.
Klarna`s AI assistant (as the favorite guinea pig of OpenAI) will evolve and integrate customer support Level 2 at least. The operating cost savings from this level of AI deployment of course are not going to be as steep as Level 1 which is actually a low-level automation deployment rather than anything revolutionary.
The potential Intelligence from a properly designed LLM-powered Customer support chatbot with Level 2 and Level 3 integration, are underestimated and not well understood.
Harvesting the learning from the language-based customer support inquiries at level 2&3, can be extremely valuable in improving the User Interface, for example. If all the LLM intelligence gathered from Level 2&3 customer inquiries stays siloed within the Customer support walls, then we remain stuck in the current business thinking and we aren’t advancing towards an AI-native business model.
We need to recognize that even if AI use cases are confined only in the Customer area, Customer support deployments are just scratching the surface of the potential for Intelligence. What if, we deploy GenAI to extract intelligence from the Customer Support area to those designing the User Interface, to those focused on Supply chain efficiencies and Human resources issues?
Klarna has not even scratched the surface of deploying GenAI in the Customer Area.
I can confidently say that not even 1% of what is possible is done. Just think of the Klarna Customer Area as a shopping marketplace and the extent that Klarna could integrate GenAI for contextual discovery helping customers make their shopping decisions, and lowering the customer acquisition cost. This is just exploiting technology at the Customer Discovery dimension, similar to how Instacart has done with `Ask InstaCart`.
Think of Klarna using GenAI at the Customer area to extract Intelligence across various countries and improve personalization country by country in all business areas — Marketing, Sales, Supply chain, Human resources, and Finances.
In our framework for starting the AI-Native Transformation journey, we emphasize that AI will enable complex and intelligent business archetypes to emerge. For now, we need to start the journey by breaking down the use cases in the three main areas and across five dimensions but always keeping in mind that a lot of value lies in the interconnections of this matrix/framework.
The outline of our framework from `Thinking like an AI-native (co-authored with Hari Abburi) from our new Wiley book: 𝐓𝐡𝐞 𝐅𝐚𝐬𝐭 𝐅𝐮𝐭𝐮𝐫𝐞 𝐁𝐥𝐮𝐫𝐬: 𝐃𝐢𝐬𝐜𝐨𝐯𝐞𝐫 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐯𝐞 𝐈𝐧𝐭𝐞𝐫𝐜𝐨𝐧𝐧𝐞𝐜𝐭𝐢𝐨𝐧𝐬 𝐒𝐡𝐚𝐩𝐢𝐧𝐠 𝐭𝐡𝐞 𝐅𝐮𝐭𝐮𝐫𝐞 captures the breadth and potential of such a transformation.
Klarna`s current positioning in the Transformation journey towards becoming an AI-native business is by no means revolutionary or transformational for its business.
Explore AI Opportunities with Our Guidance
If you’re considering how to effectively integrate Generative AI into your business strategy, we invite you to explore our comprehensive framework. Our workshops are designed to provide insightful evaluations of your current roadmap alongside introducing our pioneering framework.
Learn how we help companies and leaders transform towards becoming AI Native and reach out to us.
📌 Subscribe to my YouTube Channel with my insights and industry leader interviews. New video every Wednesday: https://www.youtube.com/EfiPylarinou
📌 Twitter: https://twitter.com/efipm
📌 Apple Podcast 👉 https://buff.ly/3P1Cp1Z
📌 Spotify Podcast 👉 https://buff.ly/3xPyWaV
📌 Linkedin: https://www.linkedin.com/in/efipylarinou/
📌 TikTok: https://www.tiktok.com/@efiglobal