How Klarna is using AI for cost savings, changing ‘extremely frustrating’ creative processes

Throughout the first quarter of 2024, Klarna used AI to generate 1,000 images using generative AI tools like Firefly, Midjourney and DALL-E. During that same time period, Klarna decreased its spending by 25% on external marketing suppliers responsible for services like translation, production CRM as well as social agencies.

It is using those savings to invest more in “working” areas like paid media. CEO and co-founder Sebastian Siemiatkowski noted these savings in a now viral tweet about the company’s use of AI in marketing as well as its ability for its smaller marketing staff to reportedly work more efficiently with the use of AI.

To get a better sense of how Klarna is using AI, what its savings actually look like and how the company’s marketing organization is functioning now, Digiday caught up with CMO David Sandstrom, who noted that the company had not reduced its marketing headcount but had seen a reduction in marketing staff due to normal churn as well as choosing to re-hired in other areas.

The conversation below has been edited and condensed for clarity.

Earlier this week, there was a now viral tweet from CEO and co-founder Sebastian Siemiatkowski that stated Klarna had saved $6 million to produce images by now using AI. Is that the total Klarna was spending on images prior to using AI?

Yes, that’s the total. All of the money spent on ideation, which is usually, at least in the old world, was one of the bigger hassles, in my opinion, when it comes to creating marketing. You talked about ideas but there was never a way to really visualize ideas without paying quite a lot of money. You start visualizing ideas. Scrap them. Then we visualize more. Scrap them. So what we’ve done with AI in the ideation stage has been extremely helpful to get ideas up on the table as quickly as possible. And then in the creation stage we do loads of images. Bear in mind that we’re active in 45 markets and we have hyper-personalization for our consumers. We do a lot of imagery. 

What is the cost now to create imagery? If you’re cutting out stock imagery like Getty, was that included in the $6 million saved? 

I have to be honest with you, I’m reading his tweet for the first time. I’m not sure we had a Getty image account. I’m not sure he knows the details of how we operate fully within marketing. I will not be able to tell you what our cost for Getty images is if we had a stock photo account like that. You probably think I run his account. 

I don’t think that. … How are you measuring customer response to AI images?

It might be hard to say that it’s exactly the AI images. What we do within marketing when it comes to AI, AI imagery is a small proportion of that. Our ambition is that we have hyper-personalized communication to 150 million consumers in 45 markets in 14 languages. We have not been able to produce at that scale.

What AI has helped us do is personalized messages. It would be hard for me to say, yes it’s the imagery or yes it’s this. But what we do know is that our ability to almost in real time be able to cater to your interests, talk to you about your interests, products, brands, merchants that you might be interested in versus mine that differ gives a huge uplift in engagement. What we’re seeing is much higher engagement at a much lower cost. We’re obviously monitoring things like social, reviews, customer support. Nothing there. People think our campaigns are as beautiful as ever. 

If there was a customer response where it seemed like they weren’t happy with the use of AI would there be any change in what you’re doing? 

I don’t know. If consumers complained about our use of Photoshop would we have changed Photoshop? Or if they complain — I don’t think that’s the case — AI is a tool for us that we use to enable marketeers to be way more efficient and get way more done at a bigger scale than ever before. That is what AI is. We do not — and we don’t have the ambition to — press a button and go grab a coffee and when we come back there’s a fully fleshed out campaign ready to go. That’s not how any of this works. We still have very talented marketeers, very talented designers. We’ve empowered them throughout their workflow to utilize AI to craft something amazing.

What is the cost of using Midjourney, Dall-E or other tools? What does that account for with the budget? 

It’s a fraction. We obviously have enterprise accounts with the providers. There’s a cost to the enterprise account, there’s usually a cost to the API call that we’re making but it’s an absolute fraction of the cost of an agency.

Is there any worry of becoming too reliant on these services if those costs go significantly up? 

You mean they would increase the prices because we’ve saved money? We haven’t thought about that. Also, we build logics on top of these platforms, which means we’re not deeply integrated into one of these platforms. This field is developing on a weekly basis. I don’t think this is the right time to go all in and put all of your eggs into one of these platform baskets. You see that in other fields where you don’t want to be platform dependent. Just look at the creator space. When Instagram changes their algorithm a lot of people lose out. We’re obviously not platform dependent. We’ve built our massive logics on top and we’re agnostic to these platforms. That also means we can be fairly flexible when new services, new platforms, softwares pop up and they pop up on an almost hourly basis. 

Have you added team members to be responsible for the logics you’re building? 

Yeah, of course. There are a lot of new jobs in general in the world popping up as a consequence of what’s happening. By the end of the year, I think we’re going to see a significant amount of new job descriptions, everything from prompt engineers to AI ethics managers to you name it.

How are you breaking up your marketing budget now? Has the overall media allocation shifted as you’ve been using more AI assets? Where are you spending your ad dollars?

One thing that hasn’t changed is the price of distribution. We’ve been putting a lot more money and effort into media and distribution. One of the dimensions we look into is working and nonworking capital. Nonworking capital is obviously the money we spend on agencies, on production, on strategy, on translations, on the things that don’t actually reach the consumer. Then we have working capital, which means this money is actually moving the needle for our P&L and our numbers. We have an ambition to have as big of a proportion as possible working for us. AI has been great in that sense. It has enabled us to move more money into working capital. 

What is the breakup of your ad spend right now? 

We tried all sorts of things in the past but now we’re reverting back to more traditional marketing. I would say maybe we do 20% social, 40% offline including TV, linear, OOH and print and then 40% digital video. That’s a very rough estimate. We were probably much bigger on social [prior]. 

Are these assets made by AI and in-house teams? Or are these created by external agency partners? 

It depends. Again, AI is a tool we work with throughout. It’s not like here are the AI things and here are the human things. AI is a tool that we work with throughout. We do that with everything from our CRM send outs to the bigger campaigns we’re doing. A lot of them have started with an idea that has been sketched and mocked with AI. It’s across the line. AI is still fairly weak on video. But that is probably going to change in the next quarter. That’s the place if you exclude avatars that’s not as advanced. 

Why are you going back to a more traditional mix? 

The brand awareness and reach side of our marketing plan is prioritized right now. 

Would you say the reductions in spending you’ve made due to AI are being reallocated into some of the more traditional media now? 

Yes, absolutely. Our ambition is to free up money so we can invest in things that actually perform for us.

The end result, the video we’re seeing or the content that we’re seeing on TV for brand awareness, those aren’t yet AI videos are they? 

Not on the video side. But on the image side, yes. I think about this differently than you do. To me it’s not binary. It’s not a computer, press a button, out goes the campaign. Have we utilized AI to create a lot of the imagery, even for traditional campaigns? Yes we have. But that doesn’t mean that the image is completely AI created. There are a lot of humans who have put their hands on that. We have creative directors, everything that happens in post-production, there’s so much happening there. We’re not doing anything new with AI. Nothing new. The only thing we’re doing is that we’re doing it faster and at a lower cost.

At a lower cost because of what? 

Because of a lot of things. One simple example is how much money we’ve been wasting in the past due to producing things we didn’t know would turn out good. When you do something completely offline without AI, the only thing you have is a description of what this might become. You have an idea in someone’s head. The idea in your head is different from the idea in my head. In order to get it out of your head, I have to spend a lot of money, maybe do a couple of test shoots, maybe do prototyping, storyboards, something and three weeks in and $100,000 later I might see it and not like the idea. Now what we do is sketch it within an hour. We look at it jointly. I might say I like it. That can still then lead to a photoshoot. We might shoot this. But there is so much money to be saved. I don’t think you should think about this as some sort of binary process where it’s either/or. There is a way to improve all of our existing processes. Although a lot of people seem to be fans of current creative processes, they are severely broken. And they are frustrating for everyone, for the client, for the agency … Most of the processes are extremely frustrating to everyone. 

It’s a change in the process to avoid a full production of something you might scrap if you didn’t like how it turned out.

Ask these big companies who are spending a ton of money. Ask Procter & Gamble or Pepsi how much they’ve spent on storyboards. Storyboards are drawn images by humans that are made to be thrown away. Or pitch decks. Whatever it is. There is so much waste in this industry which means that all the humans who actually work in this industry will be able to work on things that actually matter, that actually will be produced, that will actually have an impact instead of doing all this unnecessary work we’ve been doing for years.