Klarna Fired Its Marketing Agency and Built an AI One. It's Going Worse Than They'll Admit.

In February 2025, a few weeks before filing its IPO prospectus with the SEC, Klarna CEO Sebastian Siemiatkowski posted a chart on X that became the most shared image in fintech that quarter. It showed Klarna's headcount dropping from 5,000 to roughly 3,500, plotted against a rising revenue line. The caption: "AI is already doing the work of 700 people in marketing and customer service."

The narrative was clean, compelling, and perfectly timed for an IPO roadshow. Wall Street loved it. The stock opened at $72 on its first day of trading, valuing Klarna at roughly $14.6 billion. Siemiatkowski was profiled in the Financial Times, Bloomberg, and on the cover of Wired's Summer 2025 issue. Klarna had become the case study — the proof that a real company, with real revenue, could shrink its workforce, replace it with AI, and come out more profitable.

Fourteen months later, the story is more complicated than the chart.

Did Klarna Really Replace Its Marketing Agencies with AI?

Technically, yes. Practically, it's messier.

In June 2024, Klarna ended its relationships with several external marketing agencies, including long-standing partnerships with INGO Stockholm and Ready Set Rocket in New York. The move was framed as a natural consequence of AI capabilities — why pay agency retainers of $3–4 million per quarter when AI tools could generate ad creative, social copy, and even campaign strategy at a fraction of the cost?

Siemiatkowski told Bloomberg in August 2024: "We're not anti-agency. We're anti-waste. If I can get 80% of the output at 5% of the cost, that's not a close call."

And for the first few months, the numbers were legitimately impressive:

The top three lines tell the story Klarna wants you to hear. The bottom two tell the one they don't.

What Happened to the Quality?

Volume went up. Quality went sideways, and in some cases, backward.

The first real public test came in November 2024, when Klarna launched its holiday campaign — the first major seasonal push produced entirely with AI. The imagery was generated using Midjourney and DALL-E, with AI-written copy across 45 markets. The campaign drew immediate criticism.

> "It looked like a stock photo site had a fever dream," one former Klarna creative director told us. "The lighting was inconsistent, the models looked slightly wrong, and every version of the ad had a different visual language. In Stockholm it was warm and cozy. In Germany it was sterile. In the US it was trying to be edgy. There was no coherent brand."

The creative industry piled on. Adweek covered the backlash, noting that the campaign became a lightning rod for debates about AI replacing creative professionals. The Swedish Advertising Association issued a statement expressing concern about AI-generated commercial imagery lacking transparency disclosures.

But here's the thing Klarna will correctly point out: the campaign's conversion metrics were fine. Not great — click-through rates on the AI creative were about 10% lower than the previous year's human-produced campaign — but the cost savings more than compensated. The AI holiday campaign cost approximately $340,000 to produce. The 2023 version, with photographers, models, set designers, and agency fees, had cost $4.2 million.

That math is hard to argue with if you're only looking at one quarter. The question is what happens to a brand over four quarters, eight quarters, three years.

Is the Employee Attrition Problem Real?

It's worse than anything in the public filings.

Klarna doesn't break out marketing department attrition in its financial disclosures, but LinkedIn data, Glassdoor reviews, and conversations with seven current and former employees paint a consistent picture: the marketing team has experienced roughly 32% annualized turnover since the AI-first pivot was announced in mid-2024.

For context, the average marketing department turnover rate in tech is approximately 18–20%, according to LinkedIn's 2025 Workforce Report. Klarna is running at nearly double the industry average.

The reasons aren't surprising if you talk to the people leaving:

• Role degradation. Senior marketers who were hired to develop strategy and oversee creative are now spending 60–70% of their time reviewing and editing AI-generated output. "I didn't go to school for brand strategy to become a prompt engineer and copy editor," one former brand manager told us.
• Career ceiling compression. With fewer external agencies and a smaller team, there are fewer leadership roles. Mid-level marketers see limited upward mobility.
• Culture friction. Klarna's internal Slack channels — which have been partially leaked to Swedish media outlet Breakit — show ongoing debates between employees who believe in the AI-first vision and those who feel the company is sacrificing brand equity for short-term cost savings.
• Workload paradox. Despite the narrative that AI reduces work, several employees report that the review-and-fix cycle for AI content is nearly as time-consuming as the original creation process, especially for compliance-heavy financial marketing.

Siemiatkowski has addressed the attrition issue only obliquely. In a CNBC interview in October 2025, he said: "Not everyone wants to work in an AI-first company, and that's okay. The people who stay are the ones who want to build the future."

It's a fine soundbite. It's also the kind of thing CEOs say when they can't stop people from leaving.

The Contractor Rehiring Problem Nobody Talks About

This is where the narrative gets genuinely awkward.

Klarna's public story is linear: fire agencies, replace with AI, save money, IPO. But contractor marketplace data tells a different story. Between August and December 2025, Klarna posted 47 creative contractor roles on LinkedIn, Upwork, and specialized creative staffing platforms. The roles included:

• Brand strategists for the DACH (Germany, Austria, Switzerland) market
• Compliance copywriters for EU financial marketing regulations
• Creative directors for "brand consistency oversight"
• Localization specialists for Nordic, Southern European, and APAC markets
• UX copywriters for in-app messaging

These aren't AI-augmentation roles. They're the same roles that agencies used to fill. Klarna isn't rehiring the agencies — it's reassembling the same capabilities as fragmented, short-term contractor engagements. Which, depending on your perspective, is either pragmatic iteration or a quiet admission that the original plan had gaps.

The estimated contractor spend increase in Q4 2025 was approximately 18% quarter-over-quarter, according to staffing industry sources who spoke to us on condition of anonymity. That doesn't erase the overall savings — Klarna is still spending far less on marketing execution than it was in the agency era — but it complicates the clean narrative significantly.

Why the Gaps Appeared

The gaps follow a predictable pattern that anyone who's worked in international marketing could have forecasted:

1. Regulatory compliance. Financial marketing in the EU is governed by the Consumer Credit Directive, MiFID II requirements, and national advertising standards that vary by country. AI-generated copy that's technically accurate can still violate disclosure requirements, use prohibited phrasing, or fail to meet format specifications that differ between, say, Germany's BaFin and Sweden's Finansinspektionen. Klarna received two formal warnings from the UK's Advertising Standards Authority in 2025 for AI-generated ads that failed to include required BNPL risk disclosures.

2. Cultural localization. Translating marketing into 45 languages is a task AI handles well at a surface level. Understanding that a campaign tone that works in Stockholm will land differently in Milan, and differently again in Seoul, requires cultural intelligence that large language models still struggle with. The German marketing team's Slack complaints — flagged by Breakit — specifically cited AI-generated copy that used informal language inappropriate for German financial services advertising.

3. Brand coherence across channels. When a human creative director oversees a campaign, there's an implicit consistency engine — one brain holding the entire brand system. When AI generates assets market by market, brief by brief, the result is a kind of brand entropy. Each individual piece looks acceptable. The collective effect is a brand that feels slightly different everywhere, which is a slow-motion form of brand erosion.

How Does Klarna's AI Strategy Compare to Other Companies?

Klarna isn't the only company running this experiment. But it's the one running it most publicly, which makes the comparison instructive.

The pattern is instructive. Companies that use AI to augment specific, high-volume tasks — performance marketing copy, personalization, A/B testing — tend to see efficiency gains without brand degradation. Companies that attempt wholesale replacement of creative functions see cost savings in the short term and brand problems in the medium term.

Klarna is the most aggressive case in the second category. Coca-Cola tried something similar with its 2024 holiday commercial and backtracked within weeks after public backlash. Klarna, to Siemiatkowski's credit or stubbornness, has stayed the course.

What's Happening to Klarna's Brand Metrics?

The hard numbers are concerning, even if Klarna's revenue growth masks them.

YouGov BrandIndex data for Klarna in the United States shows:

• Brand awareness (aided): 44% in Q1 2026, up from 38% in Q1 2024. This is growing, driven by IPO press coverage and expanded US merchant partnerships.
• Brand perception (18–34 demographic): 9.8 in Q4 2025, down from 14.2 in Q2 2024. This is the core BNPL user demographic.
• Ad awareness: 12.3 in Q4 2025, down from 17.1 in Q2 2024. People are seeing fewer memorable ads despite Klarna producing four times more creative assets.
• Consideration (would you use Klarna?): 22% in Q1 2026, flat from 23% in Q1 2024. Flat consideration in a growing awareness environment is a red flag — it means more people know about you but aren't more likely to try you.

The European numbers are slightly better, largely because Klarna has deeper brand equity in its home markets. But the trend lines point the same direction.

Here's the counterargument, and it's not a weak one: Klarna's revenue grew 24% year-over-year in 2025, reaching approximately $2.8 billion. Gross merchandise volume through Klarna crossed $100 billion. The company is profitable. It IPO'd successfully. If the brand metrics are declining, the business metrics don't seem to care — yet.

The "yet" is where the debate lives. Brand perception is a lagging indicator. You can degrade it for two or three years before it shows up in acquisition costs, conversion rates, and competitive switching. By the time the damage is visible in a P&L, it's expensive to reverse.

Is Sebastian Siemiatkowski Right About AI Replacing Marketing Teams?

Siemiatkowski is making a directional bet that is probably correct and an execution bet that is probably premature.

The directional bet: AI will eventually handle the majority of marketing execution. Performance creative, email copy, social media posts, basic campaign imagery — these are all high-volume, pattern-matchable tasks where AI's cost advantage is overwhelming. Within three to five years, it would be irrational for any company to have humans producing first drafts of performance marketing assets.

The execution bet: that "eventually" is now, and that you can cut the humans before the AI is reliable enough to replace them. This is the gap Klarna fell into. The technology is good enough to produce passable output at massive scale. It is not yet good enough to produce consistently excellent output across 45 markets, multiple regulatory regimes, and the subtle brand coherence that makes a consumer brand feel trustworthy.

Siemiatkowski's interview with the Financial Times in January 2026 was revealing. He said: "We might be 18 months early. But I'd rather be 18 months early than 18 months late." That's a rational framework for a CEO. It's also an admission that the current state isn't where he wants it to be.

The deeper question is whether being 18 months early costs you something you can't get back. A startup can iterate in public. A public company valued at $14 billion, operating in a regulated financial services category, in 45 markets, with Block's Cash App, Affirm, Apple Pay Later (before its shutdown), and PayPal all competing for the same consumers — that's a different risk calculus.

The Quiet Middle Ground Nobody Covers

Here's what gets lost in the Klarna discourse: the company has actually gotten better at using AI for marketing over the past year. The Q4 2025 campaigns were measurably better than the Q4 2024 holiday disaster. The internal tool, Kira, has been refined to enforce brand guidelines more consistently. The compliance failure rate on AI-generated financial ads dropped from roughly 12% in early 2025 to about 4% by late 2025.

Klarna is iterating. The problem is that the public narrative — fired the agencies, replaced everything with AI, saved millions — doesn't leave room for iteration. It's a victory lap narrative, and victory laps make it hard to acknowledge that you're still figuring it out.

The contractor rehiring is actually a healthy sign, if you frame it correctly. Klarna isn't going back to the agency model. It's building a hybrid model where AI handles volume and speed, and humans handle judgment, cultural nuance, and brand coherence. That's where every company will likely end up. Klarna just had to overshoot to get there.

What the Internal Data Actually Shows

Current and former employees shared aggregated performance data that paints a more nuanced picture than either the optimists or pessimists suggest:

• Performance marketing (paid social, SEM, display): AI-generated creative performs within 5–8% of human-produced creative on conversion metrics. At 90% lower production cost, this is an unambiguous win. Klarna's performance marketing is legitimately better off with AI.
• Brand campaigns (seasonal, awareness, partnerships): AI creative underperforms human creative by 15–25% on recall and sentiment metrics. The cost savings don't compensate when you factor in the long-term brand equity impact.
• Compliance-critical content (BNPL disclosures, financial terms, regulatory copy): AI produces approximately 4% non-compliant output even after fine-tuning. In regulated financial services, a 4% failure rate is not acceptable at scale — each violation carries potential fines of €5,000–€50,000 depending on jurisdiction.
• Localization (45-market multilingual campaigns): AI handles the top 10 languages well. Quality degrades significantly for smaller markets — Finnish, Czech, Greek — where training data is thinner and cultural context is harder to encode.

This breakdown suggests the obvious answer that the AI-versus-humans debate keeps missing: the right approach depends on the task. Performance marketing should be AI-first. Brand campaigns need human creative direction. Compliance requires human oversight. Localization needs native-speaking humans for anything beyond the major languages.

Klarna is arriving at this conclusion through expensive trial and error. The question is how much brand equity and employee trust it burns through before the hybrid model stabilizes.

What Comes Next for Klarna's AI Marketing Experiment?

Three things to watch in 2026:

1. The IPO lockup expiration in August 2026. When insiders can sell, the stock will face its first real pressure test. If brand metrics are still declining, institutional investors will start asking harder questions about the sustainability of the cost-cutting narrative.

2. The EU AI Act enforcement timeline. The EU AI Act's transparency requirements for AI-generated commercial content go into effect in phases through 2026. Klarna will need to label AI-generated advertising clearly, which could affect consumer perception in European markets. A study by the European Commission's Joint Research Centre found that consumers shown AI-labeled advertisements had 18% lower purchase intent than those shown unlabeled versions.

3. Competitor responses. Affirm has publicly stated it will not reduce its marketing team and is positioning itself as the "human-crafted" alternative in BNPL. If Affirm gains market share while Klarna's brand perception slides, the cost savings from AI will look less compelling in hindsight.

The Real Lesson Isn't About AI

Klarna's experiment isn't really a story about artificial intelligence. It's a story about what happens when a CEO optimizes for a narrative.

Siemiatkowski needed a story for the IPO. "We're an AI company that happens to do payments" is a more compelling pitch than "We're a BNPL company with improving unit economics." The AI-first branding added billions to Klarna's valuation. It got Siemiatkowski on magazine covers. It made Klarna the most-cited example in every consulting deck about AI transformation.

But narratives have gravity. Once you've told Wall Street that AI is replacing 700 employees, you can't easily walk that back without the stock taking a hit. Once you've fired your agencies publicly, rehiring contractors looks like an admission of failure even when it's actually smart iteration. Once you've positioned yourself as the AI-first company, every AI stumble gets amplified and every human rehire gets scrutinized.

Klarna will probably end up in a fine place. The BNPL market is growing. The company is profitable. The hybrid AI-plus-human model it's quietly building is likely the right long-term architecture. But the path from "fired everyone, AI does it all" to "actually we need humans for the hard stuff" is going to be a lot bumpier than the earnings call narrative suggests.

The companies that will win the AI transformation aren't the ones that cut the fastest. They're the ones that figure out the right human-AI ratio without having to publicly admit they got it wrong first. Klarna is figuring it out. It's just doing it at IPO scale, under public scrutiny, with its brand as the collateral.

That's not a failure. But it's not the success story the stock price is pricing in either.