Klarna CEO: SaaS Is Dead in an Agentic World

Why Klarna's CEO Thinks Every SaaS Company Is Overvalued

Sebastian Siemiatkowski is one of the few public company CEOs willing to say the quiet part loud. In this wide-ranging conversation with Harry Stebbings, the Klarna founder lays out his case for why AI agents are about to collapse the switching costs that have propped up SaaS valuations for two decades — and what he's doing about it inside his own company.

The cost of software is going to zero: "The cost of creating software is going down to zero. Everyone will be able to generate software at any point of time." But Siemiatkowski argues this is only phase one. The real disruption hasn't arrived yet — it's the death of data switching costs. When AI agents can migrate your entire CRM, accounting, and project management data in one click, the moats that SaaS companies rely on simply evaporate.

SaaS multiples have further to fall: "If you look at utilities, normal companies that are more utility, they may trade at one to two. Is it likely they could come down to one or two? Yes, I think so." He draws a direct line from Chegg's collapse (trading at 0.2x sales after ChatGPT) to what awaits enterprise software companies still trading at 5-10x. The age of 20-30x price-to-sales multiples for software is over.

The "company in a box" experiment: Siemiatkowski personally built a prototype where he put an open-source accounting system and CRM into a workspace, layered a Claude agent on top, and asked it to bookkeep invoices and manage customer accounts. "It worked really, really well. The plumbing firm of the future won't vibe code this themselves — they'll buy off-the-shelf products. But the winner will be extremely broad." His vision: AI agents sitting on top of open-source infrastructure replacing entire categories of business software.

Klarna cut 50% of its workforce through AI: The company went from over 7,000 to below 3,000 employees, mostly through natural attrition. Critically, Siemiatkowski didn't ask the board for a single additional dollar to simultaneously launch banking, trading, peer-to-peer payments, and international remittances. AI absorbed the work. Employee compensation per head grew nearly 50% during the same period — the gains were shared.

Customer service proved the thesis: Klarna's AI customer service replaced the equivalent of 700 agent jobs in its early deployment. But the deeper insight was about context: "Customer service isn't just 'I need an agent that answers questions.' Sooner or later you want it to read the source code and explain to the customer how it works." This drove Klarna to conclude that customer service must be deeply integrated into the tech stack, not bolted on from a third-party SaaS vendor.

6 Takeaways from Siemiatkowski on AI and Enterprise Software

• Data switching costs are the next domino — Once AI agents can migrate your data between vendors in one click, the SaaS moat disappears overnight
• Systems of record are vulnerable — ERPs, CRMs, and project management tools built on data lock-in face existential repricing toward utility-level multiples
• AI-native means one stack, not many silos — Klarna rebuilt its entire tech platform because siloed SaaS tools couldn't give AI enough context to perform well
• Headcount will keep falling — Klarna shrinks 20% per year through natural attrition, with no plans to backfill. By 2030, Siemiatkowski expects "even less" than 2,000 employees
• Human connection becomes the premium — The future of VIP customer service is human interaction, not AI. Klarna recruits its own customers as part-time support agents via an Uber-like model
• Per-seat pricing is dead — Investors should avoid any company selling per-seat. The winning business model displaces labor, not licenses

What Klarna's AI Transformation Means for Every Enterprise

Siemiatkowski's argument is not theoretical — he's running the experiment at scale inside a public company. The implication is clear: organizations that consolidate their data into AI-native stacks will outperform those still paying for siloed SaaS. And as AI agents eliminate the friction of switching vendors, the defensive moats that justified premium software valuations will wash away. The companies that survive will be the ones that wake up every morning fighting for their customers — just like a restaurant or a retailer — because the money printing machine in the basement is about to stop working.