Klarna CEO Sebastian Siemiatkowski on AI, Software and the Future of Banking

Klarna cofounder and CEO Sebastian Siemiatkowski joins Harry Stebbings to discuss AI, software, customer service, fintech competition and his experience building the company. His claims about Klarna and competing products reflect his position as its CEO; financial figures and predictions are not independently verified in the interview.

Siemiatkowski argues that AI is dramatically reducing the cost of producing software, and predicts that agents will also make it easier to move data between providers. Today, he says, customer data and its vendor-specific structure create switching costs even when an interface can be replicated. The proposed one-click migration is his forecast, not a demonstrated capability for every system.

He believes lower switching costs could pressure software valuations, while acknowledging that established products will not disappear overnight. He compares historical price-to-sales multiples with utilities and offers a possible lower range. These are time-specific approximations and speculative valuations, not an estimate of any particular stock’s fair price.

Stebbings challenges the assumption that large enterprises will build critical systems themselves, citing permissions, hierarchy and operational needs. Siemiatkowski expects reusable open-source components and standardized building blocks to reduce the effort, but agrees small businesses may buy packaged services rather than write their own software. How safe and reliable such systems become remains an open question.

The CEO describes an experiment he calls a company in a box: an agent on top of open-source accounting and CRM software. He says he tried tasks such as recording an invoice and creating a customer account. The example is his own small test, not proof that an agent can reliably or lawfully replace professional accounting in all circumstances.

For Klarna, Siemiatkowski says the goal is to give AI better context by reducing fragmented data and rethinking the bank’s technology stack. He says the company still uses some external software, including Slack. Stebbings asks whether agents that connect systems might also reduce the need to build everything internally; the guest says the eventual product model is uncertain.

On customer service, Siemiatkowski recalls an early AI deployment that he says handled work equivalent to roughly 600 agents. He emphasizes that many questions were simple and says the people worked for outside service companies and were reassigned. His account of work volume and employment outcomes is not an independently verified headcount audit.

He argues that good support requires context about the company’s products, potentially including how its software implements rules such as interest calculations. Klarna concluded it needed to develop its own system, he says, although other enterprises may buy strong third-party tools. He does not claim that all businesses must follow Klarna’s approach.

Stebbings asks about the reaction to headlines describing hundreds of jobs replaced. Siemiatkowski says some people were understandably angry and disputes the characterization that Klarna abandoned AI. He also says the company had placed too much emphasis on cost and wanted to ensure that customers who prefer human contact could obtain it.

Stebbings questions whether human service will automatically become thoughtful and personalized. Siemiatkowski describes recruiting enthusiastic Klarna customers for flexible support work and claims high satisfaction for those interactions. The interview does not independently assess the program’s working conditions, customer outcomes or comparison with other support channels.

Stebbings says his investment strategy increasingly favors products that replace labor rather than charge per seat. Siemiatkowski expects a significant employment transition and possible short-term turmoil, while hoping AI eventually supports a wealthier society. Both the host’s investment interest and the guest’s employer interest matter; neither establishes a particular employment outcome.

Siemiatkowski recalls deciding in 2015 that Klarna should become a digital financial assistant after struggling in payments competition. His imagined assistant might help a customer find savings and manage paperwork. This is a product vision rather than an existing guarantee of financial advice, approval or savings.

Asked about Revolut, he says Klarna has around 110 million customers globally compared with roughly 65 million at Revolut, while acknowledging that Klarna users engage less frequently. These are his approximate customer counts at the time; the interview does not establish that the companies define or measure customers identically.

He says Klarna wants to turn an occasional payments relationship into more frequent banking use. Unlike a trading-led product, its entry point is shopping, and he describes the brand as closer to a digital lifestyle service. His claims about accelerating adoption are company statements, not a comparison independently established in the interview.

Siemiatkowski recalls declining an opportunity to advise Nubank founder David Vélez because he lacked time and jokes about missing an investment. He argues that fintech firms enter from different markets and may mainly take share from incumbent banks, although he acknowledges competition among the challengers. Both the missed-return idea and competitive outcomes are speculative.

He describes the United States as central to Klarna’s ambition to become a global financial assistant. He says transaction-level product information can help tailor recommendations and cites full digital receipts as an advantage. This is his account of the company’s data and strategy; useful personalization also depends on how data is handled and what customers authorize.

The CEO says US incumbents have improved digital services, making entry difficult, and estimates Klarna has nearly 30 million US users. He recalls roughly two to three million active US cardholders in the available prior-quarter figures, while warning that new earnings had not yet been released. His estimates should not be treated as current audited metrics.

Asked whether Nubank or Revolut will do better in the US, Siemiatkowski first backs his own company and then guesses Nubank may be more focused while Revolut is expanding across more markets. He praises both leaders. The comparison is the opinion of a direct competitor, not evidence of either company’s future performance.

On being a public-company CEO, he says Klarna already had substantial reporting obligations as a bank, although he would prefer full private ownership if that were possible. He describes a board plan to broaden banking services, including transfers and deposits, and says AI changed how much additional spending he expected to request.

Siemiatkowski says Klarna had fallen from a peak of more than 7,000 people to fewer than 3,000, primarily through attrition after earlier layoffs. He says he sought no additional investment for the expansion plan because he expected AI to help the existing organization deliver it. These are management’s characterizations, not proof that AI alone caused every staffing change.

Asked about 2030 headcount, he predicts fewer than 2,000 employees may be possible. He expects merchant relationships and human customer service to remain important and says the company still hires occasionally despite an overall decline. A projection is not a staffing commitment, and it does not establish how contractors or future business changes will be counted.

He estimates annual attrition at about 20% and says average employee compensation per head has risen almost 50% over the period he describes. He attributes the change in part to sharing productivity benefits. The conversation does not specify a full compensation methodology or establish individual employees’ experience.

Asked about stock-based compensation, Siemiatkowski sees large differences between American and European technology companies and says Klarna increased compensation to compete for talent. He predicts that technology and financial firms will face stronger competition as switching costs fall. His comparisons and estimates are broad opinions rather than a measured industry survey.

He expects some incumbent banks to adapt to digital services and AI while others struggle, depending on leadership and investment decisions. Discussing Goldman Sachs’s Marcus, he offers his own interpretation of the difficulties of defending a long-term initiative in public markets. He acknowledges that the people involved may disagree with his account.

On Klarna’s earlier high valuation, he clarifies that some shares were purchased at the cited $45 billion level. He says founders should watch whether valuation multiples are expanding faster than revenue, and regrets not being more cautious about hiring before later layoffs. His retrospective does not establish that any one valuation was appropriate.

Siemiatkowski describes how Sequoia’s Chris Olsen contacted Klarna, and says Sequoia took a 25% stake at a $100 million valuation. He recalls challenging Olsen about why more partners had not attended, followed by a call from Michael Moritz and an offer to join the board. This is the founder’s recollection of the transaction and relationship.

He praises Moritz’s ability to identify important details and recalls a 2019 conversation that sharpened Klarna’s focus on the US. He credits Sequoia’s people and brand with helping the company. These are the views of a founder whose company received Sequoia investment, not an independent assessment of investor performance.

Stebbings asks whether buy-now-pay-later is a difficult business. Siemiatkowski recalls Klarna’s early financing and says it achieved a long period of profitable growth. He also says he became concerned when late fees represented a material revenue line and decided to change the product rather than leave the business.

The CEO argues that Klarna’s fixed-installment offering can be healthier than revolving credit and says the company removed revolving credit in the Nordics, giving up revenue. He reports that some UK late fees returned after an earlier period without them. His comparison is a commercial claim from a lender, not a general determination that BNPL is safer for every consumer.

Siemiatkowski acknowledges that customers can overextend themselves and that a lender must help people in distress. He presents a combination of debit and occasional installment payments as his preferred alternative to credit cards. The interview does not establish outcomes across products, eligibility rules or individual borrowers.

Asked whether a European founder must relocate to the US, he says no and recounts a company’s unsuccessful attempt to move engineering to Silicon Valley after investor advice. He argues the appropriate location depends on the business and its talent. His anecdote is not a general test of every expansion decision.

Siemiatkowski believes AI investors should try building with current tools before evaluating companies. Stebbings predicts a steep decline for Cursor; the guest disagrees and says Klarna uses it alongside Claude Code for different tasks. Both are personal judgments, and the predicted revenue decline is not a reported result.

Asked to choose between hypothetical investments in Anthropic and OpenAI, Siemiatkowski says he prefers Anthropic’s product direction for his own needs, while describing OpenAI as more consumer-oriented. He wants an assistant willing to challenge his views. His characterization of products and company strategy is subjective and does not prove differences in accuracy or bias.

He says his investment vehicle Flat has made an investment in defense and that he now sees traditional software investing as riskier. Stebbings offers a bullish view of data-center demand. Their respective financial interests and forecasts are relevant to the exchange and do not establish future investment returns.

Siemiatkowski presents a personal theory of AI as a form of compression, comparing repeated enterprise data with a consolidated source of knowledge. He believes organizations could reuse information and code more effectively. His explanation of how model training stores information is a conceptual analogy, not a technical demonstration that models preserve all facts or eliminate duplication.

He sees two opposing possibilities for compute demand: enterprises may economize by reducing redundant work, while new entertainment and other generated content may require substantial additional capacity. He explicitly says he does not know which force will dominate. His suggested reduction in infrastructure need is therefore a conditional scenario, not a forecast of actual demand.

He uses Wikipedia’s approach to creating articles as an analogy for checking whether work already exists before starting it again. He expects AI eventually to help companies find and reuse existing information and code. He also cites an unverified claim about new search queries and acknowledges that he does not know how much novelty people create.

Asked what CEOs say privately, Siemiatkowski says he spends limited time in CEO gatherings and prefers working with his teams. He believes leaders should experiment with building. He describes using Claude to create an animated explanation of a complicated financial concept and says the combination of skills made a strong impression on him.

He praises Grok and describes using it to check claims circulating on X. That is his experience, not a controlled evaluation of accuracy or a guarantee that any model reliably verifies information. He also reiterates his denial of a rumor that Klarna was offering buy-now-pay-later for rent.

Siemiatkowski says he has revised his expectations about the pace of AI adoption: changing habits and workplace processes can take longer than he expected. Stebbings suggests consumers may move faster than enterprises, and the guest agrees. Their discussion distinguishes technology capability from actual adoption.

He rejects criticism that he is focused on a quick exit or indifferent to customers and says he cares about their financial well-being. He presents Klarna’s products as better than the alternatives. These are his responses to criticism and his commercial assessment, rather than independent findings about lending outcomes.

If Klarna were private, he says, he would spend less time communicating with investors but pursue the same financial-assistant strategy. He acknowledges that communicating a clear message is difficult, while Stebbings favors candid conversation over a rehearsed pitch. Both describe their own preferences.

He recalls the strain of a falling valuation and layoffs and describes an intense television interview during that period. Listening to Queen’s “Under Pressure,” he says, helped him frame the experience as part of the demanding career he had chosen. He acknowledges both difficult emotions and gratitude for the opportunity.

Siemiatkowski recalls an email written six months into Klarna imagining international expansion and competing with banks. Stebbings argues that fixed visions can constrain founders; the guest disagrees in his own case, while acknowledging he did not know how the early vision would be achieved.

He describes an immigrant childhood, family financial strain and early enthusiasm for entrepreneurs. As a child, he believed making money could resolve problems at home; he later learned that money did not solve his father’s alcohol dependence. He also recognizes the practical privilege that financial security gives his family.

Looking ahead, Siemiatkowski says he is optimistic that AI can improve lives but cannot know exactly what the world will look like in two years. His ambition is to build a financial service that helps customers save time and manage money. That is the CEO’s stated goal, not a guarantee of customer outcomes or future product performance.