Klarna 'Replaced' Salesforce With AI. Then Its CEO Told You Not to Copy Him.

If you've sat in any finance meeting in the last year, someone has used Klarna as the example: "they replaced Salesforce with AI, the era of buying software is over." It's become the go-to proof that you can rip out your stack and let an AI run the work instead. There's just one problem with that story - it's not really what happened, and the person saying so loudest is Klarna's own CEO.

This matters for you because the Klarna myth is quietly driving real decisions about accounting and finance technology. People are using a misread of one company's experiment to justify some genuinely risky calls about their own ERP, their close tools, their whole stack. And in accounting that's not a small thing - across the last 90 days of job postings we track, nearly 60% explicitly name a major system like NetSuite, SAP, Oracle, or Workday. Six in ten accounting jobs are defined by the software stack. So when someone says "AI is about to make all that irrelevant," it's worth knowing whether the story they're citing actually says that. It doesn't.

What people think happened

The popular version: in 2024 Klarna announced it was shutting down Salesforce and Workday, cut a big chunk of its workforce, pointed an OpenAI-powered chatbot at customer service that reportedly did the work of hundreds of agents, and declared software vendors obsolete. AI ate the stack. The end of SaaS. You've heard it.

Pieces of that are true - Klarna did move off Salesforce's CRM and off Workday, and it did go hard on AI in support. But the conclusion everyone drew from it is where it falls apart.

What actually happened, in the CEO's own words

Here's the part that didn't make the headlines. In March 2025, Klarna CEO Sebastian Siemiatkowski directly pushed back on the idea that his own move spelled doom for enterprise software. Asked whether other companies would follow Klarna off Salesforce, he said plainly: "I don't think it is the end of Salesforce; might be the opposite." And on whether everyone should copy the playbook: "Will all companies do what Klarna does? I doubt it."

Then he dismantled the core of the myth. Klarna didn't replace its software with AI. As Siemiatkowski put it, "we did not replace SaaS with an LLM, and storing CRM data in an LLM would have its limitations." What Klarna actually did was consolidate data out of various SaaS platforms into its own internal infrastructure - using graph database technology and other tools - and then let AI work on top of that consolidated data. And the things it "shut down"? It largely swapped them for other software: Workday for HR went to Deel, the CRM became a blend of third-party and in-house tools. It even kept its Salesforce partnership and kept using Slack.

So the accurate headline isn't "Klarna replaced its stack with AI." It's "Klarna re-architected its data layer and swapped some vendors, with AI running on top." Way less sexy. Way more true. And for anyone running an accounting function, it tells a completely different story about what to actually do.

The part nobody quotes: the customer service walk-back

There's a second piece that got buried. Klarna went all-in on AI for customer support - and then, in 2025, started rehiring humans, with Siemiatkowski acknowledging the all-AI approach had led to lower-quality service. The company that supposedly proved AI could replace its people quietly decided that, actually, some of those people were doing something the AI couldn't.

If you're a controller or CFO who's felt the pressure to "just automate the close" or "let AI run AP," sit with that. The single most-cited "AI replaced everything" case study involves a CEO who's embarrassed by how the story got told, who says it might be good for the software he left, and who hired people back when the pure-AI version underdelivered. That's not a cautionary tale about being too slow on AI. It's a cautionary tale about believing the hype version of someone else's experiment and applying it to your own books.

The actual lesson for your accounting stack

Here's what I take from it, and it lines up with where the smart money in finance is heading anyway. The Klarna story isn't "AI replaces your GL." It's "the company that controls its own data gets to do interesting things with AI on top of it." Klarna's win wasn't killing vendors - it was getting its data out of a dozen walled gardens and into something it controlled and could build on.

That's the real build-vs-buy reframe for accounting. The question was never "build our own close software vs buy a suite." It's "do I control my financial data and workflows well enough to put AI to work on them?" You can absolutely do that with accounting software you bought - if your ERP, your sub-ledgers, your AP tool are open, composable, and let you reach your own data. You cannot do it with a closed suite that traps your transaction data behind a UI, no matter how many features the demo had. This is the same reason finance teams are drifting toward composable, API-first tools - and there's now even an open standard, Anthropic's Model Context Protocol, built specifically so AI can plug into the systems where your data lives.

And here's the accounting-specific catch most people miss when they fantasize about copying Klarna: your stack probably isn't yours to rip out cleanly. Remember that nearly 60% of accounting roles are tied to a specific system - your team's expertise, your processes, your hiring are all wired into NetSuite or SAP or whatever you run. Klarna consolidated its data with graph databases and internal engineering most finance teams simply don't have. "Rip it all out and build" isn't a plan for a controller with a five-person team and a month-end deadline. "Own your data and choose tools an agent can actually operate" is.

What I'd do

Don't copy Klarna's headline. Copy the part Klarna actually did right: get your financial data out of the walled gardens and into something you control, and choose accounting tools - bought or built - that an agent can operate on top of. When you're evaluating your next ERP, close tool, or AP system, the question to lead with isn't the feature list, it's: can I get my own data out of this programmatically? If the answer is no, that's the red flag, not whether the vendor's marketing says "AI-native." The era of buying accounting software isn't over. The era of buying accounting software you can't get your data out of should be.

This is the exact lens we built our software directory around - so you can evaluate accounting and finance tools on whether they fit a modern, data-portable, automatable stack, instead of drowning in another feature-comparison grid.

Frequently asked questions

Did Klarna replace Salesforce with AI?

Not exactly. Klarna moved off Salesforce's CRM but did not replace it with AI. As CEO Sebastian Siemiatkowski stated, "we did not replace SaaS with an LLM." Klarna consolidated its data from various SaaS tools into its own internal infrastructure (using graph database technology) and ran AI on top of that, while swapping some vendors for other software - Deel replaced Workday for HR.

Did Klarna's CEO say the era of Salesforce is over?

The opposite. Asked whether other companies would leave Salesforce, Siemiatkowski said "I don't think it is the end of Salesforce; might be the opposite," and added "Will all companies do what Klarna does? I doubt it." He has pushed back on the popular reading of Klarna's move.

What does the Klarna story mean for accounting software decisions?

That controlling your own financial data matters more than whether you build or buy. With nearly 60% of accounting roles tied to a specific system (NetSuite, SAP, Oracle, Workday), your stack choice is high-stakes. The lesson is to choose ERP and close tools that are open, composable, and let you access your data programmatically - so AI can work on top of it - rather than a closed suite that traps your transaction data.

Did Klarna rehire humans after going all-in on AI?

Yes. After heavily automating customer support with AI, Klarna began rehiring human agents in 2025, with the CEO acknowledging the all-AI approach had produced lower-quality service in some cases - a caution for finance teams tempted to fully automate judgment-heavy work like the close.

Should accounting teams avoid all-in-one ERP suites?

Not automatically, but scrutinize them. The risk with a closed all-in-one is the walled garden - if you can't reach your financial data programmatically, you lose the ability to put AI to work on it later. A composable set of tools that connect cleanly often ages far better than a closed suite, no matter how complete the feature list looks at demo time.

We organize the accounting and finance software landscape so you can evaluate tools on whether they fit a modern, data-portable, automatable stack - not just feature checklists. Explore it on the Audit Friendly software directory.