The Spreadsheet Problem No One Talks About

There is a number sitting in a spreadsheet somewhere right now. A revenue figure that jumped 10%, or a cost line that crept upward by a fraction, and nobody in the room can say with any certainty why it moved.

This is not a technology failure. It is, arguably, a design flaw that has been baked into financial planning for four decades; one so familiar that most finance professionals have simply stopped noticing it.

Evan Rumpza noticed. And then he decided to do something about it.

Rumpza is the CEO of Galdera Labs, a European startup that has raised 1.5 million euros (about $1.8 million) in seed funding to build what he describes as a “living model.” The tool aims to be a financial planning system that doesn’t just run calculations but holds context, traces decisions and articulates the reasoning behind the numbers in real time. The pitch sounds straightforward. The problem it is solving turns out to be surprisingly deep.

The 90-to-3 Story

The company’s origins trace back to Klarna, the Swedish payments giant that processes transactions for millions of merchants across Europe and beyond. Rumpza and his future co-founders built internal financial systems there that compressed a team of 90 analysts down to three, without sacrificing the quality or accuracy demanded by an organization preparing for an $800 million fundraise and an eventual public offering.

That compression was not achieved by automating the arithmetic. Spreadsheets already handle that. The breakthrough was building systems that could hold the narrative alongside the numbers. Those systems could remember why an assumption changed, who initiated it and how it cascaded through the rest of the model.

“A formula doesn’t tell you why this cell changed,” Rumpza explained during a conversation with PYMNTS CEO Karen Webster. “That context, right now for most organizations, lives in the heads of various stakeholders across the company.”

After watching artificial intelligence (AI) reshape the way Klarna’s engineering teams worked, Rumpza saw the same transformation waiting to happen in finance. The question was how to do it without breaking the one thing finance refuses to compromise on: accuracy.

The Accuracy Problem

Webster put the constraint plainly during their conversation. “You can’t just be 90% right,” she said. “You have to be 100% right because so many decisions hinge on the accuracy of that workflow.”

Rumpza’s answer is not to make AI more accurate at math. It is to stop asking AI to do math at all.

Galdera’s architecture keeps deterministic computation separate from AI reasoning. The core calculations, the numbers themselves, are handled by traditional systems that produce identical outputs from identical inputs, every time. AI is applied to everything else: navigation, context, interpretation and the translation of numerical outputs into something that can explain itself to a CFO.

Connecting the two is a graph-based data structure that maps the relationships between every assumption, change and outcome in the model. When a number moves, the system can trace exactly why, and surface that trace without anyone having to reconstruct it from memory.

“The math stays deterministic,” Rumpza said. “And then the context and the navigation is where the AI really helps.”

Transparency as the Product

In a technology landscape full of companies promising seamless AI behind polished interfaces, Galdera is taking a deliberately different approach. The company’s go-to-market strategy is built around showing CFOs exactly what is happening inside the system, not abstracting it away.

“It all comes back to simplicity,” Rumpza said. “There’s a lot of jargon in this space … but it’s really keeping everything really simple and also exposing a lot of data to the end user.”

The target audience is finance leaders who are drawn to what AI can offer but unwilling to stake their credibility on a black box. For that audience, a system they can interrogate is worth more than a system that simply projects confidence. The transparency is not a feature; it’s the entire basis of trust.

What the Technology Still Cannot Do

Rumpza is candid about the limits. Large language models, he acknowledged, still struggle with the kind of multi-step, sequenced analytical reasoning that defines serious financial planning. The improvement in mathematical reasoning over the past few years has been real, but the challenge of orchestrating complex workflows that combine structured calculation with contextual judgment remains unsolved.

“It’s still this combination of contextual reasoning with also the deterministic structure,” he said, “where there’s a lot of potential and room to grow.”

This acknowledgment shapes the entire product philosophy. Galdera is not claiming to have replaced financial expertise with machine intelligence. It is claiming to have built a framework that routes each task to whatever handles it best, and that does so transparently enough that the CFO sitting at the other end of the system can follow the logic without having to take anything on faith.

“It’s not so much about competing with the foundation models outright,” Rumpza said, “and more about serving the right context to the user in the right moment.”

The Next Interface

The electronic spreadsheet didn’t just change how finance teams worked. It changed what kind of financial thinking was possible. VisiCalc, released for the Apple II in 1979, was one of the first software applications that could justify buying a personal computer for business. Excel, launched in 1985, effectively defined a generation of corporate decision-making.

What Rumpza is proposing is not a replacement for those tools. It is a new layer above them, one that takes the output of deterministic computation and transforms it into something a CFO can actually use to lead.

For organizations managing rapid expansion across multiple currencies, product lines and regulatory environments, the gap between knowing what happened and understanding why it happened is not a minor inconvenience. It is where decisions get delayed, where assumptions go unchallenged, and where the wrong number becomes the basis for a very consequential choice.

Galdera’s bet is that closing that gap is worth $1.8 million in seed capital, and that the CFOs who have spent years explaining their spreadsheets to each other will recognize, almost immediately, what it means to have a model that can finally explain itself.

PYMNTS CEO Karen Webster is one of the world’s leading experts in payments innovation and the digital economy, advising multinational companies and sitting on boards of emerging AI, healthtech and real-time payments firms, including a non-executive director on the Sezzle board, a publicly traded BNPL provider. She founded PYMNTS.com in 2009, a top media platform covering innovation in payments, commerce and the digital economy. Webster is also the author of the NEXT newsletter and a co-founder of Market Platform Dynamics, specializing in driving and monetizing innovation across industries.