The SaaS Grim Reaper: 3.4% OPEX reduction in Klarna.

Between 2%-5% of OPEX used to go to SaaS at this size of the company on average:)

Death is coming for your enterprise software stack.

And unlike that horror movie cliché where the oblivious teenager insists on investigating the strange noise in the basement, this is one murder scene you'll want front-row tickets to.

The Great SaaS Reckoning: Your CFO's Secret Fantasy

Let's be honest—your company's software ecosystem resembles a digital hoarder's paradise. That "essential" project management tool you adopted in 2019? Nobody's logged in since last August. The CRM your sales team swore would transform revenue? They're still keeping the "real" data in spreadsheets.

And don't get me started on those 17 different communication platforms where important messages go to die.

But wait, you protest, SaaS is the backbone of modern business!

Without our precious subscription software, we'd be back to the stone age of on-premise solutions and—gasp—actual human coordination!

Here's the contrarian bomb I'm dropping:

What if your entire SaaS stack is just an overpriced, overcomplicated workaround for problems that AI could solve more elegantly?

Ridiculous?

Tell that to Klarna, the Swedish fintech giant that just ceremoniously dumped Salesforce and approximately 1,200 other SaaS solutions from their tech stack.

Yes, you read that correctly—one thousand two hundred separate software subscriptions, consigned to the digital dustbin.

Their CFO probably needed medical attention when they realized how much they'd been spending on redundant software. But here's where it gets interesting: cost-cutting wasn't even their primary motivation.

Strategy Shift: From Digital Hoarding to AI Prosperity

What Klarna's leadership team realized—and what might send shivers down the spine of every SaaS CEO—is that the fundamental promise of enterprise software has gone unfulfilled. Instead of creating order from chaos, SaaS has birthed new forms of chaos: data silos, disjointed workflows, and an archipelago of disconnected information islands.

Sebastian Siemiatkowski, Klarna's CEO, put it rather bluntly: "Feeding an LLM the fractioned, fragmented, and dispersed world of corporate data will result in a very confused LLM."

Translation for the AI-uninitiated: garbage in, garbage out.

Even the most sophisticated artificial intelligence can't make sense of your company's knowledge when it's scattered across hundreds of different platforms, each with their own proprietary data structures, access controls, and incomplete information.

The strategic insight here isn't just about cost-saving—it's about recognizing that the SaaS model itself has become the obstacle to the next evolution of enterprise productivity: truly intelligent, AI-powered operations.

Case Studies: The Early SaaS Exodus

Klarna's Knowledge Graph Revolution

When Klarna embarked on their journey away from SaaS dependency, they didn't just rip out Salesforce and call it a day. They fundamentally rethought how corporate knowledge should be organized.

As Siemiatkowski explained:

"We started exploring a few key concepts: What of our data was actually valuable? What data was duplicative, incorrect, or contradicting? Why was it like that?"

Their exploration led them to Neo4j and the world of graph databases—technology that organizes information not as isolated records in tables, but as interconnected nodes in a knowledge network. This became the foundation for a unified corporate intelligence layer that could actually feed meaningful, contextual information to their AI systems.

The result? A consolidated knowledge graph where their AI tools could leverage the entire corpus of corporate information, rather than the fragmented pieces visible through individual SaaS applications.

The Enterprise Followers

Klarna isn't alone in this transformation.

As one commenter on Siemiatkowski's LinkedIn post noted:

"We went through a similar exercise at one of the largest companies in the MENA region. At first, the idea of shutting down major SaaS providers seemed daunting, but the autonomy it gave us over our operations was transformative."

Another company leader shared:

"We are in the same journey to remove 3rd party SaaS/Vendor solutions to go in-house in a mix of open source, AI, and in-house developed platforms.

Reinvented our platforms."

The pattern emerging here isn't just cost-cutting—it's a fundamental rethinking of how enterprise software should function in an AI-first world.

Executing Your Own SaaS Jailbreak

Ready to stage your own software revolution?

Here's how to methodically dismantle your SaaS dependency while building something more valuable in its place:

Step 1: The Great Data Audit

Before you can meaningfully replace SaaS with AI, you need to understand what data actually matters to your business. Start by:

• Creating a comprehensive inventory of every SaaS solution in your organization

• Mapping data flows between systems to identify redundancies and dependencies

• Prioritizing datasets based on business value and uniqueness

• Identifying the "systems of record" for critical data categories

Remember, this isn't just an IT exercise—it's about understanding what information drives your business outcomes.

Step 2: Knowledge Graph Implementation

The foundation of your AI-first architecture will be a unified knowledge representation layer. This typically involves:

• Selecting appropriate graph database technology (Neo4j, like Klarna, or alternatives like Amazon Neptune or TigerGraph)

• Defining your ontology — the conceptual structure that represents your business entities and their relationships

• Developing data pipelines to extract, transform, and load information from legacy systems

• Establishing governance processes to maintain data quality and relationships

This stage requires both technical expertise and deep business domain knowledge to ensure you're modeling information in ways that reflect how your organization actually operates.

Step 3: AI Integration & Interface Development

With your knowledge foundation in place, now comes the exciting part:

• Selecting and integrating appropriate LLM technology — either via commercial APIs like OpenAI's or implementing open-source models

• Building custom interfaces using tools like Cursor (Klarna's choice) or similar development environments

• Training AI systems on your now-consolidated, high-quality data

• Creating domain-specific prompt libraries that encode business rules and common workflows

Step 4: Progressive SaaS Elimination

Now comes the satisfying part—methodically eliminating SaaS subscriptions:

• Prioritize replacements based on cost, user complaints, and data centrality

• Implement replacement functionality in phases, ensuring no critical capabilities are lost

• Maintain dual systems temporarily during transition periods

• Collect and analyze feedback to continuously improve AI-powered replacements

Results You Can Expect

Organizations that have embarked on this journey report several common outcomes:

1. Significant cost reduction — Often 30-50% of total software spend

2. Dramatically improved data quality through consolidation and standardization

3. Enhanced employee productivity from reduced context-switching and unified interfaces

4. More agile business processes that can evolve without vendor dependencies

5. Competitive advantage through unique, AI-powered workflows tailored to your specific business needs

Common Pitfalls to Avoid

This transformation isn't without challenges.

Watch out for these common obstacles:

1. Underestimating data migration complexity — Legacy SaaS often contains critical business data in proprietary formats

2. Resistance from power users who have built their workflows around specific tools

3. Integration gaps where your new system lacks specific functionality needed by certain departments

4. AI hallucination issues if your knowledge graph is incomplete or contains contradictory information

5. Implementation exhaustion when the project scope expands beyond initial estimates

The key to avoiding these pitfalls is a phased approach focused on quick wins and continuous improvement rather than a "big bang" replacement.

The Future Enterprise: AI-Native, Not SaaS-Dependent

What emerges at the end of this transformation isn't just a cost-optimized version of your old business—it's fundamentally a different kind of organization. One where:

• Knowledge flows freely across departmental boundaries

• AI augments human decision-making at every level

• Business processes evolve organically based on actual usage patterns

• Innovation happens continuously rather than waiting for vendor release cycles

As Siemiatkowski noted:

"The same thing as is always true, used to be mobile first, now it is AI first."

Just as the mobile revolution forced companies to rethink customer experiences, the AI revolution is forcing us to rethink our operational foundations.

The SaaS giants won't go quietly into this good night, of course.

They'll fight for relevance by bolting AI features onto their existing products. Some may even successfully transform into AI-first platforms.

But many won't make the transition, becoming the Blockbusters and Kodaks of the enterprise software world.

The question isn't whether this transformation will happen—it's whether your organization will lead it or be dragged along reluctantly as competitors gain advantage.

The SaaS emperor has no clothes.

And AI just handed out telescopes to the entire kingdom.