I’d describe myself as tech-savvy, even though I have next to no clue about actual coding. I am a classic SaaS heavy user in my private life, but even more so in my professional environment.
After several years in the world of Big Law firms with not much exposure to SaaS services (beyond Microsoft’s Office suite), I’ve spent nearly three years working with startups and scale-ups, which, at their core, are almost always B2B SaaS. And during this time, I reviewed countless SaaS providers (from a legal perspective, either clients wanted to procure them or comply with applicable law; in either case, I had to understand what those providers do and why clients need them).
If you follow the current market narrative of a dying SaaS industry, you might think the offices of these companies are filled with a sense of impending doom. But the opposite is true.
None of the founders, engineers, or product leads come to work with slumped shoulders because they fear AI will take their jobs or their entire company tomorrow.
Instead, they see AI as the greatest opportunity since the invention of the cloud. They are trimming internal processes for efficiency and working feverishly to integrate AI features into their own products.
Now, one could argue that these people are disillusioned, like the frog in the beaker, enjoying the warmth while the water slowly reaches a boil. It’s a fair point, and certainly worth discussing.
But let’s switch perspectives and look at how these companies actually operate. They all use a complex stack of B2B solutions layered on top of classic infrastructure such as AWS, Azure, or Google Cloud.
Every sales-driven company needs a CRM. Whether it’s Salesforce, HubSpot, or Microsoft’s solution hardly matters. Without these platforms, a modern sales organization simply won’t function.
Sometimes I honestly wonder what investors, who have never configured or operated these products themselves, think these tools actually do.
Salesforce or HubSpot aren’t just digital phone books with a fancy email interface. They map the entire sales cycle: from initial lead generation to demoing and complex contract negotiations, all the way to creating quotes and securing archiving.
Sometimes one feature is solved more elegantly by the competition, sometimes another provider has the better dashboards. Yet 99.9% of all companies that need such solutions will never even consider developing them in-house.
Regardless of the vibe coding hype, it is simply far too complex and, from a business perspective, pure madness. In my daily life, I see absolutely no tendency toward building it yourself.
Quite the opposite: the dependency on major players is growing. Salesforce and HubSpot are integrating AI features so deeply into existing workflows that switching or building an alternative becomes even more absurd than it already was.
Of course, the hype is real, and rightly so. The development of LLMs is breathtaking, and what clever colleagues are now solving via agentic structures for internal workflows is impressive.
A simple example: product updates are gathered via a collaboration tool and a bot shares them in a reader-friendly format in the relevant Slack channel. It saves time and nerves.
But what do my colleagues use for this automation? They use Zapier, they use Notion, they use Slack. All SaaS. Slack, by the way, is owned by Salesforce.
Our marketing pitch deck looks incredible today thanks to AI support, but is our team using an in-house-coded tool? No, we use AI-enabled SaaS.
This pattern repeats across every department, including HR. I don’t see anyone in the industry suddenly building their own HRIS or handing over high-stakes services outsourced to giants like ADP or Sequoia to an experimental agent.
Everyone talks about how great automation is, but if you ask an experienced HR manager whether they want to let an autonomous bot handle the payroll or the critical API between their HRIS and payroll provider, you’ll get a tired smile and a very clear “Nope.”
The trust in the reliability and legal security (not to mention liability for data breaches) offered by an established SaaS provider carries far more weight than the promise of cost savings through home-grown AI.
I won’t even start on the massive gap between vibe coding a product copy and actually maintaining, monitoring, and updating that infrastructure 24/7.
What we are witnessing right now is, in my opinion, a violent market correction of the kind that happens every few years.
Companies built on shallow foundations are being washed away. The prime examples are the countless wrappers that do nothing more than take an existing LLM, add a thin prompt layer and a halfway decent UI, and sell it as the next AI revolution.
These firms have no moat. When the underlying models integrate the feature natively, these wrappers vanish overnight.
At the same time, I think it’s highly likely we’ll see a massive shift in pricing. The classic pay-per-seat model might increasingly be replaced by pay-for-results or pay-for-compute.
This reminds me of Microsoft’s transformation in the early 2010s. Back then, the giant successfully pivoted from one-time license sales to the cloud subscription model.
It was painful, but it was the only way to survive the new reality. SaaS companies must now prove that their value lies not in the number of user accounts, but in the quality of the results their AI-powered workflows deliver.
Those who manage this shift will not be replaced by AI. They will become indispensable because of it.
I could be wrong, of course.
One might say we are only at the very beginning.
What was sloppy text generation yesterday and agentic workforce elements today could be self-learning, self-improving systems tomorrow.
Fair enough, I don’t have a crystal ball, and I am humble enough to operate with caution. But I want to point to a few general heuristics.
The burden of proof lies clearly with the new development. I don’t have to prove why an AI feature for redlining in a contract management tool doesn’t work yet.
The feature has to prove to me that it actually takes work off my plate rather than just creating an additional layer of oversight.
If the AI diligently recognizes redlines but ignores a crucial question added as a comment by the opposing counsel, then the tool is useless in a professional context. It’s a neat toy, but not a tool.
It might be fixed eventually, but it illustrates my distrust of the current market when a tech CEO approaches me and says AI can now do contract redlining.
I have yet to see a single use case that provides serious evidence that platforms like Salesforce, ServiceNow, or Intuit are doomed.
Anyone who thinks these giants are currently imploding should look at the financial reports. The hard numbers tell a very different story.
Looking at the most recent earnings reports, these companies are not just surviving. They are often accelerating.
Salesforce reported revenue of $10.3 billion for its Q3 2026, representing 9% year-over-year growth. That is stable and shows the platform’s power remains unbroken.
ServiceNow absolutely crushed expectations in Q4 2025, with revenue growing 20.5% to $3.57 billion, proving their AI Control Tower approach is winning in the enterprise segment.
Intuit delivered an impressive 18% revenue increase to $3.9 billion in Q1 2026, driven by 21% growth in its Online Ecosystem (next earnings are scheduled for February 26).
And for those who think infrastructure-adjacent SaaS is suffering, look at Datadog (see their investor day presentation here). In Q4 2025, their revenue surged 29% to $953 million. That is the strongest growth since early 2024.
These numbers aren’t signs of a structural collapse. On the contrary, they show that legacy SaaS players are already riding the AI wave, while many pure AI startups are still struggling to find a business model beyond burning GPU hours.
Take a quick look at Booking Holdings (rather B2C). Anyone who believes the travel industry, as a classic intermediary, will be immediately disrupted by AI agents booking vacations autonomously should look at the numbers from February 18, 2026.
In Q4 2025, Booking increased revenue by 16.1% to $6.35 billion compared to the previous year. The company is actually accelerating. Total revenue growth for 2025 was 13%.
Booking Holdings already uses generative AI extensively in customer service, reducing costs per booking by about 10%. Meanwhile, management is investing an additional $700 million in 2026 into GenAI
For Q1 2026, the company expects a further 14% to 16% revenue jump.
My second heuristic is that, ultimately, all (!) AI features run on models from maybe three or four companies that are becoming increasingly similar. We are moving toward a commodity structure.
All these models are LLMs, which have inherent limitations due to their probabilistic nature. They predict the next word based on probabilities.
This statistical nature largely excludes them from high-risk workflows. If I’m working on a nine-figure M&A deal in a Big Law firm, I can’t use a solution that hallucinates correctly 95% of the time.
In fields like compliance or payroll, almost always right is factually wrong. If a broken clock is right twice a day by chance, that doesn’t make it a reliable timekeeper.
No employer can afford an agent getting creative with social security contributions just because a specific case was slightly ambiguous. Think of the legal implications of individual cases like parental leave, state subsidies, or individual pension funds across different jurisdictions.
This glass ceiling of reliability is why we aren’t seeing a mass slaughter in the B2B sector. As long as models are not deterministic, meaning they don’t guarantee the exact same, rule-based result for the same input, they remain what they are today: highly capable assistants, not autonomous decision-makers.
Lawyers, by the way, understand this, with fees rising more than ever.
The established SaaS giants understand this quite as well. They use LLMs for interaction and vibe, but the critical business logic, the databases, and the legally compliant workflows remain within their proprietary, deterministic code.
This is also the answer to the Booking Holdings numbers. An AI bot can give you wonderful travel inspiration, but when it comes to the final transaction and liability, we trust the platform that has held the infrastructure for years and decades.
The LLM providers supply the engine, but the SaaS companies own the chassis, the brakes, and the insurance policy. Betting that an LLM wrapper will replace Salesforce or Booking underestimates the weight of liability in the global economy.
I draw a very clear personal conclusion from this analysis. My portfolio has grown by about 11% p.a. since 2017, and I have documented every step on my blog since 2020. I’m a rather conservative investor with a fetish for dividends.
But looking at this mix of absurd doomsday scenarios for cash-flow-heavy players and their actual fundamental strength, I see a historic opportunity.
I am actively using this phase to invest heavily, in multiple tranches, in this exact B2B SaaS sector. Especially with Salesforce, ADP, and ServiceNow (and slightly behind them, Intuit), I see massive potential.
As always, I will present the details of these investments in my upcoming monthly update.
You can support me by subscribing to my Substack.
