When Satya Nadella apparently declared on a podcast in December 2024 that “SaaS is dead,” his comments largely flew under the radar. Fast forward to February of this year and his nuanced response to a complex question has been resurrected as the software sector shed roughly $1 trillion in market capitalisation as investors apparently took the Microsoft CEO at his word.
The headline travelled. The nuance, as usual, did not.
I should be upfront: I run technology at a SaaS platform, which might make me a less-than-neutral voice here. But I've sat with this argument long enough to conclude that the “SaaS is dying” narrative isn't just wrong – it's being used to obscure something far more interesting and, for financial adviser firms thinking seriously about their technology stack, considerably more important.
The numbers rather get in the way of a good funeral
The global SaaS market was valued at over $400 billion in 2025. Forecasts put the 2026 figure at anywhere between $375 billion and $465 billion, with compound annual growth rates of 13% to nearly 19% projected through to the mid-2030s. Gartner expects that by the end of this year, 85% of all enterprise software spending will be SaaS-based. Forrester, meanwhile, projects global SaaS spending growing from around $318 billion in 2025 to over $512 billion by 2028.
Those are not the numbers of an industry heading for the departure lounge.
What has happened – and this is worth being clear-eyed about – is that a specific category of SaaS company is under genuine pressure. Thin-wrapper applications. Horizontal tools that do many things adequately and few things well. Businesses whose primary value proposition is the convenience of a monthly subscription seat rather than any deep product differentiation. Yes, those are vulnerable. But conflating that very real pressure with an existential crisis for the SaaS model as a whole is a bit like watching a cluster of high street chains go under and declaring the death of retail. The thing that's dying isn't SaaS. It's the undifferentiated version of it.
Follow the money
It's worth asking who actually benefits from the “SaaS is dead” narrative, because the answer is instructive. The hyperscalers do – AI workloads justify the extraordinary capital expenditure the five largest US cloud and technology companies have committed to through 2026, reportedly approaching $690 billion. The foundation model labs benefit too, because every dollar redirected from enterprise software budgets toward AI APIs validates valuations that would otherwise be difficult to justify. As one analyst in The Next Web put it recently: “The headline travelled. The correction did not. An industry of analysts, venture capitalists, and foundation model CEOs built a year of marketing on the louder half.”
There's a third party here as well: venture capital, whose portfolio repricing depends on a compelling story about AI-native companies outrunning incumbents. Which they may, in some markets. In others – and financial services sits close to the top of that list – the picture is considerably more complicated.
Evolving isn't the same as dying
To give Nadella his due, his actual remarks were more considered than the coverage suggested. His real point was about AI agents potentially absorbing the workflow logic currently embedded in SaaS applications – not that the platforms themselves would evaporate. Myles Suer, a respected analyst and tech journalist framed this well: SaaS “isn't disappearing – it's becoming the system of record that agentic AI relies on to reason and to act.”
That evolution is real. Pricing models are genuinely shifting – the traditional per-seat, per-month structure is under pressure as AI changes how work gets done, and IDC predicts that by 2028, around 70% of software vendors will have refactored their pricing strategies around new value metrics. But that's a maturation story, not a mortality one.
The more seductive part of the “SaaS is dead” argument is the idea that AI now makes building software so cheap and fast that any sufficiently motivated team can replicate in weeks what took others years to build. In narrow, well-defined contexts with low regulatory complexity and minimal integration requirements, there is some truth to that. The problem is that financial services is almost the exact opposite of that context.
Financial services isn't a good testing ground for the “just build it” thesis
In financial advice, software doesn't just need to work. It needs to operate within a dense, evolving, and largely unforgiving regulatory environment. FCA authorisation requirements. Consumer Duty obligations. HMRC compliance for tax-efficient investment structures. Suitability documentation. Audit trails. Data governance. These are not problems you can prompt your way through, and they are not problems that will tolerate the “ship fast and iterate” ethos that makes AI-accelerated development so appealing elsewhere.
The alternative investment space – EIS, SEIS, VCT and IHT portfolios – is particularly unforgiving in this regard. The regulatory complexity is substantial. The product universe is fragmented across dozens of managers, each with their own structures, timelines and reporting requirements. The interactions between tax reliefs, HMRC advance assurance, investment timing and individual client circumstances are genuinely intricate. Building something that handles this correctly isn't a sprint; it's an extended process of iteration with regulators, product providers, advisers, and ultimately the end investors who have real money and real tax positions at stake.
There's an observation from Benchmark's Sarah Tavel that has been circulating in enterprise tech circles and it has stuck with me. Discussing AI-era enterprise software, she noted that the fastest-growing companies have often kept engineering teams lean while significantly scaling their relationship teams – because enterprise deals “have never been a pure technology problem. They're about trust, customisation, and understanding what keeps a CTO up at night.” That rings entirely true from where I sit.
What actually creates a moat
Imagine an AI-powered startup arriving in the alternative investment platform space tomorrow. With modern tooling, they could probably produce a functional interface in a matter of weeks. What they cannot produce is a decade of established relationships with the investment managers whose products they would need to host. They would have no institutional knowledge of HMRC and FCA compliance edge cases accumulated over years of live operation. No deep understanding of how financial advisers actually use these tools day-to-day – the specific moments in the advisory process where software friction either builds or destroys client confidence. No track record that the regulator recognises.
This matters because the real value of a platform in this space has never been primarily the code. It's the accumulated understanding of a market that the code reflects. Data architecture that maps the genuine complexity of alternative asset portfolios. Workflow logic shaped by years of input from the practitioners who use it. Integrations built through long-standing relationships with providers. A compliance posture developed through sustained, direct engagement with an evolving regulatory framework.
None of that is reproducible from a prompt. It is the product of time, relationships, and the kind of domain knowledge you only develop by being genuinely embedded in a market.
The real question for adviser firms thinking about their technology is not “will SaaS survive AI?” The better question is “which platforms understand our business deeply enough to keep pace with everything coming next?” The answer to that has very little to do with how the underlying software was built. It has everything to do with who built it, how long they have been doing it, and how closely they have been listening to the advisers who depend on it day in, day out.
Software built with AI is still just software. A platform shaped by years of genuine market immersion is something else entirely – and in regulated financial services, that distinction is not a minor one.