The Death of FinTech SaaS: Hype, Reality, and What Investors Should Watch

"AI will kill SaaS" has become one of the most persistent narratives in financial technology. Agents replace software interfaces. Entire product categories get wiped out. We have seen similar claims before. Robo advisors were meant to replace wealth managers. They did not. So the real question is whether this time is different, or whether this is another hype cycle with a catchy headline.

This article is based on a recent LinkedIn Live hosted by Marvin Labs with Alex Hoffmann (Co-Founder & CEO, Marvin Labs), Steven Carroll (Founder, Carroll Consulting & Advisory Services, formerly London Stock Exchange Group), and James Yerkess (Former Global Head of Transaction Banking & FX, HSBC Wealth Management).

Real in places, overstated everywhere else

The "death of SaaS" claim is rooted in something real. AI can already augment and, in some cases, automate process-driven functions. Middle and back office operations, trading settlement, CRM management. These are areas where it is making a measurable difference. But there is an important distinction between process automation and decision automation.

Like any hype cycle, it's based on kernels of truth. There are significant areas where AI can augment human capabilities or automate process-driven functions. But the distinction between process automation and decision automation is really important.

For equities content, where most information is publicly available and reasonably standardized, the case for disruption is strongest. Move into other asset classes with different structures, heavier regulation, or multiplayer coordination requirements, and the narrative weakens fast.

Why financial services moves slower than the narrative

Most financial services does not adopt technology the way consumer software does.

Banks transmit credit into the real economy. Regulators do not want that process to seize up. Corporate executives will not take risks that produce negative headlines. For insurance, lending, and credit decisions, firms need to explain how a decision was reached, prove there was no bias, and show that the AI was blind to factors it should not consider. Neural networks make that kind of reverse engineering difficult, which alone explains much of the sluggishness in insurance adoption.

Then there is coordination. Settlement systems involve hundreds of counterparties. AI might improve the user interface on one side, but it does not resolve the coordination needed between hundreds of banks in a shared settlement system. AI solves a little of the software engineering problem and a little of the user interface problem. It does not solve the coordination problem.

And the operational risk appetite of large banks is genuinely conservative.

If I take one major global bank, it's still using core banking systems that have the old green crosshairs moving around the screen because it was coded so long ago. I don't see any movement of AI taking out that type of service, because the risk of change is too great.

The social media gap

A large portion of the "SaaS is dead" cycle is driven by platform incentives, not operational evidence. LinkedIn and Substack reward extreme positions. A post declaring "SaaS is dead!" with 500 bullet points gets more circulation than someone saying SaaS is evolving but structurally important. Steven Carroll has written about this polarization in detail, dismantling the pantheon of AI myths that range from imminent catastrophe to claims of total invulnerability.

There's a big gap between what is being said publicly on social media and what is actually happening operationally on the ground. That gap is where this debate about SaaS being dead is filled as an easy route.

For investors, this matters. The consensus view on AI disruption in FinTech SaaS is partly a social media artifact. Firms are experimenting. They are running pilots and evaluating use cases. But the notion that SaaS categories have already been eliminated is ahead of reality on the ground.

Where AI is making a real difference

The clearest wins are in unstructured data processing. Earnings calls, press releases, regulatory filings: AI can consume these at volume and present material information to decision-makers in minutes rather than hours. It is easy to underestimate how much of the work in financial services is just reading.

Investment research is seeing a structural shift in how qualitative work gets done. The quantitative side has been deeply developed for decades: risk models, Excel, financial modeling. The qualitative side, the business analysis, the "what does the company actually do" work, was always left behind because it required reading and manual effort. AI is bringing these two sides closer together, enabling risk management on qualitative data and repeatable analysis across larger coverage sets.

You can do risk management now on qualitative data. You can run a way to track if your investment thesis is still up to date, still working. You can repeat the same analysis that you've done in one company on 100 companies.

Firms are also using AI internally to map their own processes, find duplication, and identify where new technology could be introduced. Less visible externally, but producing real savings.

Investment banking and equity research are seeing heavy tool adoption, whether for IPO processes or coverage reports. The underlying content set is similar across these workflows, and tools that synthesize information from primary financial content are gaining traction. The disproportionate number of AI tools targeting investment banking is not a coincidence.

Data vendors face hard questions

Financial data vendors are at an inflection point, and investors evaluating these businesses should be asking harder questions about defensibility.

If a vendor's primary offering is end-of-day pricing or company filings from public sources, that moat has been breached. Two years ago these businesses had a modest competitive advantage. Today, AI can access and process much of this content directly. If you are trying to market a financial data set with no proprietary content, you have zero moat right now.

Proprietary data is different. Complex surveys, industry-specific datasets, content behind licensing or IP protection: these retain defensibility. The question for each vendor is where the revenue mix falls between commoditized and proprietary.

Commercial models matter as much as the content itself. Per-seat pricing looks increasingly exposed as AI reduces the number of humans touching the software. S&P Global and LSEG have been more open to enterprise-type deals, while other parts of the industry remain attached to per-user pricing. Per-seat models look far more vulnerable than pricing that is agnostic to the delivery channel.

The other dimension is AI strategy coherence. Vendors that are open to letting customers access data through AI-native tools (MCPs, CLIs, agent frameworks) are in a stronger position than those restricting access to protect legacy distribution. Locking data behind proprietary interfaces is no longer a viable AI strategy.

I think everybody needs to figure out: do I want to push my own agent, my own system? Do I want to cooperate with people? Or do I just have a pricing model that is attractive, where I'm licensing with my customer and they can access the data in the AI system of their choice?

Startups and incumbents operate in an ecosystem

Does AI favor startups or incumbents? The grass is always greener.

Incumbents have distribution, customer relationships, data assets, regulatory familiarity, and procurement clearances that take years to build. Startups have speed, AI-native architecture, and the ability to focus the conversation entirely on product.

Both sides have an advantage. If you're a startup, you love to have the discussion just on the product. But the big innovations are not coming out of incumbents. I'd love to have the sales teams, the distribution, the data sets as first-class access, which is the advantage they have.

Financial services tends to operate as an ecosystem rather than a winner-take-all market. Startups take higher risks, fail-test approaches, and occasionally get absorbed by incumbents who scale what works. Revolut disrupted foreign exchange by offering transparent pricing. Over a decade, large banks absorbed many of the innovations Revolut pioneered, from progressive onboarding to faster cross-border UX. Revolut itself took ten years to get through the regulator to be classified as a bank.

For investors, the question is not which category wins but whether a specific firm's competitive position is improving or deteriorating. Strategic coherence in AI strategy matters more than whether the company was founded in 1990 or 2020.

How to evaluate FinTech SaaS in this environment

Start with the IP. If a vendor collects exchange-traded information or publicly available filings, that content has limited defensibility. Proprietary content with licensing protection is a different story. This is the single biggest driver of long-term pricing power.

Look at the commercial model. Per-seat faces structural headwinds. Consumption-based or enterprise models that work regardless of whether the end user is a human or an AI agent are more durable.

Evaluate M&A history for signals of strategic foresight. As Steven noted, S&P acquired Kensho in 2018 when AI was barely on the radar. Morningstar acquired PitchBook. MSCI positioned itself in private markets indices. These moves reflected management teams picking up assets before they became expensive, and that kind of foresight is a strong signal when evaluating information services firms.

Test whether management can articulate an AI strategy that goes beyond cost cutting. Do they see AI as a way to improve their product, expand content, and serve customers better? Or do they treat it as a threat? Firms that see AI only as a cost lever are more exposed than those building with it.

And do not forget the basics. Revenue quality, retention, unit economics. AI hype does not change the fact that these metrics still separate good businesses from bad ones.

Where this leaves investors

The "death of FinTech SaaS" narrative is a useful provocation but a poor investment thesis in its bluntest form. Process automation is real. Friction is being removed from workflows. But the regulated, multiplayer nature of financial services means wholesale replacement happens slowly.

The productive question for investors is which specific businesses are gaining or losing position as AI reshapes how financial content is created, distributed, and consumed. The firms worth owning have proprietary content, flexible commercial models, and management teams that see AI as something to build with. The firms most at risk sit on commoditized data and are clinging to pricing models built for a world with more humans in the loop.