Why Deep Specialism Beats Generalist AI in Customer Experience

Have you noticed that some of the leading and publicly listed customer technology vendors have been having a bit of a torrid time on the stock market lately? Indeed, many of them have been losing up to, and sometimes more than, 50% of their share price since the turn of the year.

What has happened (and is still happening) is that market investors, with the emergence of agentic workflow capabilities such as Anthropic’s Claude Cowork, are losing confidence in the traditional Software-as-a-Service (SaaS) business models that technology vendors have relied on for the past two decades.

In the market’s mind, selling software licenses based on the number of employees using the app is less attractive than it once was, especially when you consider that if 10 human operators using AI agents can manage the workload previously handled by 100 people, say, then the enterprise customer suddenly doesn’t need another 90 software licences.

A Real SaaSpocalypse?

This whole phenomenon has been described as a “SaaSpocalypse”, a term coined by Jeffrey Favuzza, an Equity Trader at Jefferies, in a Bloomberg report, in early February 2026.

However, while the term emerged from financial markets, over recent months it has quickly become the shorthand for a structural crisis of confidence in the traditional SaaS business model across the broader business community.

Does that mean that these tools, and particularly those in the customer experience domain, are in terminal decline?

Not necessarily.

The problem is that we have seen these types of things before in recent years, and tech history is full of ‘extinction narratives’ where a new paradigm emerges, markets react, and analysts predict the complete collapse of established business models.

The reality is that, in almost every case, while the initial threat was real, the market response was often exaggerated because it overlooked how incumbents adapt and integrate new technology.

For instance, here are three prominent examples from the technology industry that share the same traits as the “SaaSpocalypse” thesis:

1. ‘Open-Source Software Will Replace Paid Software’ (Late 1990s–2000s)

Remember when, in the late 1990s and early 2000s, as Linux, Apache, and MySQL gained momentum, market observers argued that ‘free’ open-source software would destroy the business models of commercial software vendors? At the time, Steve Ballmer, then CEO of Microsoft, famously called Linux a “cancer” on its business model, fearing that nobody would pay for operating systems or enterprise middleware if open-source alternatives were available for free.

2. ‘The Web is Dead / Apps Will Replace the Web’ (2010)

Remember when, in 2010, after the launch of the iPhone and iPad, Wired published an article claiming that “The Web is Dead” and predicted that closed, native mobile apps would render web browsers, URLs, and traditional websites obsolete, threatening companies built around the open web, such as Google.

3. ‘The Cloud Will Destroy On-Premises Enterprise Giants’ (Early 2010s)

Remember when, in the early 2010s, the emergence of Amazon Web Services (AWS) and pure-play cloud software took off, many on Wall Street declared legacy enterprise giants such as Microsoft, Adobe, and Oracle obsolete. The narrative, at the time, was that no company would ever buy traditional software licenses or maintain on-premises databases again, and that agile cloud startups would wipe out the incumbents.

Despite all the hyperbole at the time, nothing happened as predicted. Open source didn’t replace paid software. Instead, it became the foundation of modern enterprise IT infrastructure when enterprise clients realized they were not just paying for the code itself, but for security, compliance, SLA guarantees, data governance, and support.

Native apps did explode in number, but the open web adapted and developed modern standards, such as HTML5, which enabled developers to build faster, more accessible, and more interactive web experiences as well as browsers powerful enough to run desktop-grade applications.

Cloud didn’t destroy on-premise enterprise giants. Instead of dying, many of the incumbents executed massive pivots. For example, Adobe shifted Photoshop and the rest of its portfolio from boxed products and perpetual software licenses to the Creative Cloud subscription model and expanded its portfolio into digital marketing management. Meanwhile, Microsoft pivoted radically under Satya Nadella to build Azure, it’s cloud computing platform, and Office 365 (now known as Microsoft 365), it’s cloud-based productivity, communication and collaboration suite.

Generalism v Specialism

So, what’s likely to happen with the SaaSpocalypse, particularly when it comes to the customer engagement, experience and support space?

Balaji Balasubramanian, President and CPO of SAP CX, has been in the tech industry for multiple decades and believes that we are witnessing “a step change that we have never seen before” and that “the hockey stick growth that we are seeing in technology capabilities is breathtaking.”

However, while he believes that AI agents and agentic capabilities will enable us to do phenomenal things, he cautions that they need to be grounded in context and enterprise-wide data. That, in SAP’s and Balasubramanian’s mind, means that “a great customer experience can only happen when you marry the front office domains, whether it’s selling or marketing or service, to your operational back end.”

Meanwhile, Dan O’Connell, CEO of Front, says that while he thinks that the “SaaSpocalypse is overblown”, he also believes that “AI agents will become more capable and handle more complex tasks over time”, but also that brands will continue to find “value in platforms and predictability of pricing.”

Balasubramanian agrees and adds that combining AI agents with application context and enterprise-wide data will allow them to “solve much more phenomenal problems for our customers,” but adds that anyone who thinks they can build an app to replace their SaaS providers ignores the “foundational reality.”

O’Connell is not so sure about the data aspect when he says that while he thinks “there are some data moats that exist”, he questions how much protection they offer given how fast the frontier models are advancing.

Instead, he believes one way to think about whether your existing application or provider will get eaten up by AI is to do with “the denseness of workflows because getting people to change their behaviors and the workflows within systems becomes pretty entrenched and the hard thing to do is to go and rip out and change the behaviors.”

For me, what seems to be emerging is a battle between generalism and specialism. For example, you wouldn’t ask the same person to design and install a new kitchen for you, and then ask the same person to fix your car. Fundamentally, that’s not going to happen because of all of the domain-specific skills and expertise and experience that are required for both jobs.

General-purpose AI agents are effective at synthesis and broad cognitive tasks, but they often lack the ‘procedural gravity’ needed for specialized enterprise operations. While a generalist model can easily produce a summary or answer straightforward questions, it struggles with complex, multi-step workflows that require industry-specific compliance, legacy data validation, and organizational rules integration. These ‘dense’ workflows serve as significant barriers. They are not just user interfaces, but operational frameworks that are costly and time-consuming to replicate. As a result, a specialized SaaS platform’s value extends beyond being a simple data repository to becoming a ‘system’ that enforces intricate, often messy operational logic that a generalist AI cannot understand without extensive, proprietary context.

As a result, domain expertise, proprietary data, unique context, and the density of workflows that applications address are likely to be among the characteristics of the SaaS providers that survive the “Saaspocalypse.” However, as we have seen with other tech disruptions, markets tend to price in technological disruption instantly but forget that enterprise adoption moves slowly and that incumbents adapt by integrating the disruptive technology into their existing platforms and offerings.

Plus ça change, plus c’est la même chose…..maybe.