A few weeks ago, I got to interview Salesforce about its new Headless 360 platform. Without question, this was a significant announcement for the company. But the interview also signaled a bigger shift in the software industry. The answer to one of my questions, in fact, made me quake a little in my boots.
The exchange went like this. I asked John Kucera, senior vice president of product management, for his perspective on Salesforce’s competitive position in the current environment, where agentic AI is actively threatening to kill the SaaS business model.
“We are going to win by continuing to disrupt ourselves, including with our openness strategy,” Kucera said. “To win, we have to innovate faster and make adoption possible for more of our customers.”
It was a great answer. But it was his response to my next question — asking where Salesforce sits three years out — that gave me the jolt. I expected something like, “We’re a venerable franchise.” Instead, I was told: “No one can look that far out.”
Wow. When you hear something like that from a top executive at a leading software company, you realize just how fast the ground is shifting.
This level of uncertainty matters to CIOs because at least one-third of their yearly funding is tied up in software applications, particularly SaaS apps. And they need to have a strategy, or at least a hunch about the road forward.
Anthropic aims to disintermediate SaaS
Without question, Anthropic is aggressively expanding Claude beyond a foundational model into a true enterprise software tool. This has been necessitated by the massive post-money valuations they and OpenAI have obtained from their recent funding rounds — and their aspirations to go public by the fourth quarter of this year. It’s also becoming clear that they must build enterprise applications to deal with the constant commoditization of their foundational models.
To make the move to an enterprise tool, Anthropic’s most recent architecture enables Claude to target the requirements of specific enterprise departments through pre-built workflows and model context protocol (MCP) integration. This is designed — Anthropic has indicated — to displace work currently handled by specialized SaaS applications, particularly broader enterprise capabilities that teams have historically stitched together using low-code tools. Anthropic supports this through Claude Cowork and Claude Code plugins, as well as a cookbook for Claude Managed Agents, enabling Claude to manage complex workflows.
Because of their emerging expanded market charter, many market watchers, including myself, have worried that SaaS is going away or being disintermediated. Adding fuel to the fire, Anthropic CEO Dario Amodei said at a recent company event that individual SaaS companies could “lose market value, go bankrupt, completely go bust.” Amodei argued that SaaS companies can no longer rely on the complexity of their software as a protective moat: “I think if your moat is ‘our software is complex and difficult to write, and we can write it, and others can’t match it,’ I think that’s going away.”
The salvo appears to take aim at legacy vendors like SAP. However, Amodei claimed that the incumbent software companies that pivot to a new kind of moat may even do better than they did before, while those that don’t pay attention “are going to have a really bad time.”
None of this means enterprise applications disappear. But it may fundamentally change where their value — and pricing power — comes from.
Why enterprise applications still matter
To be clear, I don’t believe that large transaction systems are just going to go away. CIOs have spent years of their lives using them and lost too much sleep implementing them. At the same time, CIOs are not interested in managing the development and maintenance of even more software — they would rather make only software that adds a competitive advantage.
Philipp Herzig, SAP CTO, argued, however: “AI creates even more strategic relevance for business applications. As agents take on longer, more autonomous work, the application layer becomes more — not less — critical because it provides the systems of execution, proven workflows and business logic required for reliable execution at scale.”
In other words, the debate is no longer whether enterprise applications matter. The more important question is whether SAP and other SaaS companies — without a radical shift – can continue extracting the same economic rents from what they deliver today.
SaaS companies must become data companies
So how can SaaS companies win, and what should CIOs look for in their roadmaps? I think a path forward for the industry is to make the data coming out of their applications more useful and more accessible. Historically, business intelligence has had to do a significant amount of work to make data ready for use.
In this model, their value shifts away from application code and toward proprietary data, governance and the business logic embedded in enterprise workflows. So, the opportunity for SaaS companies is to become providers of essential curated data sets and specialized knowledge bases made accessible via APIs, virtualized access and MCP services.
In this world, they become the infrastructure that AI agents depend on, regardless of which vendor makes the agent. This is a step beyond Salesforce’s Headless platform, but it seems likely where Salesforce, among others, is going.
Here, companies that survive combine three things with their application chops:
-
Proprietary data that AI agents need to function.
-
Deep domain knowledge encoded in that data.
-
Pricing models tied more closely to outcomes than subscriptions.
As such, data is the foundation of the new moat — but it only becomes a real competitive advantage when it’s unique, refined and operationally essential to the agents doing enterprise work.
Making the same argument, Herzig said , “AI agents require the application layer, including semantically rich data and embedded governance, to operate effectively. Without this context, they produce outputs disconnected from business reality, limiting productivity and revenue growth. Agentic AI and AI-assisted development may change the UI and how software is built, but agents depend on enterprise applications, they are not replacements for them.”
So, the essence of this is that Anthropic is causing SaaS companies to become data companies. By doing so, this will effectively increase competitive pressure on pure-play data companies. And by doing so, it will accelerate industry consolidation. Given this, here are some relevant examples of giant software incumbents acquiring specialized data players to survive the AI shift:
What is less clear is what this will do to data platform companies and integration companies. But as Boomi CTO Matt McLarty pointed out , an AI agent is entirely limited by the data it can access, which turns integration into a core “competitive differentiator.”
What CIOs should do now
It is hard to judge exactly how competitive dynamics will change or precisely what actions CIOs should take at this stage. A few principles stand out.
Pursuing shorter-term vendor deals makes sense while your data and AI strategy is still taking shape. Longer-term commitments are only justified when a vendor’s economic model aligns with your firm’s needs and the direction of your AI strategy.
Finally, if your organization is behind on delivering AI capabilities, you should aggressively pursue vendors that can accelerate your AI applications and technology roadmap.
