Technology strategy
Making sense of the SaaSpocalypse
For more than a decade, the default response to a new business requirement was often to add another application or expand an existing suite. This was fast and convenient, but it came with a trade-off. Instead of the software adapting to how the organization worked, organizations were frequently forced to change their workflows to fit the software. While this approach may work well for a commodity process, it becomes a constraint when the workflow is core to the business – limiting innovation, slowing the pursuit of new opportunities and hindering the shift towards a more data-driven enterprise.
Apart from this, over time enterprises were left with sprawling software estates, overlapping capabilities, expensive customizations and workarounds built around spreadsheets and email , making it near impossible for many organizations to have a clear end-to-end picture of their processes and workflows.
Generative AI and agentic workflows are now opening up a different choice: giving organizations the ability to build software around the way the business needs to work. Across industries, CEOs and technology leaders are beginning to ask a question that few leaders would have seriously considered only a few years ago. Do we still need all the SaaS products we use today (in which for some products businesses were paying for an entire bundle of capabilities which they may not have been using)? Could these capabilities be rebuilt in-house using AI – which would now only allow these systems to be built , operate and evolve at a fraction of the cost , but more importantly help unlock business innovation.
It doesn’t mean SaaS is disappearing. It means SaaS is being unbundled. Enterprises can increasingly rebuild the capabilities that genuinely differentiate their business, while continuing to use SaaS for commodity functions and retaining systems of record where replacement may not add economic or operational value. The question for decision-makers is no longer simply which software to buy. It is which capabilities make sense to buy – and which ones the enterprise should own and control. The advent of generative AI and agentic workflows is making selective SaaS replacement not just a theory, but an economically viable possibility. This is what people increasingly mean when they talk about the “SaaSpocalypse”.
The SaaS world : horizontal vs vertical
To understand where the market is heading, it helps to differentiate between horizontal and vertical SaaS.
Horizontal SaaS includes large, foundational platforms such as Salesforce for customer relationship management or Workday for human resources. These systems are more likely to withstand the SaaSpocalypse. Instead of being replaced, many will serve as the systems of record around which enterprises build new agentic workflows.
Vertical SaaS, on the other hand, consists of more specialized systems tailored to particular industries or business domains. These are likely to come under pressure to be replaced by building them in-house. Organizations already possess much of the domain knowledge on which these products are built.
In the past, packaging that knowledge into software was prohibitively expensive, driving companies to rent off-the-shelf vertical solutions. Today, AI is beginning to make it possible to build these targeted capabilities more rapidly. The cost of building and maintaining custom solutions could increasingly compare favourably with the continuing licensing and customization costs of external products.
The push for workflow autonomy
A major driver of this transition is that businesses are tired of conforming to rigid software. Over the years, organizations have been forced to adapt their internal operations to match the workflows dictated by third-party SaaS products.
In an AI-driven, agentic world, that compromise is becoming less necessary. Companies want the freedom to define their own workflows and build systems that natively support them.
We are already seeing this thesis begin to play out in a number of industries across the world.
A large retailer in Southeast Asia is replacing a proprietary marketplace platform with a more modular solution assembled from open-source commerce technologies. This allows the organization to retain the capabilities it needs while gaining greater control over its workflows and future development.
Similarly, a large financial-services organization is developing a framework to decide which existing systems it should retain, which it should progressively hollow out and which it should replace. The objective is not simply to reduce cost, but to regain control over the workflows that are important to the business.
Why this isn't just another SOA false dawn
Skeptics might compare this trend to the early days of service-oriented architecture (SOA). Back then, the industry promised that companies could easily disassemble giant monoliths and build systems tailored to their needs. In reality, this proved terribly complicated, and organizations spent vast amounts of time and money without achieving their goals.
So, what makes the SaaSpocalypse different?
Historically, one of the biggest challenges enterprises faced was a lack of internal visibility. Critical knowledge about capabilities was locked away in emails, fragmented documents, chat groups and the minds of employees. Today, AI tools can act as a knowledge fabric, extracting and organizing this scattered information. Modern AI-assisted engineering tools can also help decipher legacy systems, making the disassembly process significantly easier than it was during the SOA era.
Furthermore, organizations with a strong digital foundation already have many of the required endpoints available. In an agentic system, the organization can define the goal, the available endpoints and the necessary constraints, allowing AI agents to orchestrate tasks across them. This level of autonomous orchestration was practically impossible in the non-agentic world.
The “hollow out the core” strategy
Enterprises do not need to execute a high-risk, big-bang transformation to escape SaaS lock-in. Instead, they can employ a strategic pattern known as “hollowing out the core.”
This approach involves progressively moving selected capabilities out of major platforms and rebuilding them around the core. Rather than trying to rip out a massive ERP installation overnight, companies can identify the specific capabilities they use, rebuild them outside the platform and gradually reduce their dependence on the incumbent system.
AI-assisted engineering can accelerate this by:
Using curated business and engineering context to design and deliver reusable capabilities more quickly.
Combining internal implementation patterns with a reverse-engineered understanding of existing features and open-source technologies.
Speeding up the reconstruction of workflows, business rules and interfaces.
Consider the marketplace example. A traditional marketplace platform may bundle seller onboarding, catalogue management, moderation and order workflows. An enterprise may not need to recreate the entire product. It can rebuild only the elements it requires, resulting in a leaner and more adaptable alternative. As more capabilities move outside the proprietary platform, the incumbent core becomes thinner, less critical and eventually easier to replace.
Overcoming the agentic trust gap
If AI allows us to build these bespoke systems more easily, why haven’t agentic workflows completely taken over?
The answer lies partly in our tolerance for error. As consumers, we can often live with false positives from AI: a third nostril might make an AI-generated profile picture unusable, but it is more likely to raise a smile than cause serious harm. However, if you direct an autonomous agent to rebalance a financial portfolio, no one inherently trusts it to act without making critical mistakes. Until large language models improve their ability to handle context reliably, enterprises will need strong guardrails and evaluation frameworks. This may involve multiple checks, deterministic controls and human oversight before consequential actions are taken. By building these capabilities into shared engineering platforms, organizations can avoid reinventing the same guardrails for every workflow and domain. Agents and guardrails are fast evolving and the trust is slowly building, enabling the march from simple tasks to complex ones and long running activities.
The path forward
The SaaSpocalypse is not an endpoint. It is a transition away from buying whole platforms towards owning targeted, differentiated capabilities. Organisations which have tied themselves to large monolithic systems now potentially have an easier way our to build composable architectures and take control. Organizations which have tied themselves to a large monolithic system now have an easier way out to build composable architecture to take control of their total platform costs and reduce their dependence on external vendor roadmaps. This approach can reduce technological lock-in and enable faster delivery of business-specific functionality.
AI is making staged replacement economically realistic, giving organizations a safer path to modernization and greater control over how their software – and their business – evolves.
