In early 2026, the stock valuations of leading SaaS companies declined significantly after capital markets absorbed the increasing capabilities of AI coding software, such as Claude Code and OpenAI Codex. Commentators labeled this downturn in valuations the “SaaSpocalypse,” suggesting that enterprises may no longer need to buy software if they can build it themselves.

We believe this label overstates the implications.

Why did the SaaSpocalypse happen?

The central argument of SaaSpocalypse is this: If enterprise IT can build any software using AI, why buy it from a vendor?

That view depends on a few key assumptions:

• The cost of building and maintaining software is less than, or at least comparable to, the cost of licensing and operating SaaS solutions.
• Enterprise IT teams can replicate commercially available software and continuously enhance its value.
• New entrants can disrupt established SaaS providers and thus increase competition—either taking away market share from incumbents or compressing pricing.

Why we believe the market overreacted

These assumptions overlook a critical point: enterprise software is not just code. AI is making it easier to write code. It is not eliminating the complexity of running enterprise software. Expertise and deep domain knowledge play an important role in making effective enterprise software.

Coding is the easy part. Institutional knowledge is much more difficult to replicate.

Building software is not just coding

Beyond coding, building SaaS applications requires understanding domain rules, regulatory fluency, embedded understanding of business processes with their nuances across regions, products, brands and organizations—and then building architecturally robust, scalable, performant, secure and production grade reliable software.

Each technical area requires specialized systems expertise. For example, reliability demands disaster recovery, failover, distributed systems design that handle workload elasticity and compliance.

Enterprise software must ensure access control, auditability, traceability, rollback mechanisms and data governance.

Institutional knowledge’s importance for enterprise applications

Enterprise applications also encode institutional knowledge. They represent decades of operational decisions, calculation logic, compliance interpretation and edge-case handling across many deployments.

This knowledge is difficult to replicate because it is not fully captured in code. It emerges over time through real-world usage, exceptions and regulatory constraints, and depends on core disciplines required to build and run enterprise software:

• Domain modeling
• Architectural integrity
• Production operations
• Governance
• Security and compliance liability

AI expands what IT teams can prototype, but it does not eliminate these requirements. AI accelerates software development for SaaS companies as well. The productivity delta does not accrue exclusively to enterprise IT.

2 historical analogies: Open source and industrial automation

Open source didn’t end software companies. Instead, it expanded the surface area of what application and other infrastructure companies could build and how they built it. It commoditized certain layers and forced business model evolution in others. We expect AI coding to do the same, to take away value from undifferentiated software and increase the value of the rest.

Likewise, robotic manufacturing and assembly lines reduced the price and increased the throughput and quality of building everything from cars to chairs. It enabled car companies to design better cars that lasted longer, gave better fuel efficiency, reliability and comfort. A vehicle is not just assembled parts; it is architecture, the design, supply chain coordination, safety engineering, regulatory compliance, the ecosystem that services its life cycle and brand trust.

Complex software application is similar. AI will make building more sophisticated software easier for companies whose primary pursuit is to build sophisticated, valuable systems. Customers will benefit from rapid innovations, better user experiences and increased reliability.

Our predictions for the future of SaaS software

Below are three predictions on where pricing pressure, competition and new agentic categories will emerge.

1. The commoditization of low-complexity software

Low-complexity, undifferentiated tools will experience pricing pressure or internal replacement. This aligns with historical compression patterns.Multiple venture capitalist investors have documented margin compression in undifferentiated, easily substituted software. For example, A16Z identified a broad trend of enterprise software startups running surprisingly low margins of 30%-40%, where low price points create a cost anchor against competitors.

2. Increased competition from AI-native entrants

AI-native startups will reimagine business processes. This mirrors how SaaS disrupted on-premise client-server systems. However, disruption does not automatically equal displacement.

Customers will benefit from faster innovation, but switching costs will not reduce substantially due to data migration, integration complexity, organizational retraining and porting customizations.

3. The expansion of agentic applications

As agents automate manual, repetitive workflows, the market for agentic applications will expand. Rather than “software as a service,” we may see a shift toward “service as software,” where domain workflows are embedded in orchestrated AI systems.

Why ZAIDYN remains hard to replicate in the AI era

These dynamics are especially relevant for platforms like ZAIDYN, where differentiation is not in the code itself, but in the accumulated domain expertise and operational scale behind it. For ZAIDYN, this includes:

• Decades of life sciences-specific business process logic, data management, data modeling expertise and calculation variations learned over hundreds of implementations.
• Six Sigma accuracy and operational consistency in high-stakes domains such as incentive compensation.
• Proven deployment at scale across more than 100,000 reps, billions of dollars of incentive payouts and thousands of customer engagement recommendations.

These advantages will carry forward in an AI-first world as well. Instead of a deterministic workflow, ZAIDYN will build expertise in creating and orchestrating a new system of agents. In fact, agentic architectures can be even more complicated than established software systems due to:

• Nondeterministic behavior
• Rapidly evolving models and tooling
• Jagged frontiers between dependable and hallucination-heavy tasks

Take the medical, legal and regulatory (MLR) review as an example of agentic workflows. As these agentic systems analyze thousands of documents, images, tags and citation sources and evaluate constantly evolving regulatory rules, systems must be optimized for prompt engineering, context window optimization, evals, model choice tradeoffs for cost and latency, guardrail design and observability frameworks.

The design and operation of these systems will only become more complex over time.

How consulting insight becomes operational systems

In practice, domain expertise is not just documented. It is encoded into software systems and, increasingly, into agents. This requires a combination of consulting expertise, product engineering capabilities and deployment experience to translate expertise into something that can operate at scale.

Build or buy? 4 questions leaders should answer before committing to new tech

Ultimately, the choice to buy software instead of building it in-house will boil down to how enterprise CIOs and CFOs answer four questions:

1. Just because we can build a full-fledged, proprietary software application, do we want to?
2. Is this the best use of our next $10M of investments over three to four years?
3. Is this our company’s core business or does it provide a meaningful competitive advantage?
4. Do we want long-term ownership of evolving AI systems, governance, compliance liability and model retraining cycles?