SaaS Without a Moat: The Enterprise Shift to AI-Native Custom Builds

Salesforce’s recent stock slide is a canary in the coal mine. The company beat on earnings but issued weaker guidance, particularly around its AI platform. Investors punished the stock, wiping out nearly a quarter of its market value in 2025.

This matters because Salesforce is more than a software vendor — it’s an emblem of SaaS itself. If its growth narrative falters, it raises questions about the model that’s dominated enterprise IT for two decades. The question isn’t whether AI adds value. It’s whether AI undermines the very foundations of the SaaS rent-based model.

As an enterprise data architect and data scientist working in financial services, I believe the answer is yes. SaaS as we know it is losing its moat. Generative and agentic AI are enabling enterprises to build and govern custom capabilities faster, cheaper, and safer than renting generic platforms.

SaaS Was About Speed and Standardization

The appeal of SaaS was straightforward:

Faster time to value.
Standardized capabilities that avoided reinvention.
Predictable subscription pricing.
Vendor-managed infrastructure and updates.

For banks and insurers, SaaS has historically offered access to enterprise-grade CRM, HR, or analytics without long internal build cycles. The model has made sense in a world where development was typically slow, talent was generally scarce, and compliance meant proving a vendor’s credentials.

AI Collapses the Rationale

That world is ending.

Developer throughput is accelerating. Studies show AI tools boost coding productivity by 20–50%, and nearly half of code in some enterprise repos is now machine-suggested.
Agentic frameworks extend the reach. Beyond autocomplete, agents can test, refactor, and deploy code — compressing cycles that used to require teams of specialists.
Compliance can be designed in. AI isn’t just spitting out code; it’s increasingly capable of embedding lineage, explanations, and audit trails directly into builds.

For enterprises, this means the historical argument for SaaS — “we can’t afford to build or maintain this ourselves” — is eroding.

Why pay per-seat for a one-size-fits-all CRM when you can generate a custom, compliant workflow on demand, tied directly to your data and tuned to your regulators’ expectations?

SaaS and Services Under Pressure

The implication isn’t that enterprises need no external help. Rather, the nature of services is shifting:

From integration to orchestration. Instead of gluing SaaS modules together, services firms will help orchestrate AI-native pipelines and evaluation harnesses.
From configuration to compliance. The new premium is assurance — proof of explainability, bias testing, drift detection, lineage.
From rent to IP accelerators. Enterprises won’t pay for generic dashboards; they’ll pay for accelerators, governance templates, and domain-specific ontologies.

In this model, SaaS subscriptions become harder to justify. Services survive, but as enablers of custom AI-native builds rather than as implementation arms of external platforms.

A Scenario View: SaaS Under Pressure

What portion of today’s SaaS spend survives the AI transition? The numbers won’t be exact, but directionally we can see where it’s heading.

By 2030, I expect enterprises will keep 60–70% of their current SaaS footprint. SaaS survives in commodity categories like payroll or HR, but for critical workflows, AI-native builds begin to displace vendor subscriptions.
By 2040, the balance could flip. Only 30–50% of today’s SaaS may remain. In the most aggressive scenario, SaaS is relegated to niche utilities while most enterprise-critical functions are rebuilt and maintained in-house with AI dev pipelines.

The likely path is the middle one: SaaS doesn’t vanish, but it shrinks dramatically as enterprises reclaim ownership of their technology stack.

Lessons from Past Transitions

We’ve seen this movie before.

Mainframes to client/server: Enterprises moved off vendor-controlled monoliths to distributed architectures.
On-prem to cloud: Hardware vendors lost ground as AWS, Azure, and Google Cloud commoditized infrastructure. Integrators didn’t vanish — they shifted to cloud consulting.

AI represents the next cycle. Just as cloud commoditized hardware, AI commoditizes SaaS functionality. Enterprises regain control of build-vs-buy — and this time, the balance tilts toward build.

What It Means for Banks and Regulated Industries

For banks, this is not just a cost play. It’s a risk and compliance decision:

Regulatory alignment. When we build, we can embed controls at the source. SaaS always required negotiating around vendor certifications and black-box processes.
Data governance. AI-native builds allow tighter control over sensitive data, reducing exposure inherent in multi-tenant SaaS.
Flexibility under regulation. Regulators move faster than SaaS roadmaps. In-house AI-native systems can adapt compliance logic in weeks, not years.
Talent models. This doesn’t mean armies of coders; it means smaller teams augmented by AI, supported by compliance engineers and data scientists.

The business case is straightforward: custom builds are no longer prohibitively slow or expensive. They may, in fact, be safer.

The Services Industry’s Next Act

For services firms, I suggest that survival depends on shifting quickly:

Develop reusable AI accelerators that embed domain knowledge.
Provide compliance-as-a-service — assurance wrappers, evaluation frameworks, lineage artifacts.
Become orchestrators, not coders — managing swarms of AI agents across the SDLC.

The firms that cling to SaaS implementation hours will find themselves in the same position as hardware resellers in 2010: staring down commoditization.

The Enterprise Takeaway

Salesforce’s stumble highlights a broader truth: the SaaS model is running out of runway. AI changes the math. Enterprises can now build faster, govern more tightly, and adapt more quickly than renting a generic platform allows.

This doesn’t mean services disappear — but it does mean SaaS as the default model for enterprise software is under real threat.

A Closing Question

If your major SaaS contracts expired tomorrow, could you rebuild their core capabilities in-house with AI-native tools — and would that be a better fit for your compliance and integration needs?

That’s the real longevity question enterprises must start answering.