The “SaaS is dead” narrative is getting louder.
AI coding tools can now generate functional applications in minutes. Agentic AI can complete workflows across systems without human intervention. Investors are watching revenue growth slow and drawing conclusions.
Two arguments are driving this conversation. Both deserve attention. Neither supports the headline.
What’s happening isn’t extinction. It’s economic pressure and interface transformation. And enterprise buyers should be focused on very different questions.
The First Argument: AI Coding Tools Make SaaS Obsolete
Tools like Claude Code and Codex have changed the economics of software development. They compress timelines. They lower the skill floor. They make it plausible to build usable applications far faster than even two years ago.
That matters.
But here’s the part the narrative skips: enterprise development capacity is limited. The strategic question isn’t whether AI can build a CRM or ERP. It’s whether rebuilding commoditized infrastructure is the right use of the engineering capacity AI unlocks.
In most cases, it isn’t.
AI coding tools create leverage when they’re applied to differentiated workflows — the integration no marketplace covers, the process unique to how your business operates, the capability that reflects your competitive advantage.
Rebuilding Salesforce from scratch doesn’t create advantage. It recreates infrastructure.
And the total cost of ownership rarely shows up in the demo.
An AI-generated application still requires:
- Hosting and monitoring
- Security reviews
- Compliance certifications (SOC 2, HIPAA, FedRAMP)
- Identity and access management
- Audit logging
- Disaster recovery
- Long-term ownership
SaaS vendors absorb these responsibilities at scale. Custom-built applications start at zero. Enterprises have seen this movie before. It was called shadow IT.
AI coding tools are an accelerant for custom development. They are not a procurement strategy for replacing mature systems of record.
The Second Argument: Agents Reduce the Need for SaaS Seats
This argument is more serious.
Agentic AI introduces automation at the workflow level. An agent can complete an expense report, resolve a service case, trigger procurement actions, or update a CRM pipeline without human intervention.
Today, agents sit on top of SaaS platforms. They read and write into systems of record.
They consume SaaS. They don’t replace it.
But here’s the real investor concern: if agents complete end-to-end workflows, how many human seats are displaced? And if pricing is per-seat, what happens to revenue?
That’s a legitimate question.
The pressure here is economic, not structural. Systems of record don’t disappear. Monetization models may evolve.
Vendors that rely entirely on per-seat pricing for task execution will feel compression. Vendors that move toward hybrid pricing — consumption, workflow-based, or outcome-linked — have a path forward.
The disruption is about revenue architecture.
It is not about eliminating the operational backbone of enterprise software.
Where SaaS Is Actually Vulnerable
Not all SaaS vendors are equally positioned.
Higher risk categories include:
- Single-function tools with minimal integration depth
- Applications without proprietary datasets
- Products lacking compliance or regulatory infrastructure
- Software that can be recreated easily with generative tools
If a product can be rebuilt with a prompt, defensibility becomes questionable.
Lower risk categories include:
- Deep systems of record
- Platforms with extensive integration ecosystems
- Industry-specific regulatory workflows
- Vendors with cross-customer benchmarking data
In these cases, AI often increases platform value rather than replacing it.
What Would Actually Have to Change for SaaS to Decline?
For SaaS to become structurally less relevant, several conditions would need to materialize:
- AI-generated code becomes reliably maintainable at enterprise scale — not just generatable.
- Compliance infrastructure becomes automated and commoditized.
- Integration becomes dynamically AI-negotiated across counterparties.
- Enterprise observability and governance wrap automatically around new AI applications.
- Legal liability for AI-built systems becomes predictable and manageable for enterprises.
These conditions are evolving. They are not yet universal.
Until they are, the “SaaS is dead” narrative runs ahead of the evidence.
What Enterprise Buyers Should Do Instead
Rather than debating extinction, enterprise leaders should focus on allocation and evaluation.
First: Where should AI coding capacity be applied?
Point it toward differentiated capability — not recreating commodity infrastructure.
Second: Which vendors are genuinely building for an agentic world?
Evaluate them on:
- Production-ready agentic functionality
- Pricing flexibility beyond per-seat
- Ability to convert historical data into operational intelligence
Avoid vendors that simply layer conversational interfaces onto legacy systems without meaningful workflow automation.
Demand proof of measurable business impact.
The Bottom Line
SaaS revenue models are under pressure. That pressure is real.
But pressure is not death.
The enterprise stack is shifting toward agent-driven interfaces and automation. Systems of record remain foundational. Compliance infrastructure still matters. Integration ecosystems still matter. Accountability still matters.
The question isn’t whether SaaS survives.
The question is which vendors adapt quickly enough to remain relevant — and which enterprises deploy AI capacity where it actually creates advantage.
That’s the strategic lens buyers should use.
A version of this Is SaaS Dead was also posted in my LinkedIn newsletter. You can subscribe to the AI Decoded newsletter here.
Lopez Research 