The SaaS Structural Assessment: Five Verdicts on What Survives the Agentic Transition

The Counter-Argument: Why the Old Chain May Be More Resilient Than Feared

The bearish narrative has real structural logic. But several counter-arguments deserve honest consideration.

Data gravity is real. Enterprises spent decades embedding operational data into Salesforce, SAP, and Workday. Migration costs are enormous. Agents operate over this data, but they still need it to exist somewhere reliable. The system of record doesn’t disappear — it becomes infrastructure — as explored in the economics of AI compute infrastructure — .

Compliance creates stickiness. In regulated industries, the system of record is a compliance artifact. Auditors reference it. Regulators inspect it. Changing it requires extensive validation processes measured in years, not months.

The the shift from SaaS to agentic service models — -value-chain-six-categories/”>orchestration layer is unproven at scale. Camunda’s research found 80% of organizations say most AI agents are still chatbots or assistants, not autonomous operators. Only 11% of agentic AI use cases reached production in 2025.

Incumbents can embed intelligence. Microsoft embeds Copilot across Office. Salesforce integrates Agentforce with CRM data. ServiceNow positions itself as the semantic layer. These aren’t passive incumbents waiting to be disrupted.

Five Structural Verdicts

1. The system of record survives as infrastructure, not as the primary value capture point. Owning data still matters, but it is no longer sufficient. The data layer becomes necessary but commoditized — essential, but not where the premium margin lives.

2. The orchestration layer is the new commanding height. Whoever controls how agents interpret intent, coordinate across systems, and deliver outcomes captures the largest share of enterprise technology value.

3. Agent identity and governance is the sleeper moat. In regulated industries, this becomes the most defensible position. Agent IAM will be as sticky as Active Directory was in the PC era.

4. The middle gets hollowed gradually, not suddenly. UI-dependent SaaS applications lose value over years because enterprises move slowly and agents aren’t yet reliable enough for fully autonomous operation. But the direction is clear and irreversible.

5. The quiet killer is headcount compression. The real disruption isn’t agents replacing software — it’s AI reducing the headcount that uses the software. If 10 agents do the work of 100 people, you need 10 seats, not 100. Revenue falls 90% without model shift. The software survives. The revenue doesn’t.

This is part of a comprehensive analysis. Read the full analysis on The Business Engineer.