SaaS Must Shift to Agent-Ready APIs as Teams Still Need Shared Systems of Record

4 articles · Updated · O'Reilly Media · Jun 1

SaaS is not being displaced by prompt-built apps, the report argues; it remains essential because companies need shared data, common metrics and collaboration across teams rather than one-off tools for individuals.
Agent-built software breaks down when each worker creates a slightly different CRM or workflow, leaving incompatible schemas, fragmented reporting and new data silos that make company-wide visibility harder, not easier.
To stay relevant, SaaS vendors need more than dashboards: they must deliver raw, machine-readable data through APIs built for agents, including structured state, permissions, relationship graphs and reliable update paths.
That shift also requires enterprise plumbing that agentic tools still lack—sharing, requirements management, collaboration, testing, versioning and security—especially as models drift and prompt-injection risks remain unresolved.
The piece says incumbents such as Salesforce and Google have the data expertise to build that next-generation stack, but large vendors risk being blindsided if they adapt too slowly.

As AI agents fragment enterprise data, can legacy SaaS giants truly become the new 'system of record' for machines?

Will AI orchestration platforms become tech's next giants, or just features absorbed by incumbents like Google and Salesforce?

With CIOs facing looming budget cuts, how can they prove the financial value of AI agents before the 2026 deadline?