In 2010, SAP paid $5.8 billion for Sybase. By 2014, the Sybase name was gone. The database was absorbed. Database capability became something a platform had rather than something a company bought. The standalone vendors that survived did so by moving up-stack into management, reliability, and enterprise support. The product became a feature. Value migrated.
Agents are entering that same current, and the water is moving fast.
The protocol layer is settling. MCP was donated to the Linux Foundation. Google's A2A attracted over a hundred enterprise supporters. These are no longer competing standards. They're shared plumbing. And once plumbing standardizes, the ability to connect pipes stops being worth paying for.
Constellation Research's Michael Ni put it plainly late last year: agentic AI is "merely a feature." The real game, he argued, is decision velocity. The agent is the mechanism. And pricing is accelerating the absorption. Enterprise software vendors, anxious about losing seat-based revenue to consumption models, are bundling agents into existing license agreements. Constellation noted that agentic enterprise licenses are becoming the norm as CIOs push back on unpredictable consumption pricing. When agents arrive inside an existing contract rather than as a separate procurement decision, they become a platform capability by default. CIOs wanted cost predictability. They got bundling.
This creates a specific problem for the current wave of agent startups. If the agent layer commoditizes the way databases did, "we have an agent" stops being a differentiator roughly as fast as "we have a database" did in the early 2000s.
The database parallel suggests where to look next. What survived commoditization was the infrastructure that made databases reliable, observable, and governable at scale. Monitoring. Access controls. Performance tuning. Backup and recovery. The boring stuff that determined whether the database actually worked in production or just worked in a demo.
Gartner projects agentic AI will resolve 80% of customer service issues without human intervention by 2029. That projection assumes agents running reliably, thousands of times, across changing conditions, with audit trails and cost controls. The distance between "agent that works once" and "agent that works at enterprise scale" is an infrastructure gap. And infrastructure gaps are where durable value tends to concentrate.
The operational layer that makes agents reliable, auditable, and governable across changing conditions is where differentiation is quietly collecting. The agent itself is becoming table stakes.
Nobody buys a standalone database anymore.