Is Software Going Headless?

Salesforce launched a headless product, making a bet that it's the data layer, not the UI, that sits at the center of value. In a structure where agents read and write directly to the DB without going through a human-facing UI, the advantage of UI-based user habits that sustained SaaS for the past 20 years is weakening.

Stickiness Factors of the SaaS Era

The moat of traditional Systems of Record was based on user habits formed through the UI, undocumented workflow rules, internal and external dependencies, and compliance significance. Switching costs were so high that replacing a CRM was likened to open-heart surgery, and replacing an ERP to open-heart surgery during a marathon.

Shift in Defensibility in the Agent Era

Defensibility is moving downward toward data models, permission systems, workflow logic, and compliance, and upward toward network effects, proprietary data generation, and real-world execution capability. Improvements in LLM reasoning ability and MCP's standardization of tool access have made this possible.

New Standards for AI-Native Systems

• Agent-friendly data models: schemas centered on capturing reasoning and actions, such as task, intent, thread, policy, and outcome
• Per-agent permission management: defining who can do what, through which agent, and under what policy
• Owning the action layer: implementing the closed loop of action → outcome capture → feedback → improvement
• Expanding real-world execution: connecting to physical execution such as coordinating field workers, logistics, and service teams
• Meaningful proprietary data: behavioral data uniquely generated by the product, benchmarks, and agent performance tracking

Three Paths for Buyers

Buyers must choose among adding agents to existing systems, building a DIY system of record, or purchasing an AI-native replacement. The most promising next-generation businesses are those that extend beyond simple data storage into real-world execution and multi-party coordination.

Categories deeply tied to compliance (payroll, ERP) will remain difficult to switch away from for a while, but for verticals that are technically underserved (manufacturing, construction, field service), there is significant opportunity for AI-native solutions. The key structure is one where data sits in the background while real-world execution comes to the forefront.