Enterprise software companies built their businesses around one assumption that went unquestioned for decades. People are the users. Employees log in, click through interfaces, update records and move between applications. The software earns its seat license because a human being sits in front of it.
That assumption is starting to crack. AI agents are becoming capable of retrieving information, making decisions and executing workflows on behalf of employees. The question circulating across enterprise software boardrooms is what happens to the business model when the human in the browser tab is no longer the primary user.
Why AI agents could make enterprise platforms more important
Salesforce is already confronting this. In April, the company launched Headless 360 at its TrailblazerDX conference. The architecture exposes every Salesforce capability as an API, MCP tool, or CLI command so that AI agents can access data, workflows, and business logic without opening a browser.
Co-founder Parker Harris framed it directly: “Why should you ever log into Salesforce again?”
That question points at something counterintuitive. The screens may matter less. The systems underneath may matter more.
An enterprise CRM or IT service management platform holds years of accumulated business rules, permissions, workflows, and records. That is exactly what an AI agent needs before it can act. The interface depreciates. The infrastructure underneath it appreciates.
“The platforms and systems of record are not losing value here,” Yoav Kolodner, CEO of Tribal and a former VP of engineering at Salesforce, told TheStreet.
“If anything, they matter more, because what an agent actually needs is the object model, the business logic that took fifteen years of edge cases to accumulate, the permission model, the audit trail. None of that is easy to rebuild, and all of it is exactly what an agent consumes.”
Tribal is an enterprise AI platform that maps the metadata and dependencies inside systems of record. It lets AI agents operate within existing permissions and governance frameworks.
The readiness problem Kolodner describes is real. An agent that sees three fields called Status, Status_c, and Stage_Old_c will act on whichever one it finds first.
A human admin knows which one is live, but an agent does not.
