Zain Hoda’s core argument: your entire Salesforce instance probably fits on a thumb drive. When an agent can clone all that data in seconds and sit on a complete, synchronized copy, you stop talking to Salesforce — you talk to the agent. The SaaS product becomes a “dumb write endpoint” where data gets created before being pulled into the agent’s context. Data moats evaporate because “we’re where the data lives” stops being a valid competitive answer.
This reframes what systems of record actually are. The rigid schema, the structured tables, the mandatory fields — these exist because the application could only do a fixed set of well-defined actions. The database had to match those actions exactly. But when agents become the interface layer, they can reason over unstructured data directly. The rigid app was the constraint that demanded the rigid schema.
This connects to Files are the universal interface between humans and agents — when agents can reason over raw artifacts (contracts, emails, invoices), the elaborate data entry pipeline becomes unnecessary. The agent derives structure on demand rather than requiring it upfront. And it explains why B2B becomes B2A — agents become the buyer accelerates the shift: agent-consumers don’t need pretty dashboards or structured input forms, they need data access.
Rate-limiting is a losing strategy. Systems that fight agent access by restricting APIs are just making their product worse. The implication: as agents get better at reasoning over raw data, the elaborate schema and the rigid app on top both become scaffolding destined for obsolescence. And Context is the product, not the model takes on new meaning — the context isn’t just your own data, it’s every system of record your agent can clone and reason over.