Summary
In this episode of the a16z podcast, partner Sema Amble and board partner Steven Sonowski — a former Microsoft executive — dig into what it means for enterprise software to go “headless” in an agentic world. The conversation starts with Salesforce’s Headless 360 announcement and quickly broadens into a structural argument: when AI agents interact with software through APIs rather than UIs, the locus of value shifts from the interface down to the data and business logic underneath.
The guests push back hard on the idea that a Postgres database plus a few REST endpoints can replicate something like SAP. They argue that decades of encoded business logic — the kind that captures how a specific company actually runs — is irreplaceable in the short term, even if the presentation layer becomes optional. The Excel and CSV export anecdote is a sharp illustration: enterprise software almost always can generate any report a user needs, but humans can never figure out how to ask for it. Language models, the guests suggest, finally solve that usability gap.
The broader takeaway is that today’s ad-hoc AI workflows — dumping PDFs into a model, querying data in natural language — are the embryonic form of tomorrow’s enterprise software products, much like spreadsheet-based CRM preceded Salesforce itself. The episode is essential listening for anyone tracking how incumbents like Salesforce, SAP, and Notion are repositioning for a world where agents, not humans, are the primary software users.
📺 Source: a16z · Published July 07, 2026
🏷️ Format: Podcast
