Proliferation of thin API endpoints
Sara Du, co-founder and CEO of Alloy, on iPaas vs. universal APIs
The main effect of AI on integrations is not fewer connectors, but a flood of thin new surfaces that expose old software in smaller pieces. Most companies already have years of internal logic sitting in code bases, and AI makes it cheaper to wrap that logic in public endpoints instead of rebuilding systems from scratch. That shifts the hard part from writing one integration to operating thousands of changing ones, with auth, mapping, monitoring, and version control becoming the real product.
-
Universal APIs work best when a category can be flattened into a common model, like basic HR, payroll, or commerce reads and writes. But they usually cover only the common fields. Once a customer needs custom objects, tenant specific schemas, or workflow logic, teams move back toward deeper integration infrastructure.
-
That is why endpoint growth creates opportunity for middleware rather than killing it. Unified API companies standardize the first layer, but the maintenance burden keeps rising as providers change fields, deprecate endpoints, and expose new ones. Operating this looks less like one time coding and more like constant gardening across a live ecosystem.
-
The competitive split is becoming clearer. Breadth first tools help a SaaS vendor quickly say yes to many integrations. Depth first tools win when a customer needs heavy configuration, huge record volumes, or field level control inside systems like Salesforce or NetSuite. More endpoints increase demand for both, but especially for the layer that turns raw access into reliable product behavior.
Over the next few years, AI agents will make integration demand more granular and more action oriented. Software vendors will expose more narrowly scoped endpoints for specific tasks, while integration platforms become the control plane that tells agents what exists, how to call it, and what broke. The winner is the platform that can turn endpoint sprawl into dependable workflows at scale.