# How to Design an API Portfolio for Developers and AI Agents

This session explains how API programs should evolve for the AI era by shifting from governing project-based interfaces to designing portfolios of business capabilities that serve both human developers and AI agents. Mo Amin and Igor Gorbunov of EPAM argue that traditional API programs were built around assumptions that applications are the main consumers and that software delivery is constrained by cost and time, but AI changes both sides: agents discover and choose interfaces at runtime based on semantics, while AI-assisted development makes building APIs and MCP servers fast and cheap, making curation, ownership, authorization, and observability more important. They propose organizing APIs around durable business capabilities, defining a shared capability contract, and exposing it through two consumption layers - REST APIs for developers and MCP tools for agents - without forcing a one-to-one mapping. They also demonstrate an EPAM accelerator that turns capability discovery, value definition, API design, and MCP design into delivery-ready artifacts in days rather than weeks.