Recent advances in large-language-model tooling have propelled AI agents, autonomous, goal-driven software entities, into everyday business operations. While research proposals for new ‘agent-aware’ notations abound, industry still relies on the well-established standards BPMN and PASS to document and govern its processes. This paper, therefore, addresses a pragmatic question: Are those existing notations already sufficient to model AI agents? We first extract four canonical agent capabilities—environment perception, autonomous action, goal orientation, and temporal persistence—and map them to BPM constructs: subjects in PASS and pools and lanes in BPMN. A conceptual analysis, a hands-on modeling study of an intelligent call-center scenario, and empirical validation through expert interviews show that (i) PASS offers an immediately natural fit, and (ii) BPMN attains functional parity when collaboration diagrams and disciplined message flows are applied. No structural limitations were identified, and perceived BPMN weaknesses can be mitigated by modeling guidelines. The findings, supported by expert insights, reinforce the hypothesis that mainstream BPM notations already cover the descriptive and analytic needs of AI-agent integration. Rather than inventing new diagram types, practitioners can extend their current tool chains and governance routines. Future work should catalog domain-specific use cases, enhance conformance tooling, and explore oversight mechanisms for self-learning agents.