Conversational Actionability

The third required capability of an APM agent (sources/2026-calvanese-agentic-bpm-manifesto, §3.3): the ability to combine interaction and enactment — to coordinate with other agents and humans, and to link conversations to concrete process actions.

Two requirements (from [ref 17] in the manifesto)

Process-aware conversation — interact with users and external agents via a conversational interface to support, trigger, and guide actions related to process enactment.
Process-aware actionability — conversations trigger process executions: decisions, actions, or evolution (priorities, process changes).

Four enactment roles

Depending on the role of the agent, actionability may cover:

Role	Purpose
Query	Provide information on processes at model or execution level
Recommend	Suggest adaptations / future evolutions of process instances
Create	Elicit models from domain knowledge and process data
Execute	Trigger actions that move process instances to a new state

Interaction modalities

Natural language (between agents and humans)
Domain-specific languages (process diagrams, dashboards)
Formal communication languages (multi-agent interaction protocols, FIPA, KQML)
Emerging human-readable token encodings (e.g., TOON) and protocols (MCP, ACP)

Federated vs centralised

APM systems cater to federated perspectives — unlike traditional BPM's single point of truth — requiring consensus resolution or tolerance of variance among agents' responses.

concepts/agentic-bpm · concepts/explainability-apm · concepts/self-modification