Perceive–Reason–Act

The micro-level control loop that every agent in an APM system runs (sources/2026-calvanese-agentic-bpm-manifesto, §2.2.2). An agent alternates between:

Sensing its environment to update its internal state
Reasoning to select actions aligned with its goals
Acting to influence the environment

Three conceptual modules

Perception module

Observes the state of the environment, context (situational constraints, location), and other agents. Manages sensors, percepts, and belief update functions. Keeps the agent's world model current.

Reasoning module

Core of knowledge representation and decision-making. Continuously informed by the framing mechanism; maintains:

Mental model — the agent's knowledge, perceptions, beliefs, including its view of "the process" as perceived through the frame.
Intentional model — the agent's goals as shaped by social and normative dispositions (obligations, roles, collective goals, prohibitions).
Identity (agent ID, type).
Behavioural dispositions and plans (policies, triggers, reactions).

LLMs are explored as implementations, though their sufficiency is disputed (see refs [41][42] in source).

The reasoning module houses the four APM capabilities: framing, explainability, conversational actionability, self-modification.

Action module

Performs actions that change environment state and sends messages to other agents. Defines the agent's capabilities (skills, resources) and ability to socially interact.

Theoretical foundations

Draws on foundational ontologies — DOLCE, GFO; agent-oriented modelling — *[[frameworks/tropos-i-star|i, Tropos]]**; classical AI constructs — BDI, FIPA; mental-model theory of reasoning.

concepts/agentic-bpm · concepts/framed-autonomy · concepts/process-awareness