What is the difference between reactive and proactive AI agents?

Reactive agents respond to explicit triggers (user queries, events, API calls) while proactive agents autonomously initiate actions based on observations and goals. Reactive pattern: user asks a question, agent retrieves information and responds. Proactive pattern: agent monitors data streams, detects conditions matching defined goals, and takes action without being asked. Examples of proactive behavior: monitoring email for urgent items and drafting responses, watching database metrics and auto-scaling infrastructure, scanning news feeds for relevant competitive intelligence and briefing stakeholders, and detecting anomalies in financial data and flagging for review. Implementation: proactive agents need a goal system (what outcomes to pursue), a perception module (monitoring relevant data sources), a planning module (deciding when and how to act), and an action module (executing decisions). Architecture: event-driven loop that polls or subscribes to data sources, evaluates conditions against goals, plans interventions, and executes with appropriate guardrails. Challenges: avoiding alert fatigue (agents should batch and prioritize proactive actions), respecting user preferences (notification timing, importance thresholds), and managing false positives. Hybrid approach: most production agents are primarily reactive with selective proactive capabilities for well-defined monitoring scenarios.