Where's the Line Between an AI Workflow and an Agent? Builders Weigh In

Charming_Mark9257 · reddit · 2026-09-18

A Reddit discussion distinguishes LLM apps, automation workflows, and agentic systems — arguing the real engineering problems begin after the first tool call: failure handling, who decides the next action, outcome verification, stopping criteria, human approval gates, permission boundaries, and auditability. The author asks whether tool use and planning suffice, or an agent must observe outcomes and adapt.

Original post →

More from coding & agent

coding & agent channel →