AI engineering
AI agent or workflow: how to choose
A practical decision guide to AI agents and fixed workflows: task boundaries, failure cases, evaluation and the evidence to request before a pilot.
THE LAB NOTEBOOK
AI engineering
Choosing useful AI systems: task boundaries, evaluation and the path from a demonstration to an accepted result.
RSSWhere to start
Start by describing a task and an independently checkable result. The articles in this category help distinguish fixed workflows from systems that choose their next action. Compare the proposed design with an existing process using the same inputs. Keep accepted outcomes, manual correction, cost and failure modes separate. Before expanding a pilot, review the allowed tools, the person who accepts the result and the evidence available when a run fails. The related knowledge and reliability guides cover source support and operational readiness.