Model and Workflow Disagree on Next Step

What This Looks Like

The AI recommends, selects, or prepares a next step that does not match the current workflow state. The model may say to proceed when the workflow requires review, escalate when the process says to collect more information, or choose a tool or handoff that is not valid for the current stage.

Why It Matters

Workflow steps carry process authority. If the model’s recommendation conflicts with the workflow, users may not know whether to follow the AI or the process. This can cause skipped gates, wrong handoffs, premature actions, or stalled work because the decision and workflow state are out of sync.

Structural Signal

The model’s selected next action does not align with the declared workflow state or process rules. The issue is not simply a bad recommendation; it is a mismatch between model output and workflow authority.

Common Triggers

• The model sees task content but not the current workflow state
• The workflow state is implied rather than passed into the prompt
• The AI optimizes for task completion while the process requires a control step
• Routing rules are stored outside the model’s active context
• The model treats optional advice as an executable next step
• The workflow changes but prompts or examples still reflect the old sequence

When to Use This Issue

Use this Issue when the AI’s recommended or selected next step conflicts with the current workflow state, handoff requirement, routing rule, or process sequence.

When Not to Use This Issue

Do not use this Issue when the workflow itself has no next step defined. Do not use it for ordinary disagreement with the AI’s advice unless the disagreement is specifically between model output and workflow state.