Workflow drift

A measurable divergence between an AI workflow's current behavior and its intended contract after prompts, tools, data, models, or user inputs change.

Workflow drift is the gradual, often invisible divergence between how an AI workflow was designed to behave and how it actually behaves after repeated use, prompt edits, and changing inputs. Small prompt tweaks accumulate. The data it retrieves shifts in format or coverage. Users start feeding it inputs the original design did not anticipate. No single change breaks the system; the drift happens across dozens of incremental adjustments. A weekly report agent that once produced crisp summaries begins padding output with filler because someone widened its context window "just in case." A triage agent that once escalated borderline cases starts letting them through because a prompt edit softened its threshold language. The result looks like work but quietly diverges from the original intent.

Builder example

Plausible prose can hide drift when a team checks appearance rather than the contract. Versioned instructions, representative evaluation cases, source-freshness checks, output metrics, and downstream error signals can detect it before a user reports a failure. The monitor should cover the component that can change: a prompt diff cannot detect a provider model update or a data-schema shift by itself.

Common confusion: Workflow drift is different from a sudden breaking change. It is incremental and hard to detect with spot checks. A single output looks fine. The problem only becomes visible when you compare a batch of recent outputs against the original acceptance criteria.