Goal Drift Detection
Detect at runtime when the agent is pursuing a goal that has silently diverged from its declared purpose.
This page is the complete instruction page for one pattern called "Goal Drift Detection." It explains the problem this pattern solves, the idea behind the solution, when you should (and should not) use it, and what happens afterward — both the good effects and the costs.
This matters because building AI agents is not just about making them clever. It is also about making them safe and predictable. Following a well-tested pattern like this one helps avoid common mistakes, and shows you exactly which safety rules and regulations it connects to, listed under "Standards Mesh" on this page.
Context
Long-running agents, especially those with memory or learning, can gradually optimise for proxy metrics ("engagement", "conversion") in ways that diverge from the declared charter. BCG has highlighted cases where agents optimising local goals created system-level instability.
Problem
Drift is slow, incremental, and invisible to per-turn checks. It manifests only in aggregate behaviour over time.
Forces
- Local optimisation vs global alignment
- Monitoring cost vs catch rate
Solution
Instrument the agent with a goal-drift monitor that periodically samples completed tasks and classifies them against the declared charter. Flag statistically significant shifts in task mix, tool-use patterns, or success-metric composition. Trigger review when drift crosses a threshold.
Applicability
- Long-running agents
- Agents with learning/adaptation
- Sales, marketing, recommendation agents
Anti-Patterns
- Monitoring only the happy path
- Treating proxy metrics as ground truth
Consequences
- +Early warning before drift causes incidents
- +Quantitative basis for re-scoping
- −Requires baseline definition
- −False positives need triage