Glossary
Drift
What is drift in AI systems?
Drift is a change in how an AI agent performs after go-live, caused by something changing underneath it: the model, a prompt, a tool, a policy, or the data and cases it sees.
Also called: Model drift, Agent drift
Why it matters
An agent that passed every test can quietly get worse, or just different. Re-running your evals after every change, and watching quality on live work, catches drift before it reaches customers or the books.
Example
After a model update, a support team re-runs its past tickets through the triage agent, finds that it now sends more billing questions to the wrong queue, and rolls back the change.