Glossary
Error Budget
The acceptable failure rate or downtime allocated for an AI agent service, balancing reliability requirements against development velocity.
What is Error Budget?
Error budgets quantify acceptable unreliability, typically expressed as allowed downtime percentage or error rate over a time period. A 99.9% uptime target allows roughly 40 minutes of downtime monthly. When the error budget is consumed, teams focus on reliability rather than new features until stability improves. This framework balances user expectations against the reality that perfect reliability is costly and may slow innovation.
For AI agents, error budgets should account for both infrastructure failures and model-specific issues like hallucinations or incorrect outputs. Teams track budget consumption through monitoring and alerting, using incidents and degraded performance to update reliability priorities. Error budgets enable data-driven conversations about acceptable risk and appropriate investment in reliability versus new capabilities.
Example
A financial agent has a 99.95% accuracy error budget, allowing 50 incorrect decisions per 100,000 transactions monthly. After a prompt change causes 30 errors in the first week, the team pauses new features to investigate and improve validation before exhausting the monthly budget.
How Signet addresses this
Signet's Reliability dimension directly measures performance against error budgets. Agents with defined error budgets and strong tracking show operational maturity. Consistently staying within budget improves reliability scores, while frequent budget exhaustion indicates instability.
Build trust into your agents
Register your agents with Signet to receive a permanent identity and trust score.