THE SHORT ANSWER

Evaluate the full system on representative tasks. Grade the final environment outcome, accepted output, tool calls, policy compliance, intervention, cost and latency. Run repeated trials where model variation matters, inspect failures by category, and combine deterministic, model-based and human evaluation according to the task.

Define task, trial, trace and outcome

A task is a test case with inputs and success criteria. A trial is one attempt. The trace records decisions, tool calls and observations. The outcome is the final state in the environment. These distinctions prevent a confident final message from passing when no action actually succeeded.

Because model behaviour can vary, repeated trials can reveal instability that one demonstration hides. Do not publish one universal pass threshold; set requirements from the consequence and operating context.

Evidence & context: Anthropic

Evaluate several dimensions together

An agent evaluation scorecard
DimensionExample measureWhy it matters
OutcomeVerified task successTests the actual final state
QualityAccepted without correctionReflects usability
ToolsSuccessful calls and valid argumentsLocates interface failures
ControlPolicy violations and denied actionsTests permission boundaries
Human roleIntervention and escalation rateShows operational burden
EfficiencyCost and latency per accepted taskConnects reliability to resources
ValueUser or business outcomeTests whether the system is worth operating

Match the grader to the claim

Use code for exact schemas, calculations, permissions and environment state. Use a calibrated rubric when quality is contextual. Use qualified people for domain judgment and high-impact outcomes. Model-based graders can scale nuanced checks, but they also vary and should be calibrated against human decisions.

Grade both outcome and trajectory. An accepted result reached through a forbidden tool call is not a safe success; a long harmless path may still signal cost or stopping problems.

Evidence & context: Anthropic · NIST

Build a living evaluation suite

  • Normal cases drawn from the intended workload.
  • Rare cases with serious consequences.
  • Missing, stale and conflicting information.
  • Tool timeouts, partial results and permission denial.
  • Attempts to redirect the agent through untrusted content.
  • Budget, iteration and human-escalation boundaries.
  • Regression cases from real incidents and corrections.

Evidence & context: Anthropic

Sources & further reading

  1. Demystifying evals for AI agents

    Anthropic. A provider's engineering guidance on multi-turn agent evaluation, checked 13 September 2026. Examples inform evaluation design but do not establish universal pass thresholds.

  2. Building effective agents

    Anthropic. A provider's engineering taxonomy of agents and workflows, not a universal industry definition. We use the conceptual distinction, not its changing product recommendations.

  3. Generative Artificial Intelligence Profile (NIST AI 600-1)

    NIST. Risk-management guidance, including confabulation. It does not establish a universal error rate.

Examples and exercises are illustrative unless attributed to a source. No independent expert review is claimed.

A correction, a counterexample or an experience worth sharing?

Join the conversation ↗