Oracle AI for Fusion Cloud CX: AI Agent Studio Developer Training Ed 1 — Lesson 3: Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling

Learning Objectives

• Assess the dependencies, controls, and expected business outcome for Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling.
• Demonstrate the workflow and leave reproducible evidence in agent evaluation report and red-team findings.
• Recommend a preventive improvement after resolving a grounded response citing outdated customer information.

Detailed Topics

• Testing
• Evaluation datasets
• Response quality
• Hallucination controls
• Permissions
• And exception handling
• Decision ownership, prerequisites, and acceptance evidence for Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling
• Relevant access, data, integration, performance, recovery, and support considerations

Instructor Demonstration

The instructor builds a representative scenario for Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling, narrating each decision and showing how it contributes to a safe and useful Fusion Cloud CX AI agent designed, grounded, tested, deployed, monitored, and governed for business service. The demonstration records agent evaluation report and red-team findings so learners can compare configuration intent with observable behaviour.

Hands-On Exercise

Complete a guided task covering Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling. After establishing a working baseline, reproduce “a grounded response citing outdated customer information,” isolate the responsible setting or data condition, apply a reversible fix, and update agent evaluation report and red-team findings with before-and-after proof.

1. Map the current process and mark where the expected result can first be observed.
2. Run the scenario while preserving timestamps, job identifiers, and relevant output.
3. Apply a reversible correction only after evidence supports the cause.
4. Retest the original case plus an unaffected case and update the handover record.

Scenario Analysis

• Clarify the business trigger, affected roles, required records, and acceptable result for Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling.
• Compare at least two implementation or correction choices and identify which evidence would distinguish them.
• Decide how agent evaluation report and red-team findings should be reviewed, approved, retained, and handed to the next owner.

Troubleshooting Guidance

For a failure in Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling, separate access, setup, data, integration, and timing causes. Compare the observed sequence with agent evaluation report and red-team findings; if “a grounded response citing outdated customer information” is confirmed, make the smallest reversible correction and document why broader changes were rejected.

Validation Checklist

• Prerequisites and role permissions for Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling have been confirmed before execution.
• The result agrees with the expected values documented in agent evaluation report and red-team findings.
• Exceptions are explained with evidence rather than hidden by manual correction.
• The final record names the approver, monitoring point, and next improvement action.

Lesson Outcome

Learners can apply Testing, evaluation datasets, response quality, hallucination controls, permissions, and exception handling, explain the decisions made, and defend the result with agent evaluation report and red-team findings.