The correct options are Use Copilot to extract duplicated logic into reusable helpers and then manually verify behavior and performance and Ask Copilot to generate unit tests for the refactored areas to reduce the chance of regressions.

Using Copilot to extract helpers accelerates consolidation while you keep architectural intent intact. Manual verification through code review, representative benchmarks and integration tests ensures behavior and performance remain correct across modules and prevents subtle coupling issues in a long lived codebase.

Having Copilot draft unit tests for refactored code increases coverage and catches regressions early. You can guide it with edge cases and domain constraints and then refine the tests to lock in expected behavior. Running the test suite across all modules provides confidence to proceed with further refactoring.

Allow Copilot to perform refactoring and write tests without any human oversight is unsafe because it removes critical review and domain context. FinTech systems demand strict controls so human code review and validation are necessary to maintain correctness and compliance.

Depend on Copilot to automatically discover and resolve all function and module dependencies during extraction is unrealistic because tools cannot reliably infer every implicit dependency such as global state, configuration, feature flags or build scripts. Engineers must map dependencies and validate with integration tests.

Configure Cloud Build to auto merge Copilot generated refactors to production without manual approval increases risk by bypassing protection mechanisms such as required reviews and status checks. Production changes should pass through protected branches with approvals and automated checks before deployment.

Full AWS Practitioner Certification Question