## Prompt Describe a concrete engineering task in which you used an AI tool. Explain what work you delegated, what context or data you withheld, how you verified the output, and how you decided whether the workflow was better than doing the task without AI. ### Constraints & Assumptions - Use one bounded example rather than a survey of tools. - Treat generated output as untrusted until it passes task-specific checks. - Address privacy, licensing, security, or access-control concerns relevant to the example. - Distinguish speed of producing a draft from correctness of the final result. ### Clarifying Questions to Ask - Is the interviewer interested in coding assistance, analysis, documentation, or any AI-supported task? - Was the tool approved to receive the data involved? - Which quality metric or baseline should be used to evaluate the workflow? ```hint Explain the verification harness The strongest evidence is a check that could reject a plausible-looking but wrong output. ``` ### What a Strong Answer Covers - A reasoned choice of which part was suitable for assistance and which part required human ownership. - Deliberate handling of confidential inputs and tool permissions. - Specific validation such as tests, static analysis, source checking, adversarial examples, or peer review. - Examples of errors the tool made and how the workflow caught them. - A fair comparison of total effort, including review and repair time. - A stopping rule for cases in which AI output creates more risk or work than it saves. ### Follow-up Questions 1. What was the most convincing incorrect output you encountered? 2. Which information would you never place in the tool? 3. How would you evaluate the workflow over ten tasks rather than one? 4. Who remains accountable when generated code passes tests but later fails in production?