LunaPromptsThe code runs. The prompt classifies the sample inputs. The architecture has the expected boxes. Starting a completely new interview now would leave the most useful material untouched: the candidate’s own work.
A Viva creates a focused conversation around that material. It asks for the decision behind a line, the consequence of a change, or the test that would make an answer more trustworthy.
LunaPrompts AI Drill Down, also called AI Viva, asks candidates follow-up questions about their assessment work. It helps hiring teams explore the reasoning behind a piece of code, a prompt, or a system design while that work is still part of the evaluation.
Drill into a support-classification prompt
Technical interviewerIllustrative scenario01 / 03A prompt classifies the examples
Each supplied support request has one obvious category.
Suppose a candidate creates a prompt that assigns support tickets to Billing, Technical, or Account. The supplied examples are clear, so the initial output looks strong. Add a ticket that says, “My invoice increased and the service stopped working.”
Ask which category the prompt should select and why. The goal is not to hide a secret correct label. The goal is to see whether the candidate notices an undefined business rule. They might ask for a priority policy, propose an ambiguous category, or route the case for review. The answer should fit the stated workflow.
Then ask for two additional test examples. One should check the chosen rule; the other should challenge it. This connects explanation to an artifact the team can inspect. A polished verbal answer without a change to the task remains incomplete evidence.
Keep the discussion focused. One well-chosen change can reveal more than a long sequence of generic questions. For reviewers, record the original behavior, the changed condition, and the candidate’s response. Use the same level of challenge across candidates while grounding the question in each person’s actual submission.
Follow the work instead of restarting the interview
AI Viva uses the candidate's assessment as the subject of the conversation. In the configured workflow, the discussion can happen during the assessment or as a follow-up to completed work.
For coding, the discussion might explore an implementation choice or an unhandled input. For prompt engineering, it might examine an instruction, an output constraint, or a failed example. For system design, it might investigate a component, a bottleneck, or an alternative architecture.
The purpose is to add explanatory evidence to the practical result.
What useful drill-down questions look like
Explain a decision
“You chose a queue for this part of the workflow. What problem does it solve, and what new failure case does it introduce?”
This tests whether the candidate can connect an architectural choice to a requirement rather than repeat a familiar pattern.
Investigate a failure
“Your prompt works for the first document but returns an unsupported answer for the second. What would you inspect first?”
This explores diagnosis. A useful answer should connect a proposed investigation to the behavior that went wrong.
Adapt the solution
“The API can now receive the same request twice. What would need to change?”
This checks whether the candidate can reason beyond the exact problem they prepared for.
Discuss a limit
“What would you want to test before putting this into production?”
This invites the candidate to identify uncertainty and prioritize verification. Recognizing an unresolved issue can demonstrate judgment even when the initial answer is incomplete.
Use Viva to deepen review, including integrity review
An explanation can support confidence in authorship and understanding. A contradiction may identify a topic for further review. Neither a smooth answer nor a hesitant one should settle the candidate's integrity by itself.
Consider the actual submission, the permitted tool policy, and the specificity of the candidate's explanation. For a technical role, the goal is to understand their reasoning about the work, not to reward a particular speaking style.
A review framework your team can use
The following is an editorial checklist for this workflow. Adapt it to the role, the candidate journey, and your configured environment.
| Review area | What to establish |
|---|---|
| Explain | Why did you choose this rule? |
| Challenge | What happens with this ambiguous input? |
| Adapt | Show the smallest change you would make |
| Verify | Which example would show the change failed? |
Do not score speaking speed as technical understanding. Let the candidate use the submitted work to explain a concrete decision.
Put the guide to work
Choose one line, prompt instruction, or component from a sample submission. Write an explain question, a changed condition, and a verification question.
Download the editable worksheet to record the decisions and open questions with your team. The examples in this guide are illustrative, not customer results or validated selection rules. Product availability and setup depend on the workflow agreed for your account.
A focused checklist, a worked scenario, and questions to resolve together.
Frequently asked questions
How is AI Viva different from an AI interview?
AI Viva examines a specific assessment submission. An AI interview can cover broader role requirements and serve as a separate screening round.
Can Viva be used when AI coding tools are allowed?
Yes. Candidates can explain the work they directed, how they reviewed it, and which parts they changed. That discussion helps evaluate responsibility for an AI-assisted result.
Does difficulty explaining an answer prove copying?
No. It identifies a gap that needs context. Reviewers should consider the candidate's work, the question, and any communication or assessment constraints before reaching a conclusion.
Keep building your hiring workflow
- AI skills: The answer looks right. Can your candidate explain why?
- Assessment integrity: A flag is a question. Find the evidence before the verdict.
- AI interviews: More first conversations. A clearer handoff to people.
Explore ai viva with LunaPrompts or browse all 19 hiring field guides.
