What Happens When an AI Leads the Interview? An Honest Look

The question almost always comes first: does it actually work? Can an AI conduct a real conversation — or is it just a dressed-up questionnaire in the end?

That's a fair question. And it deserves an honest answer — not a marketing claim.

This article shows concretely how an AI-led interview works in Qerie: from the first message to the finished insight. With real conversation examples. And with a clear acknowledgment of the limits.

How a Conversation Begins

An organizer — for example, an HR manager — creates a project. They formulate a core question that forms the center of the conversation. For example:

"What do our employees need to work with focus again after the reorganization?"

This core question is not a survey question for participants to answer. It's the research question that gives the AI its direction.

Participants receive a personalized link. They open it when it suits them — in the evening, between meetings, working from home. There's no appointment, no video call, no human on the other side.

The conversation starts with a simple open question:

Qerie: Thanks for taking the time. Before we get started — how are you feeling about the current situation at the company?

This isn't small talk. It's a deliberate conversation opener that creates space before moving to specific topics.

How the AI Follows Up

The core feature that distinguishes AI conversations from classic surveys: adaptive follow-up questions.

When someone answers: "I feel a bit lost right now" — the AI doesn't move to the next prepared question. It follows up:

Qerie: That sounds like a real strain. What exactly gives you that feeling — is it more uncertainty about your new role, or is it the loss of contact with your former colleagues?

This isn't a prepared path. The AI evaluates the response semantically and decides in the moment which direction makes sense.

Another example. A question about decision-making:

Participant: I often don't understand why certain decisions are made the way they are.

Qerie: Can you show me that with a specific example? What was a decision where the reasoning felt absent to you — and what would have helped?

By asking for a concrete example, the conversation leaves the abstract level. What emerges isn't "52% find decision-making processes non-transparent" — but a specific situation with context that HR can understand and act on.

What the AI Doesn't Do

Here's the honest limit.

The AI doesn't evaluate. It draws no conclusions about whether a manager is good or bad. It doesn't interpret whether someone is about to resign. It makes no recommendation about what the company should do.

The AI doesn't reliably detect irony or subtext. If someone writes "Everything's going great, as always" — and means the opposite — the AI doesn't always catch that. A human moderator would follow up. The AI does too — but less often and less precisely.

The AI doesn't conduct a therapeutic conversation. When someone shares something very personal or emotionally heavy, the AI responds matter-of-factly and steers back toward the core question. That's sometimes appropriate — and sometimes not enough.

The AI has no history with the organization. An experienced external moderator brings context. They know the industry, they've accompanied similar processes, they recognize patterns from their own experience. The AI knows only what is said in the conversation.

These aren't bugs that can be fixed — they're structural limits of the format.

What Happens After the Conversation

After the conversation — typically 15 to 25 minutes — a summary is automatically generated for the participant. They see what was captured from their responses. This creates transparency and builds trust.

For the organizer, something different happens: the AI condenses the conversation into structured insights. Not quotes that are directly attributable — but themes, perspectives, patterns.

When twenty people have participated in a project, the result isn't twenty individual evaluations that someone has to manually work through. It's a coherent picture: what's on people's minds? Where is there consensus, where are there contradictions? Which themes keep coming up?

This is the step that takes weeks in classic qualitative research and requires specialized methodological expertise. Here it happens automatically — without loss of quality.

Who This Works For — and Who It Doesn't

AI-led conversations work well when:

• Participants are comfortable communicating in writing
• The topic allows for some reflection — it's about attitudes, experiences, perspectives
• Asynchronicity is an advantage — distributed teams, different time zones, people who wouldn't speak openly in the presence of others
• Confidentiality matters — no human moderator who could attribute statements

They work less well when:

• Topics are extremely complex or emotionally charged and human empathy in the conversation is critical
• Participants have limited experience with written communication
• Culturally, a high value is placed on personal relationships in conversations

This isn't a statement that AI is better or worse than humans. It's a statement that different instruments suit different purposes.

What This Actually Changes

The question this article started with — does it actually work? — can now be answered more concretely.

An AI doesn't conduct a conversation like an experienced moderator. It doesn't conduct a conversation like a friend. It conducts a structured, open, adaptive conversation — one that generates more depth than a survey, without the effort of a human-moderated interview.

Whether that's "real" depends on what you compare it to. Compared to a questionnaire: yes, significantly deeper. Compared to an experienced qualitative researcher: no, not equivalent.

But that may be the wrong dimension of comparison. The relevant question isn't: Is AI as good as a human? It's: Does AI enable qualitative insights in situations where they wouldn't otherwise be possible?

To that, there's a clear answer: yes.

---

Qerie conducts structured AI conversations with adaptive follow-up questions — asynchronous, GDPR-native, without moderation overhead. Request early access →