Experience a voice AI interview that follows the evidence.

Stage 1 has already collected fictional SNAP responses from Sam's parent and teacher. Jess will explore only the evidence threads that remain unclear, using a bounded, semi-structured conversation.

AI interviewer

Fictional scenario

10-minute prototype

This is not a diagnostic assessment.

Do not discuss a real child or provide real personal, health or school information at any point.

Before the agent begins

It should identify itself as an AI voice agent.
It should confirm that it has reached the intended participant before using your name or mentioning the fictional child.
It should reconfirm your consent to voice processing and transcript creation.
It should explain the fictional scenario and the evidence threads it will explore.

Answer naturally as Sam's teacher and invent classroom examples freely. If the agent asks about a real child, stop the session and flag it for review.

Demo launcher

Start the Stage 2 demonstration

About 10 min · 15 min maximum

Your fictional scenario

Sam Harrison, Year 4

You are role-playing as Sam's classroom teacher. Fictional parent and teacher SNAP results have already identified several questions that need clarification.

Fictional

Access code

Your first name

Used only after the agent confirms that it has reached the intended participant.

Your role

Australian mobile number

Only needed for the phone call. Leave it blank to talk in your browser instead.

Before you begin

I understand that an AI voice agent, not a person, will conduct this demonstration.

I will use the fictional scenario only and will not share information about a real child.

I consent to live voice processing and transcript creation for this demonstration.

Your number is used only to place this call and is not displayed publicly. The connected voice service processes audio and creates a transcript for prototype review. Call recording is off.

What this prototype tests

Start with what is known. Clarify only what remains uncertain.

The agent does not repeat all 26 SNAP items. It operates inside a three-layer design that focuses the conversation on evidence gaps identified before the call.

01
Fixed opening
The agent confirms identity, explains that it is AI, reinforces the fictional scenario and reconfirms consent.
02
Selective thread plan
A predetermined plan identifies the small number of evidence gaps that should be explored in this conversation.
03
Bounded clarification
The agent can ask focused follow-up questions about examples, frequency, context and functional impact.

Probabilistic information-seeking inside a deterministic safety and decision framework.

After the interview

Review the evidence, not another score.

The intended output is a structured reviewer bundle showing what the teacher reported, which questions the evidence addressed and what remains uncertain. It does not provide a diagnosis.

Threads explored

The evidence questions released for this interview

Evidence captured

Relevant examples, frequency, context and functional impact

Uncertainty made visible

Missing, ambiguous or contradictory evidence is flagged for human review

Evidence linked to transcript

The reviewer can inspect the supporting words

Important boundaries

A safe place to test the idea.

01No real child information
02No diagnosis, treatment or clinical recommendation
03No autonomous scoring or professional decision
04Invited prototype review only, never a substitute for professional assessment