An AI Actor is your AI-powered interviewer. This guide walks you through every step of setting one up.
Open the AI Actors page
- Log in to vocalis.witting.ai.
- Click AI Actors in the left sidebar.
- Click Create Actor.
The setup wizard has four steps: Actor → Purpose → Insights → Reports.
Step 1: Actor — the basics
This step defines how your AI Actor presents itself and how it sounds.
Name
A display name for your team’s reference. This is what you see in the dashboard — candidates do not see it.
Examples: “Phone Screen - Backend Engineer”, “Culture Fit - Marketing”, “Reference Check - General”
Voice and style
Choose how the AI Actor speaks:
| Setting | What it controls |
|---|
| Voice | The voice model (each has a distinct tone and feel) |
| Language | What language the AI speaks in |
| Tone | Professional, casual, warm, etc. |
| Speaking speed | How quickly the AI talks |
Test a few voice options to find one that fits the interview style. A warm, mid-speed voice works well for most screening calls.
Step 2: Purpose — the instructions
The Purpose is the heart of your AI Actor. It tells the AI exactly what to do during the interview — what questions to ask, how to behave, and what to evaluate.
A good Purpose includes:
- Context — What role is being interviewed for, what stage of the process this is.
- Questions — The specific questions to ask, in order.
- Behavior guidelines — How to handle follow-ups, off-topic responses, or unexpected situations.
- Evaluation criteria — What the AI should assess and how to rate it.
Example Purpose
You are conducting a 15-minute phone screen for a Senior Backend Engineer
position at Acme Corp.
## Context
This is the first-round screening call. The candidate has applied through
our careers page. Your goal is to assess basic qualification and interest.
## Questions
Ask these questions in order:
1. Can you walk me through your current role and what you work on day to day?
2. What backend technologies do you have the most production experience with?
3. Tell me about a challenging technical problem you solved recently.
4. Why are you interested in joining Acme?
5. What are your compensation expectations?
6. When would you be available to start?
## Behavior
- Be warm and conversational, not robotic.
- If a candidate gives a one-sentence answer, ask one follow-up to get more depth.
- If a candidate goes significantly off topic, gently redirect.
- Keep the total call under 15 minutes.
## Evaluation
After the call, rate the candidate on:
- Technical depth (1-5)
- Communication clarity (1-5)
- Role fit (1-5)
- Overall recommendation: Strong Yes / Yes / Maybe / No
Insights are the structured data points that Vocalis pulls out of each conversation. After the interview, these fields get populated based on what the candidate said.
Adding insights
Click Add Insight and configure:
| Property | What it means |
|---|
| Name | Label for the data point (e.g., “Years of experience”) |
| Type | Text, Score, Bar Chart, Line Chart, or Pie Chart |
| Purpose | A description to help the AI extract the right value |
Example insights for a phone screen
| Insight name | Type | Purpose |
|---|
| Current role | Text | Candidate’s current job title and company |
| Primary technologies | Text | Main languages and frameworks they use |
| Years of backend experience | Score | Total years of backend development |
| Salary expectations | Text | Stated compensation expectations |
| Available start date | Text | When they can start |
| Overall recommendation | Text | Strong Yes / Yes / Maybe / No |
Add a clear purpose description to each insight — it helps the AI extract the right information. “Stated compensation expectations including base and equity” is better than just “Salary”.
Step 4: Reports — how results are packaged
Configure how the interview results are summarized and presented to your team. Review the settings and click Save to create your AI Actor.
Test your AI Actor
Before sending your AI Actor to real candidates:
- Create a test meeting with yourself as the participant.
- Go through the interview as if you were a candidate.
- Review the transcript and insights — do they match what you’d expect?
- Adjust the Purpose and Insights based on the test results.
Most teams go through 2-3 test rounds before going live.
Next steps
