Skip to main content
The Purpose is where you tell the AI Actor who it is and how it should conduct the conversation. Brief your AI Actor like you’d brief a colleague before a meeting. The Purpose Builder breaks this into four fields — Role, Goals, Flow, and Guardrails — plus optional Variables for dynamic values that change per meeting.

The Purpose Builder — define your AI Actor's persona and behavior in four fields.

​
Role

The Role defines who the AI Actor is. Give it a name, a title, and a personality. The more specific you are, the more natural the conversation will feel. A good Role answers:
  • Who is this person?
  • What company or team are they part of?
  • What is their tone — warm, professional, direct?

​
Example

Give the AI Actor a real-sounding name and title. Candidates respond better to “Alex Rivera, Recruiting Coordinator” than to a generic greeting.

​
Goals

The Goals describe what the AI Actor should accomplish during the conversation. This is the mission — what does a successful call look like? A good Goals section answers:
  • What is the purpose of this conversation?
  • What should the AI Actor learn or evaluate?
  • How should results be assessed?

​
Example

​
Flow

The Flow describes how the conversation should progress from start to finish. Rather than listing rigid questions, describe the shape of the conversation — how it opens, what ground to cover, and how it wraps up. A good Flow section answers:
  • How should the AI Actor introduce itself?
  • What topics should it cover and in what order?
  • How should the call end?

​
Example

Describe topics to explore, not a script to read. The AI Actor will find natural ways to ask about each area and follow up based on what the candidate says.

​
Guardrails

Guardrails set the boundaries for the conversation. They tell the AI Actor what to do in edge cases, what to avoid, and how to handle situations that go off track. A good Guardrails section answers:
  • How should the AI Actor handle off-topic responses?
  • What should it do if a candidate asks a question it cannot answer?
  • Are there topics it should avoid?
  • How should it manage time?

​
Example

​
Variables

Variables let you inject dynamic values into your Purpose that change per meeting. They appear as @variable_name in your Purpose text and are filled in when a meeting is created — either manually or via the API. To add a variable, click Add Variable below the Purpose fields. Each variable has:
  • Variable name — a short identifier like project_id or candidate_name. Use this in your Purpose as @variable_name.
  • Input Type — the kind of value expected. Options are:
    • Text — free-form string (default)
    • File — a file upload
    • Number — numeric value
    • Email — email address
    • Date — calendar date
    • Boolean — true / false toggle

​
Example

If your Role field says:
Create two variables: company (Text) and role_title (Text). When you schedule a meeting, you supply the actual values and the AI Actor receives the fully resolved Purpose.
Variables keep your Purpose reusable. Instead of duplicating an Actor for each client or role, create one Actor and pass the differences as variables.

​
Templates

The Purpose Builder includes templates to help you get started quickly. Select a template card, customize the fields for your use case, and you are ready to go. Available templates include:
  • Technical Hiring Interview — Screen for skills and culture fit.
  • Sales Discovery Call — Qualify on pain points and budget.
Templates are a starting point — adjust the Role, Goals, Flow, and Guardrails to match your specific needs.

​
Tips for writing a good Purpose

  • Be a persona, not a script — Describe who the AI Actor is and what it should accomplish. Let the AI find natural phrasing.
  • Set the tone early — The first 30 seconds determine how comfortable the candidate feels. Make the Role warm and welcoming.
  • Match flow to time — A 15-minute call covers 4-5 topics comfortably. A 30-minute call can go deeper. Don’t overload.
  • Be specific in guardrails — If you don’t tell the AI how to handle a situation, it will improvise. For important edge cases (salary questions, off-topic answers), be explicit.
  • Test before going live — Run a test meeting with yourself or a colleague. Listen for natural flow, clear topics, and useful insights.

​
Improving over time

After each real interview:
  1. Read the transcript. Did the conversation flow naturally?
  2. Check the insights. Were the extracted values accurate and useful?
  3. Update the Purpose and run another test.
Most teams go through 2-3 rounds before their Purpose instructions consistently produce great interviews.