Deliverables

Deliverables are specific pieces of information that the agent collects from users during tasks. They define what data to extract, how to validate it, and whether it's required.

Deliverable Properties

Property	Type	Required	Default	Description
key	string	Yes	—	Variable name (snake_case recommended)
type	string	No	"string"	Data type: string, number, boolean, enum
description	string	No	""	What information to collect
required	boolean	No	true	Must be collected
acceptance_criteria	string	No	""	Validation rules
examples	array	No	[]	Example values
enum_values	array	No	null	Valid options for enum type

Deliverable Types

string

Free-form text input. The most common type.

{
  "key": "full_name",
  "type": "string",
  "description": "User's full legal name",
  "required": true,
  "acceptance_criteria": "Must include first and last name",
  "examples": ["John Smith", "Maria Garcia-Lopez"]
}

number

Numeric values (integers or decimals).

{
  "key": "age",
  "type": "number",
  "description": "User's age in years",
  "required": true,
  "acceptance_criteria": "Must be between 18 and 120",
  "examples": ["25", "42", "67"]
}

{
  "key": "weight_kg",
  "type": "number",
  "description": "Weight in kilograms",
  "required": false,
  "acceptance_criteria": "Positive number, can include decimals",
  "examples": ["70.5", "85", "62.3"]
}

boolean

True/false values. Useful for yes/no questions.

{
  "key": "has_insurance",
  "type": "boolean",
  "description": "Whether the user has health insurance",
  "required": true,
  "examples": ["true", "false"]
}

{
  "key": "agrees_to_terms",
  "type": "boolean",
  "description": "User agreement to terms and conditions",
  "required": true,
  "acceptance_criteria": "Must be true to proceed"
}

enum

Selection from predefined options. Requires enum_values array.

{
  "key": "preferred_time",
  "type": "enum",
  "description": "Preferred appointment time",
  "required": true,
  "enum_values": ["morning", "afternoon", "evening"],
  "examples": ["morning"]
}

{
  "key": "urgency_level",
  "type": "enum",
  "description": "How urgent is this request",
  "required": true,
  "enum_values": ["low", "medium", "high", "emergency"],
  "acceptance_criteria": "Agent should help user determine appropriate level"
}

Status Tracking

During conversation execution, each deliverable has a status:

Status	Description
pending	Not yet collected
partial	Partially collected (awaiting clarification or validation)
completed	Successfully collected and validated
skipped	Not collected — either an optional deliverable the user declined, or any deliverable (required or optional) whose owning task was skipped

Status Flow

┌─────────────┐
│   pending   │  Initial state
└──────┬──────┘
       │ User provides response
       ▼
┌─────────────┐
│   partial   │  Response needs clarification
└──────┬──────┘
       │ User provides complete response
       ▼
┌─────────────┐
│  completed  │  Successfully collected
└─────────────┘

Or, when not collected:

┌─────────────┐
│   pending   │
└──────┬──────┘
       │ User declines (optional), or the owning task is skipped
       ▼
┌─────────────┐
│   skipped   │  Not collected
└─────────────┘

Property Deep Dive

key

The identifier used to store and reference the collected value. Use descriptive snake_case names.

{"key": "user_email"}
{"key": "appointment_date"}
{"key": "pain_level_1_to_10"}

Best practices:

• Use snake_case: first_name not firstName or FirstName
• Be descriptive: preferred_contact_method not pcm
• Avoid generic names: user_age not data1

description

Human-readable description of what to collect. Helps the agent understand context.

{
  "key": "symptoms",
  "description": "Primary symptoms the patient is experiencing, including duration and severity"
}

acceptance_criteria

Validation rules the collected value should meet. The agent uses this to determine if a response is acceptable.

{
  "key": "phone_number",
  "acceptance_criteria": "Must be a valid US phone number with 10 digits, can include area code"
}

{
  "key": "date_of_birth",
  "acceptance_criteria": "Must be a valid date in the past, format MM/DD/YYYY or natural language"
}

examples

Sample values that guide the agent in understanding the expected format.

{
  "key": "email",
  "examples": ["user@example.com", "john.doe@company.org"]
}

{
  "key": "appointment_preference",
  "type": "enum",
  "enum_values": ["in-person", "video", "phone"],
  "examples": ["video"]
}

Complete Examples

Contact Information Deliverables

{
  "deliverables": [
    {
      "key": "full_name",
      "type": "string",
      "description": "User's full name",
      "required": true,
      "acceptance_criteria": "Must include first and last name",
      "examples": ["Jane Doe", "John Smith Jr."]
    },
    {
      "key": "email",
      "type": "string",
      "description": "Email address for confirmations",
      "required": true,
      "acceptance_criteria": "Valid email format",
      "examples": ["jane@example.com"]
    },
    {
      "key": "phone",
      "type": "string",
      "description": "Phone number",
      "required": false,
      "acceptance_criteria": "10+ digits with optional country code",
      "examples": ["+1-555-123-4567"]
    }
  ]
}

Health Screening Deliverables

{
  "deliverables": [
    {
      "key": "has_symptoms",
      "type": "boolean",
      "description": "Whether the patient has any symptoms",
      "required": true
    },
    {
      "key": "symptom_description",
      "type": "string",
      "description": "Description of symptoms",
      "required": false,
      "acceptance_criteria": "Only required if has_symptoms is true"
    },
    {
      "key": "pain_level",
      "type": "number",
      "description": "Pain level on scale of 1-10",
      "required": false,
      "acceptance_criteria": "Integer between 1 and 10",
      "examples": ["3", "7", "10"]
    },
    {
      "key": "urgency",
      "type": "enum",
      "description": "How urgent is the care needed",
      "required": true,
      "enum_values": ["routine", "soon", "urgent", "emergency"]
    }
  ]
}

Appointment Booking Deliverables

{
  "deliverables": [
    {
      "key": "preferred_date",
      "type": "string",
      "description": "Preferred appointment date",
      "required": true,
      "acceptance_criteria": "Valid date in the future",
      "examples": ["next Monday", "January 15th", "2024-01-15"]
    },
    {
      "key": "time_preference",
      "type": "enum",
      "description": "Preferred time of day",
      "required": true,
      "enum_values": ["early-morning", "morning", "afternoon", "evening"]
    },
    {
      "key": "visit_type",
      "type": "enum",
      "description": "Type of appointment",
      "required": true,
      "enum_values": ["in-person", "video", "phone"]
    },
    {
      "key": "special_requirements",
      "type": "string",
      "description": "Any accessibility or special requirements",
      "required": false,
      "examples": ["wheelchair accessible", "interpreter needed"]
    }
  ]
}

Best Practices

Naming Conventions

Good	Avoid
user_email	email1, e
appointment_date	date, appt
pain_level_1_to_10	pain, level

Acceptance Criteria

Write criteria that help the agent validate responses:

// Good - specific and actionable
{
  "acceptance_criteria": "Must be a valid US zip code (5 digits or 5+4 format)"
}

// Avoid - vague
{
  "acceptance_criteria": "Valid format"
}

Examples

Provide diverse examples that cover edge cases:

// Good - shows variety
{
  "examples": ["John Smith", "Maria Garcia-Lopez", "J. Robert Oppenheimer III"]
}

// Avoid - repetitive
{
  "examples": ["John Smith", "Jane Smith"]
}

Next Steps

• Examples — See complete plan examples with deliverables in context