Key Concepts

This guide explains the fundamental concepts that power Diaform. Understanding these will help you make the most of the platform.

Organization#

Your organization is your workspace in Diaform. It's where you manage your subscription, invite team members, and create projects.

Key characteristics:

Each organization has a unique subdomain (e.g., yourcompany.diaform.io) that appears in all conversation links
Organizations have a subscription plan (Trial, Starter, Standard, Pro, Business, or Enterprise) that determines usage allowance, projects, seats, and features
You can invite team members to collaborate on projects and view responses
All billing and usage limits are managed at the organization level
Organization roles include owner, admin, member, read-only, and billing

Your subdomain is part of your brand identity on Diaform—choose something recognizable and professional.

Project#

A project (internally called an "agent") is an AI agent configured for a specific product, service, or research topic. Each project operates independently with its own configuration.

What a project includes:

Conversation link: A unique URL where respondents access the conversation
Questions: Open text, single choice, multiple choice, NPS, or star rating prompts
Knowledge base: Background information about your product that helps the AI conduct informed conversations
Automations: Rules that trigger actions mid-conversation based on detected topics or sentiment
Appearance settings: Branding, colors, logo, and welcome messages
Project access: Optional restrictions for which organization members can open or edit the project

You can create multiple projects within one organization—for example, separate projects for different products, customer segments, or research topics.

Conversation#

A conversation is a single session between the AI and one respondent. Each conversation goes through these states:

Opened: The respondent clicked the link but hasn't started answering yet
Started: The respondent began answering questions
Completed: The conversation finished (either the respondent completed all questions or the AI determined enough information was gathered)
Cancelled: The conversation was cancelled

Each conversation generates one response with structured insights.

Usage Credits#

New organizations use AI message billing. One successful assistant response during a conversation consumes one AI message. Older organizations may still use legacy conversation billing, where one started conversation consumes one conversation credit.

You can view opened vs. started vs. completed conversations in your analytics to understand drop-off rates.

Response#

A response is the structured result of a completed conversation. Each response includes:

AI-generated summary: A concise overview of the key points from the conversation
Per-question answers: Structured answers for each question, including:
- The respondent's answer
- Sentiment analysis: Positive, negative, neutral, or mixed
- Follow-up exchanges (if any)
Answer confidence: A high, medium, or low confidence label for each summarized answer
Notable quotes: Direct respondent statements when a genuinely useful quote is present
Full transcript: The complete conversation history between the AI and respondent

Responses appear in your dashboard where you can filter, search, and analyze them.

Probing Intensity#

Probing intensity controls how deeply the AI follows up on each question. You can set this per question to balance depth vs. conversation length.

Three intensity levels:

Low: Up to 1 follow-up question—best for surface-level insights or quick surveys
Medium: Up to 3 follow-up questions—ideal for most use cases, balancing depth and respondent time
High: Up to 5 follow-up questions—for deep exploration when you need rich qualitative data

The AI adapts follow-ups based on the respondent's answers. If someone gives a detailed, complete answer, the AI may ask fewer follow-ups than the maximum allowed.

Higher probing intensity leads to richer insights but longer conversations. Use high intensity for your most critical questions.

Automations#

Automations are rules that trigger mid-conversation when the AI detects specific conditions. They enable your conversations to take action in real-time based on what respondents say.

You can create automations that trigger when:

The AI detects specific topics (e.g., mentions of "pricing", "bugs", or "customer support")
The AI detects specific sentiments (e.g., anger, frustration, churn risk, or high satisfaction)

When triggered, automations can:

Ask for an email address
Offer a discount code
Book a calendar meeting
Redirect the respondent to a URL
Send a Slack notification to your team

For example: If the AI detects pricing complaints, automatically offer a 20% discount and notify your sales team in Slack.

Context#

Your Context tab provides background information about your product, service, or company that helps the AI conduct more accurate and informed conversations.

You can add knowledge in two ways:

Text entries: Paste in product descriptions, FAQ content, feature lists, or other relevant information
File uploads: Upload PDF or DOCX files containing product documentation

The AI uses this context to:

Ask more relevant follow-up questions
Understand product-specific terminology
Provide more informed responses if respondents have questions
Better detect topics and sentiment related to your specific features

A well-stocked Context tab helps the AI conduct conversations that feel informed and personalized to your product.

What's Next?#

Now that you understand the core concepts, dive deeper into specific areas: