The Conversation Design Workflow | AI Training
People & Blogs
Introduction
Designing effective AI assistants, whether chatbots or voice interfaces, requires a well-defined workflow that includes various phases and team roles. This article outlines the entire conversation design workflow, which can be divided into two primary sprints: Happy Conversation Design and Detailed Conversation Design.
Step 1: Requirements Gathering
The workflow begins with gathering requirements, where several key considerations come into play:
- Technical Capabilities: Assess whether the organization is capable of automating conversations.
- Identifying Conversations: Determine which conversations will be automated.
- Stakeholders Involvement: Identify who will be involved in those conversations and who will represent the organization through the AI; this is crucial for defining the bot's personality.
Here, the role of a conversational copywriter becomes essential. They maintain the integrity of the bot's tone of voice and ensure that its communication style reflects the organization's values.
Step 2: Happy Conversation Design
Once the requirements are specified, the focus shifts to "Happy Conversation Design," which centers around:
- Use Case Identification: Identify specific situations where the bot will interact with users.
- Human and Assistant Needs: Determine the needs of both users and the assistant itself using a structured canvas.
- Sample Dialog Creation: Construct sample dialogues that will be transformed into flow charts.
Flow charts are validated through a "Wizard of Oz" test and refined through an expert rewrite. This iterative process aims to improve the dialog before presenting it to stakeholders for approval. Common stakeholders typically include product owners and business managers.
This phase constitutes Sprint 1: Happy Conversation Design and involves collecting feedback and achieving sign-off from stakeholders.
Step 3: Detailed Conversation Design
Following stakeholder approval, the focus shifts to Sprint 2: Detailed Conversation Design. This sprint is more intensive and encompasses:
- AI Training: A critical phase where data is leveraged to train the AI, focusing on edge cases and the proverbial "long tail."
- Error Handling: Work to ensure that any user interactions that require human intervention are managed adequately.
- Implementation: Implement the designed dialogues and AI models, ensuring everything is live and functioning as intended.
- Iteration: After the bot goes live, continuous iteration is vital for improvement based on user interactions and feedback.
Team Roles
Various roles participate in this workflow to ensure effective design, deployment, and management of AI assistants:
AI Trainers: Focus on analyzing data and training models to create a deep understanding of the interactions. They implement everything into the AI platform and continuously identify new use cases.
Conversation Designers: Responsible for canvassing user and assistant needs, creating sample dialogues and flow charts, and conducting Wizard of Oz testing.
Conversational Copywriter: This role involves creating resonating dialogues, maintaining the bot personality, and ensuring compliance with tone of voice during both the requirements and expert rewrite phases.
Together, these roles contribute to developing AI assistants capable of managing millions of conversations daily in a friendly and personalized manner.
Keyword
- AI Assistants
- Conversation Design
- Happy Conversation Design
- Detailed Conversation Design
- Requirements Gathering
- Use Cases
- Stakeholders
- AI Training
- Error Handling
- Iteration
- Conversational Copywriter
- Flow Chart
- Sample Dialog
FAQ
Q: What is the purpose of the conversation design workflow?
A: The workflow aims to create effective AI assistants that automate conversations in a personalized and user-friendly way.
Q: What are the two main sprints in conversation design?
A: The two main sprints are Happy Conversation Design and Detailed Conversation Design.
Q: Who are the key roles involved in this workflow?
A: Key roles include AI Trainers, Conversation Designers, and Conversational Copywriters.
Q: Why is stakeholder approval important?
A: Stakeholder approval ensures that the designed dialogues and use cases align with business objectives and user needs.
Q: What happens after implementation?
A: After implementation, the bot enters a phase of iteration, allowing it to improve based on real user interactions and feedback.