AI Chatbot
Writing Guidelines
Developing AI Chatbot Writing Guidelines at Amazon
As Generative AI becomes increasingly prevalent across Amazon, the need for coherent writing guidelines emerged to unify AI bots within our organisation. This is essential for ensuring consistency and establishing a distinctive tone that allows users to easily identify PEX (People Experience) products through their interactions with AI chatbots.
To address this need, I conceptualised and developed our org’s AI chatbot writing guidelines. These guidelines serve as the foundation for all content creation related to building AI chatbots within our extended org (280+ teams). They ensure a unified approach to AI personality, tone, and functionality.
What’s in the guidelines?
The comprehensive writing guidelines cover all aspects of creating AI content, including:
Tenets for the AI: Establishing core principles that define the AI's purpose, behaviour, and user interaction style.
Building an AI Chatbot's Personality: Guidelines for developing a consistent and engaging personality that resonates with users.
Naming Conventions: Standardised naming conventions to maintain clarity and cohesion across different chatbots, crucial for a seamless user experience.
UI Structure, Buttons, and Labels: Best practices for using UI elements to guide users and keep them on track during their interactions.
Assisting with Alternative Word Order and Incomplete Requests: Strategies to handle varied user inputs effectively, ensuring smooth and intuitive communication.
By creating these guidelines, I ensured that all AI chatbots within our organisation adhere to a consistent style and tone, enhancing the overall user experience. These guidelines provide a clear framework for future teams, streamlining the development process and fostering a cohesive identity for PEX products.
To ensure our teams can create AI chatbots that consistently reflect our brand's voice and values, I developed comprehensive writing guidelines. These guidelines serve as a blueprint for crafting chatbots that not only function seamlessly but also resonate with our users as authentic extensions of our organisation.
I conceptualised the user experience of our conversational AI chatbots. This involved envisioning the chatbot's personality, defining its capabilities, and understanding its limitations. By doing so, we ensured that each chatbot feels unique yet part of a cohesive family of products.
Conversation interaction styles
One key approach to scale AI chatbot design among 280+ teams was to design distinct conversation interaction styles tailored to different types of AI chat assistants.
This allowed us to create engaging, user-centric interactions that align with our brand’s ethos and provide a consistent, enjoyable user experience across all our AI-driven products.
What does this look like in practice?
Take a look —>
System-centric (Search interface)
AI interface is activated upon user input
Responses are not conversational
Responses are lists or actions
AI interface cannot reference personalised content from previous interactions
No archive
Content-centric
Longer thorough responses
User-driven responses to queries and commands
Each new user action wipes the slate clean
Visual-centric
Interface incorporates graphical elements (text format, images, buttons, etc.)
Interaction is mostly agent-driven, but involves some user input
Agent-driven simple multi-turn interaction sequences is possible
Agent might ask user if output was helpful and present button options for responses
Conversation-centric Style
Responses are “short or bite-sized” like real time conversations
Interaction is “mixed initiative,” and can be steered by both agent and user
Agent recognizes non requests like “thanks” or “what do you mean?”
Agent handles complex multi-turn interaction sequences; back and forth is natural; convo non-linear
Error handling and recovery
Misunderstandings or errors are handled gracefully
Guide users back to productive conversation without breaking the chat flow
