How the Brand Assistant works
The Brand Assistant chat gives you access to brand knowledge through a conversational chat window. Using natural language prompts, you can generate content that aligns to a specific brand kit, retrieve answers to brand questions, and look up brand guidelines.
The following sections describe how the Brand Assistant generates responses.
Generating a response for brand-aware chats
After you submit a prompt for brand-aware chats, the starts a thinking process to decide whether to answer directly or involve additional agents.
Here's how it works:
-
The user enters a prompt - the user enters a prompt in the - such as a question or an instruction.
-
The system passes information from the Brand Context - the system automatically provides information from the Brand Context brand kit section as a system prompt.
-
checks if it can answer from Brand Context - the thinking evaluates whether the information passed from the Brand Context alone is enough to answer the prompt.
-
Path selection - based on the analysis, the process continues in one of two ways:
-
Generate a direct response - if Brand Context is sufficient, the generates a direct response using that content.
-
Invoke other agents - if the Brand Context doesn't answer the prompt and more information is needed, the activates one or more agents to search and organize the information, then generate a response. For example:
-
Search agent - finds information from your brand knowledge, the web, or both.
-
Summary agent - condenses retrieved information into a concise, relevant response.
-
-
Generating a response for brainstorming requests
When you enable the Brainstorming and submit a prompt, the begins its thinking process and runs a chain-of-thought workflow with a sequence of agents. Each agent completes a specific task to produce a comprehensive response.
Here's how it works:
-
The user enters a prompt - the user enters a prompt in the - such as a question or an instruction.
-
The system passes information from brand kit sections - the system automatically provides information from specific sections of the brand kit - Do's and Don'ts, Tone of Voice, Grammar Guidelines, and Checklist as a system prompt. This ensures all generated content reflects these standards.
-
The user enables Brainstorming - when Brainstorming is enabled and the user sends a prompt, the begins its thinking process and activates the agents involved.
-
Agent sequence - the sequence of agents is activated:
-
Create an outline - the first agent reviews the user's prompt and creates a structured content outline.
-
Generate questions - the next agent lists questions needed to complete the content.
-
Search for answers - another agent searches for answers to the list of questions. It searches information from brand knowledge, the web, or both, depending on the user's selected search settings.
-
Review coverage - the next agent checks if the answers fully address the prompt. If gaps remain, not, the process can loop up to three times to refine the outline, questions, and retrieval.
-
-
Generate a response - when the information is complete, the returns a final, comprehensive response with a summary and a standalone, editable content block (an artifact).