Overview
Bolt Agents are designed to provide users with accurate and relevant answers based on their context and assigned knowledge. The key to maximizing their effectiveness lies in careful planning—balancing specialized agents with generalist ones to ensure smooth, accurate responses without overwhelming users.
In this guide, we’ll walk you through how agents work together and provide steps to help you structure your agents in a way that delivers a great user experience.
Things to Consider
Generative AI Model: Bolt Agents use AI to interpret questions and respond based on their assigned knowledge. Unlike decision-tree systems, they don’t rely on rigid rules but instead interpret the intent behind each question.
Specialization + Teams: You can assign specific skills and knowledge to agents (e.g., program-specific or general financial aid inquiries) and group them into teams to ensure collaboration.
Planning Matters: A well-planned agent structure helps avoid confusing or irrelevant responses, keeping the user experience streamlined and efficient.
Steps to Ensure Effective Planning of Your Agent
Identify Key Use Cases
Start by defining what types of questions students are most likely to ask and in what context.
For general areas of your website, like admissions or campus information, consider using a generalist agent who can handle a wide variety of questions.
For more specific sections, like academic programs (e.g., MBA or nursing), deploy specialized agent that focus on answering questions related to those areas.
Determine Where Specialization is Needed
Use specialized agents only where necessary. For instance, in the MBA section of your site, a dedicated MBA admissions agent would provide highly relevant answers.
In contrast, on a general admissions page, you wouldn’t need an agent for every individual program—one or two generalist agents should cover the most common inquiries.
Limit Overlap in General Contexts
Avoid deploying too many specialized agents in broad sections of your site. For example, if multiple program-specific agents are available in a general area, they may return confusing or irrelevant information.
Keeping the number of agents minimal in general contexts helps users get consistent and accurate answers without confusion.
Use Internal Descriptions for Better Collaboration
Ensure agents can work together efficiently by using the internal description feature. This allows one agent to direct the user to another within the same team if a question falls outside its expertise. For example, if the admissions agent cannot answer a detailed financial aid question, the internal description can guide it in passing the query to a financial aid agent.
Test the User Experience
Once your agents are deployed, test them by asking different types of questions. Make sure that specialized agents provide answers relevant to their focus and that generalists cover broader topics without unnecessary redirection.
Adjust the setup if you notice agents returning inconsistent answers or directing users unnecessarily.
Balance Generalist and Specialist Agents
In highly specific sections (e.g., MBA or medical programs), ensure your agents are tightly focused.
In more general sections, simplify the setup by using fewer generalist agents that can handle a variety of common questions.
Less is more in general contexts—too many agents can overwhelm users or lead to conflicting answers.
Additional Pro Tips
Keep the student's experience in mind. Always plan with their perspective in mind. The goal is to deliver the information they need quickly and accurately without unnecessarily sending them through multiple agents.
Don't be afraid to iterate and adjust. As you gather more insights into how users interact with your agents, continue refining their structure for improved performance.