Agent Performance Improvement

ADS automatically enhances AI agent performance through intelligent prompt engineering and context enrichment.

This happens without any additional configuration required and uses the same LLM instance provided by you for your ADS Subscriber.

How It Works

ADS uses these key data points to optimize your agent’s performance:

• Agent Description is provided during subscriber instantiation and contains detailed information about the agent’s purpose, capabilities, and domain expertise.

  It is suggested that you provide a detailed description about the tools and abilities your agent has as that will help the agent plan the tasks better according to the kind of ADS event it receives.

• Event Context comes from the publisher payload and includes the:

  • event_name (e.g., subscription_cancelled, website_crashed)
  • event_description with comprehensive details about what the event means and its implications
  • event_data which contains any additional information relevant to the event

Prompt Enhancement feature combines these elements into a seasoned prompt that provides rich context to guide your AI agent’s decision-making process.

The Enhancement Process

When an event triggers an agent, ADS automatically:

1. Contextualizes the Agent: Uses the agent description to establish the agent’s role and capabilities
2. Explains the Event: Incorporates the event name, description and metadata to provide complete situational awareness
3. Generates Enhanced Prompt: Wraps the original prompt with contextual information for better agent performance
4. Captures Execution Result: Instructs the agent to provide detailed summary of its reasoning and actions it performed
5. Sends Execution Report (If notification channels are configured): Sends a comprehensive report of the agent’s actions and reasoning to configured channels like Slack, email, etc.

You can learn more about this in the ADS Subscriber Reports documentation.