Auto Pilot is an intelligent feature that automatically selects the most appropriate AI model or function for your prompts, ensuring faster, more accurate, and contextually relevant responses—without requiring manual selection. It is designed to simplify the user experience by reducing the effort involved in choosing the right tools, improving overall response quality, and leveraging memory capabilities so users can focus more on outcomes rather than configuration.

Note: Auto Pilot is available to organizations that have subscribed to it through the Marketplace. This feature is currently in Beta, and some capabilities may be experimental and will continue to evolve over time.

How Auto Pilot Works

Auto Pilot intelligently balances relevance, accuracy, and performance when determining how to handle your request. Its process includes:

  1. Request Classification – Determines whether your input should be handled by an AI model (for reasoning or text generation) or a function (for task-specific operations such as calculations or data retrieval).
  2. Context Awareness – Analyzes conversation history, uploaded files, and—if enabled—CSE memories or data sources to fully understand intent.
  3. Model Selection – Chooses the most suitable model based on task complexity, speed, and quality requirements.
  4. Function Matching – Identifies if a specialized function is better suited for the task, such as handling real-time data, computations, or structured outputs.
  5. Policy & Safety Checks – Ensures all actions comply with privacy, security, and platform policies.
  6. Execution & Response – Enhances the prompt if needed, executes it using the selected model or function, and returns the result with clear visibility into what was used.

This intelligent routing ensures every request is handled by the most effective resource, improving both efficiency and accuracy while maintaining compliance.

Prompt Enhancement

Auto Pilot improves prompt clarity by refining user inputs to produce better results. This is especially helpful for short or ambiguous queries, where additional structure can significantly improve output quality. For example, a simple request like “summarize sales deck” may be enhanced with more context, expected output format, and level of detail to generate a more precise and useful response.

Reasoning Transparency

Auto Pilot provides visibility into how responses are generated by displaying a reasoning trace. This allows users to understand the steps taken to arrive at an answer, improving trust, validation, and interpretability—especially for complex or multi-step tasks.

Memories Integration

For users of the Cognitive Storage Engine (CSE), Auto Pilot can incorporate stored memories—such as preferences, past interactions, or contextual insights—into responses. This enables more personalized and context-aware outputs. When memories are used, they are surfaced within the chat, giving users full transparency into what influenced the response.

Advanced File Search

Advanced File Search allows Auto Pilot to locate and retrieve relevant files directly from your CSE using natural language queries. Users can request documents conversationally, and the system will return matching results with key metadata such as file name, date, size, and tags. Access is strictly controlled, ensuring that only authorized files are visible.

Best Practices