Charlie is a customizable AI agent whose knowledge can be constantly updated. This is possible as you can upload, update or delete information from Charlie's  knowledge base, which gets influenced by several key factors: the information you provide, the time invested in training, and the evolution of your business. In this article we explain the logic and hierarchy that determine how Charlie's configuration and knowledge base impact response generation.

Structure

Charlie's knowledge around your business is built on 4 main pillars:

1. 📄 Training Data: Active feedback over Charlie's responses that refines its performance over time.

2. 📄 Information loaded : The foundational documents and details you provided to build Charlie's knowledge about your business or account.

3. 📄 Objections: Specific instructions on how to handle particular topics, questions, or concerns to successfully guide conversations.

4. 📄 API : Possibility to fetch live data in real-time (optional)

Priority Hierarchy

Training data holds the highest priority—anything included here will override all other instructions.

When generating a response, Charlie follows this sequence:

1. First, reviews relevant training data

2. Then, checks loaded information

3. Finally, consults objections and API information to complement the response

Logic behind it

Charlie's logic is designed to be an ever-evolving solution capable of consistently representing your brand or business. To achieve this, we prioritize feedback and up-to-date information to maintain alignment with your current reality, supported by the fundamental information you've provided.

We recognize that business dynamics change, influencing how you need to respond to certain circumstances. This is why training data receives the highest priority—it captures your most recent preferences and requirements.

Impact

Here we explain how any adjustment, new uploads or configurations in the sources of information used by Charlie, can impact its responses:

• Training data: High impact: Everything here will overwrite previous instructions

• Loaded information: Medium impact: These fundamentals guide Charlie to ensure accuracy and alignment with your brand or account.

• Objections & API: Low impact: It'll enhance responses but don't override core knowledge.

Users involved in configuration and feedback should understand these best practices, structure, logic, and hierarchy to maximize Charlie's effectiveness.

Key insights

Training Data is King Your active feedback through training data takes precedence over everything else. Use this strategically to fine-tune Charlie's responses to match your current business needs.

Build Strong Foundations The quality of your initial loaded information determines Charlie's baseline knowledge. Invest time upfront to provide comprehensive, accurate documentation.

Team Alignment Matters Everyone providing feedback or configuration should understand this hierarchy. Inconsistent input from multiple team members can create confusion in Charlie's responses