2

Introducing Guidelines: Train Duckie to Support Customers Your Way

Valerie Li
Valerie Li
Co-founder and CEO of Duckie

What are Guidelines?

We often hear:

  • "Can Duckie follow our specific support policies?"
  • "How do I make sure Duckie uses our troubleshooting steps for certain errors?"
  • "I don’t want Duckie using certain words—can I control that?"

A one-size-fits-all AI wouldn’t work for real-world support teams. This leads to wrong answers and inconsistent responses.

You asked, we delivered. Now, you can train Duckie to respond just like your team — following your policies, handling requests exactly how you want, and continuously improving with every interaction.

How to Set Them Up

You have complete control over how Duckie learns and evolves. There are several ways to train and customize Duckie:

1. Persona – Define Duckie’s tone, response style, and level of detail

Example:

"Be very concise in your answers. Be friendly and cheerful."

2. Global Guidelines – Set rules that give Duckie instructions on the support flow, things to say and not to say, escalation policies, etc

Example:

    “Duckie has 2 versions, one is a managed service and the other is a self-hosted service. When users are asking questions, make sure to clarify with them which version of Duckie they are using, and return responses accordingly.”

3. Case-Specific Guidelines – Teach Duckie how to handle specific topics like pricing, refunds, or troubleshooting steps for certain errors

Example:

Case: Agent tool creations

Guidelines: "When people are asking about agent tool set ups, make sure to include guidance on creating API wrappers, and include code samples on constructing APIs. Default to use FastAPI for API creation unless otherwise requested by users."

4. Message-Level Feedback – You or your customers can provide feedback on Duckie’s responses, and it will learn from interactions in real-time and refine its answers.

How Your Guidelines Are Used

Your inputs are built into Duckie’s decision-making and response engines:

  1. Persona: Built into Duckie’s response engine to define its communication style.
  2. Global Guidelines: Built into both the decision and response engine, guiding response generation, escalation logic, and prioritization across all interactions.
  3. Case-Specific Guidelines & Feedback: Duckie analyzes real conversations by categorizing, clustering, and mapping feedback to specific issues. This allows it to dynamically adjust responses and refine decision-making through continuous adaptive learning.

With Guidance, Duckie adapts to your team’s workflows and policies, ensuring accurate, reliable, and on-brand support.

📢 Ready to up your support game? Book a demo with us to get started today!

Share this article: