Use AI features responsibly

Updated

AI features can assist you as you work with Customer.io, but it’s important that you verify AI-generated responses and outputs before you implement them. This page covers some of our recommendations to help you use AI features responsibly.

Verify AI-generated results

AI suggestions and generated content are designed to help you, but they’re not a substitute for your expertise.

AI can make mistakes, potentially leading to unintended consequences. You should not rely on AI-powered features as a singular source of truth, especially when it comes to features that can impact your audience—like segments that trigger campaigns or updating messages in live campaigns.

When you work with AI-powered search results, you should review the cited sources from our documentation to verify the accuracy of the information. Checking original sources can help you verify the answers provided by AI.

You can also use the thumbs up and thumbs down buttons to tell us if an AI-generated response was helpful or not. Your feedback helps us improve our AI features.

We don’t control the AI models you use with MCP features

The tools we expose to our Model Context Protocol (MCP) server are consumed by the AI-models you use—in your Integrated Development Environments (IDEs like Cursor) or other tools. While we can test the output of individual tools, we cannot control how the LLM you use interprets the outputs of our MCP tools.

When you use our MCP server, make sure that you trust the Large Language Models (LLMs) you use. You should only use MCP features with LLMs approved by your organization.

Improve AI-generated results by accurately describing your data

When you create attributes, events, campaigns, segments, and so on, you give these assets names and descriptions. Our AI features rely on these names and descriptions to understand your data and provide suggestions—just like any member of your team would. That means that clearly labeling, describing, and organizing your data helps our AI features perform better.

For example, imagine that you have an attribute called cname. This attribute might be relevant to you. But, if you don’t provide a description, the assistant might not know what it means. Is it a company name, customer name, or does it refer to the canonical name in a domain name system? That’s something you need to tell the assistant.

Here are some best practices to help you get the most out of our AI features:

  • Use descriptive names: Give attributes, events, and campaigns clear, meaningful names
  • Add descriptions: Provide descriptions for your data attributes to help AI understand their purpose
  • Follow consistent naming conventions: Use standard formats like snake_case or camelCase consistently
  • Use tags effectively: Tag similar campaigns and segments with descriptive labels
Copied to clipboard!
  Contents
Is this page helpful?