AI Copilot

AI Copilot

AI Copilot

AI Copilot is a powerful, AI-powered assistant fully integrated into your dex workspace. It's designed to accelerate every step of the analytics development life cycle (ADLC)—from writing code to debugging models—so you can deliver high-quality data products faster and more efficiently.

Copilot brings intelligent, context-aware assistance directly into your development environment, helping you write better SQL, troubleshoot issues, and automate repetitive tasks—all with natural language.

What You Can Do with Copilot

AI Copilot is built to boost productivity for data practitioners at every stage. With just a few prompts, you can:

• Generate models from scratch using natural language (e.g. “Create a model that joins customers with their last order”)

• Edit and refactor existing code without writing SQL by hand

• Fix broken models by describing the issue or pasting in an error

• Accelerate transformations by asking for column logic, join strategies, or filter conditions

• Enhance documentation with AI-suggested table and column descriptions

It’s like having a senior analytics engineer embedded directly into your modeling flow.

Where It Lives

Copilot is accessible directly from the dex code editor. Simply click the ✨ wand icon at the top of the model editor to start interacting.

You can:

• Highlight code to edit or fix it

• Use freeform prompts to generate new logic

• Ask questions about structure or transformation ideas

Data Privacy & Usage

Your privacy and data security are top priorities. Here’s how Copilot handles data:

Does Copilot store or use personal data?

No. Copilot does not require or store personal data. You control what information you enter, and Copilot operates without accessing sensitive personal inputs unless explicitly provided by you.

Does Copilot access my warehouse data?

No. Copilot does not access or query your row-level data. To provide AI assistance, it may utilize:

• Column names

• Model-level SQL

• Model names

• Inline model documentation

• Prompts entered by the user

Row-level data (e.g. actual values stored in tables) is never accessed or transmitted unless explicitly entered by the user into the Copilot interface.

Always double-check Copilot-generated code before deploying to production. Like all AI assistants, Copilot may produce incorrect or incomplete results.

Best Practices

• Use natural language prompts to get started (e.g., “Filter to the last 30 days of orders”)

• Use Copilot as a pair programmer—not a replacement for human review

• Avoid entering sensitive credentials, tokens, or row-level data in prompts

• Pair Copilot with version control to easily track and validate AI-generated edit