Skip to content

Overview

Public preview

In the Data Productivity Cloud, Copilot introduces a new era of AI-driven data productivity tools. Copilot lets you prompt generative AI to create data pipelines, improving your efficiency in managing and processing data.

Copilot is still in its early stages and is considered experimental. While it offers promising features, it's continuously evolving as we gather feedback and refine its capabilities.

Key Features

  • Create data pipelines using plain language instructions, leveraging advanced AI technology.
  • Copilot accompanies users throughout their data pipeline creation journey, providing assistance at every step.

Important Considerations

  • Exclusivity: Copilot is currently available exclusively for transformation pipelines.
  • Ongoing refinement: We're actively refining Copilot's capabilities, particularly in areas such as configuring table inputs and handling variant columns containing JSON data. To mitigate limitations, consider including specific instructions and structures in your prompts.
  • Potential challenges: Users may encounter instances where Copilot incorrectly positions components on the canvas or suggests irrelevant inputs. We're committed to addressing these challenges to enhance clarity and to minimize errors in the automation process.

While Copilot represents a promising advancement in AI-driven data productivity, it's important to approach it with the understanding that it's still in its experimental phase. We appreciate your feedback and patience as we continue to improve and enhance Copilot's capabilities.

The flow of sending data to Copilot

Your interaction with Copilot involves transmitting data for processing. This flow encompasses the following steps:

  1. Instructions and prompts sent via the Copilot chat interface, steering the creation of the data pipeline.
  2. User inputs, interpreted by Copilot, to understand the desired function and data flow of the pipeline.
  3. Copilot uses user input and its knowledge of common data transformations to select components for filtering, aggregation, and any other operations required in the data pipeline.
  4. Copilot generates metadata relative to the transformation columns, which includes specifics such as column names, data types, sizes, and precision levels. This metadata is derived from the user's instructions.
  5. Users retain the ability to customize the data and make changes in the output as required.
  6. Users review and validate the created pipeline and its components. If needed, the user should modify individual component properties to ensure the pipeline delivers the desired output.

Before you start using Copilot, read Copilot prerequisites.