Nexadata Pipelines are end-to-end workflows for transforming data through a series of organized steps. With the Nexadata Transformation & Mapping Copilots, you can create pipelines simply by describing your transformation objectives in plain language.
You provide:
Input data: structured application data from connected systems such as Salesforce or Anaplan, tabular files such as CSVs and spreadsheets, or tabular data extracted from unstructured sources such as PDFs
A structured prompt describing your desired logic
An optional output goal file to guide the final format
From there, Copilot builds the pipeline step-by-step using a deterministic, agentic AI approach. Each transformation is logically derived, reproducible, and fully visible.
Nexadata Transformation & Mapping Copilots in action
👍 Advanced users can continue to use Advanced Mode to manually define or refine transformations with full control.
Step-by-step Instructions
Step 1: Prepare Your Data
Before starting:
If multiple datasets need to be joined, they must have a common join key, such as a unique ID or email address.
(Optional) Upload a sample output file to define the structure of your desired result.
These three inputs (input data, structured prompt, and output goal) provide Nexadata AI Copilot with the context, intent, and target needed to build a pipeline.
Step 2: Build Your Pipeline with the Copilot
Nexadata Copilots use agentic AI intelligence to construct pipelines based on your transformation and mapping goals. Every pipeline is generated deterministically, meaning:
Each transformation is derived step-by-step
The same inputs will always produce the same outputs
Every action is explainable, auditable, and transparent
What You Provide
Input data: structured application data, tabular files, or data extracted from unstructured sources such as PDFs
Structured prompt: a plain-language description of your logic
(Optional) Output goal file: a CSV showing the target format
The Copilot evaluates your inputs, applies each transformation in a logical sequence, and presents you with a full, deterministic pipeline ready for review.
The Pipeline AI Copilot setup form showing the pipeline name, dataset selector, prompt field, and optional sample output upload
✏️ Example Prompts
Example 1
Convert all values in the "status" column to uppercase to ensure consistency. Convert the "last_name" column to lowercase for standardization purposes. Keep only the rows where status equals "ACTIVE" (after the uppercase transformation). Remove the "overtime_hours," "hours_worked," and "projects_completed" columns. Rename "performance_score" to "rating". Group the data by "region" and calculate average salary. Sort the grouped results by average salary in descending order.
Example 2
Build a segment hierarchy based on total order volume per customer. Customers with more than 20,000 in total sales should be classified as 'Strategic', those from the USA as 'Domestic', and all others as 'International'.
Each of these prompts leads to a deterministic sequence of transformations, ensuring full traceability from input to output.
Step 3: Review and Refine Your Pipeline (Human-in-the-loop)
Once Copilot has built your pipeline:
Review each transformation step in order.
Each step is deterministically generated based on your prompt and the output from the previous step.
A completed pipeline showing the full list of transformation steps generated by Copilot
For more control, you can select and configure individual transformations directly through the UI. Advanced Mode lets you precisely define each step by manually configuring the transformation logic. It is always accessible regardless of how a transformation was originally created, giving you maximum flexibility to adjust your pipeline at any time. Below is an example of configuring a filter transformation in Advanced Mode.
Advanced Mode showing a filter transformation being configured manually
Step 4: Insert a Transformation Mid-Pipeline
You can add a new transformation anywhere in an existing pipeline without rebuilding it. To insert a step between two existing transformations, hover your mouse between the two steps in the Transformations panel. A + button will appear at that position.
Hovering between two pipeline steps to reveal the + insertion button
Click + to open the transformation type selector. Three options are available:
Transform: Add a transformation using natural language mode. Describe what you want to do, and the Copilot will configure the step.
Mapping: Map values based on conditions using the Mapping Copilot.
Advanced: Build the transformation manually with full control over the configuration.
The transformation type selector showing Transform, Mapping, and Advanced options
Natural Language Mode (Step-Level Prompting)
When you add individual transformations using natural language prompts, you can do this without running the full Copilot.
Steps created this way are marked with a ⚡lightning bolt icon.
Hover over the icon to view the original prompt used for that transformation.
This mode works alongside Copilot and Advanced Mode, offering flexible ways to build or edit pipelines one step at a time.
👉 For more details on the difference between natural language mode and Advanced Mode, see Natural Language Input vs. Advanced Mode in Nexadata Pipelines.
Supported Transformations
Whether you build your pipeline using Copilot or Advanced Mode, Nexadata supports a wide range of transformations:
Copy: Duplicate a column and specify its position in the dataset.
Filter: Include or exclude rows based on custom conditions.
Flip sign: Automatically adjust the sign of values based on custom conditions.
Group by: Group and summarize data with functions like sum, count, and average.
Insert column: Add new columns with formulas or default values.
Insert row numbers: Add sequential row numbers to your dataset using the Insert Row Numbers transformation.
Join: Combine datasets in Nexadata using a common key, applying graphical joins (inner, outer, left, right) to produce a custom, unified output dataset.
Keep: Selectively retain essential columns while removing all others to streamline your dataset.
Lowercase: Standardize text by converting selected columns to lowercase.
Math Operations: Use the Math transformation to perform statistical operations such as sum, mean, median, and standard deviation on selected columns.
Merge: Combine selected columns into a single column with a custom delimiter.
Parse Date: Convert date values from any format into a standardized datetime format.
Remove: Remove unnecessary columns to simplify your dataset for improved focus and clarity.
Rename: Update column names, enhancing clarity and consistency across your dataset.
Replace: Find and replace values in a column, with options for exact matches or regex patterns.
Shift: Reposition columns to the start, end, or relative to another column.
Sort: Arrange dataset columns in ascending or descending order, with options for empty value placement.
Split: Divide a column’s contents into multiple columns based on a specified delimiter.
Stack Rows: Combine rows from two or more datasets where the columns and column types match.
Uppercase: Standardize text by converting selected columns to uppercase.
💡 Copilot intelligently selects the right transformations based on your prompt. In Advanced Mode, you can manually apply and sequence any of these transformations.
Summary of Steps
Connect your input data (application data, tabular files, or data extracted from unstructured sources such as PDFs)
Provide a structured prompt describing your transformation logic
(Optional) Upload a sample output goal file
Let AI Copilot build a deterministic pipeline
Review or refine steps using Advanced Mode or by inserting transformations mid-pipeline
Execute the associated Workflow