How to Use Copilot to Suggest Integration Setup

This guide explains how to use the AI-powered Copilot feature to automatically generate a complete integration configuration from a sample file.

What is Copilot Auto Setup?

Copilot Auto Setup uses artificial intelligence to analyze an uploaded file and suggest a complete integration setup. Instead of manually configuring fields, mappings, and value translations, you upload a file, optionally describe its purpose, and Copilot generates the entire configuration for you to review and confirm.

Copilot supports all file types: Excel, CSV, Text, and JSON.

Prerequisites

Before using Copilot Auto Setup, ensure:

  1. The Integration Framework Auto Setup capability is enabled in Copilot & AI Capabilities in Business Central
  2. You have the NAVX IF ALL permission set assigned
  3. You have a sample file ready for upload

How Copilot Auto Setup Works

User uploads a file and describes it
    |
Copilot analyzes file structure
    |
AI matches columns to BC tables and fields
    |
Copilot generates fields, mappings, and value maps
    |
Post-processor adds missing required fields
    |
User reviews and edits suggestions
    |
User confirms to create the integration

Step-by-Step Guide

Step 1: Open the Copilot Auto Setup

On the Integrations page, choose the Suggest Integration Setup action in the prompting area. This action has a sparkle icon indicating it is an AI-powered feature.

Step 2: Upload a File

In the Copilot dialog, choose the Attach File action to upload a sample file. Supported file types are:

  • Excel files (.xlsx, .xls)
  • CSV files (.csv)
  • Text files (.txt, .dat)
  • JSON files (.json)

The uploaded file name will appear in the dialog once selected.

In the Describe the file field, enter a description of what the file contains. This helps Copilot make better suggestions. Examples:

  • "Customer master data with customer numbers, names, and addresses"
  • "Sales orders mapping to Sales Header and Sales Line"
  • "Vendor invoices with line items"
  • "Inventory items with quantities and locations"

The more specific your description, the better the AI can match your data to the correct Business Central tables and fields.

Step 4: Generate Suggestions

Choose the Generate action. Copilot will:

  1. Analyze the file - Detect file type, delimiters, headers, column data types, and sample values
  2. Build BC metadata context - Gather relevant Business Central table and field information (up to 25 tables with up to 50 fields each)
  3. Send to AI - Submit the file analysis and BC metadata to the AI engine
  4. Parse the response - Convert the AI suggestions into integration field and mapping records
  5. Post-process - Add any missing required fields for document imports (e.g., Document Type, Line No.)

Step 5: Review the Suggestion Summary

After generation, the dialog displays a Suggestion Summary with:

Field Description
Integration Name The suggested name for the integration. You can edit this.
Description The suggested description. You can edit this.
Integration Type The detected file type (Excel, CSV, Text, or JSON). Read-only.
Field Delimiter The detected field delimiter (e.g., Comma, Tab, Pipe). Visible for non-JSON files.
Record Delimiter The detected record delimiter (e.g., LF, CR, CRLF). Visible for non-JSON files.
JSON Root Path The dot-notation path to the root array. Visible for JSON files. You can edit this.
Root Array Count The number of elements in the root JSON array. Visible for JSON files.
Column Count The number of columns detected in the file. Visible for non-JSON files.

Step 6: Review Suggested Fields

The Suggested Fields section shows the integration fields that Copilot suggests creating. Each row represents a source column from your file. You can review and edit:

  • Field number and name
  • Column type (Standard, Constant, Dynamic)
  • Constant values for fields that should always have the same value
  • JSON Path for JSON files

For document imports (e.g., Sales Orders), the post-processor automatically adds required fields like Document Type (as a Constant), Type (as a Constant set to "Item"), Line No. (as Line Numbers), and System-Created Entry (as a Constant).

Step 7: Review Suggested Mappings

The Suggested Mappings section shows how each field maps to a Business Central table and field. You can review and edit:

  • Source field number and name
  • Destination table and field
  • Whether validation is enabled
  • Whether the field creates a new document

Step 8: Review Warnings

If Copilot detected any issues, they appear in the Warnings section. Common warnings include:

  • Invalid table or field references that were skipped
  • Missing critical mappings (customer/vendor number, item number, quantity)
  • Fields that could not be matched to any BC table

Step 9: Confirm or Discard

  • Choose Confirm to create the integration with all suggested fields, mappings, value maps, and character replacements
  • Choose Discard to close without creating anything

After confirmation, the new integration appears on the Integrations page and is ready for use.

Step 10: Regenerate (Optional)

If the suggestions are not satisfactory, choose the Regenerate action to re-analyze the file and generate new suggestions without closing the dialog. You can modify your description before regenerating to guide the AI toward better results.

Copilot with JSON Files

When you upload a JSON file, Copilot provides additional capabilities:

  • Automatic root array detection - Copilot detects the path to the root data array
  • Dot-notation field paths - Fields are mapped using JSON paths like customer.address.city
  • Array mode detection - Nested arrays are identified with appropriate flatten/concatenate modes
  • JSON Sample storage - The uploaded file is stored as a JSON Sample on the integration record for future field discovery

What Copilot Creates

When you confirm the suggestions, Copilot creates:

  1. Integration record - With name, description, type, direction, delimiters, and JSON paths
  2. Integration fields - One per source column, with appropriate column types and constant values
  3. Integration mappings - Linking each field to a BC destination table and field
  4. Field value maps - Any value translation rules suggested by the AI
  5. Character replacements - Any find-and-replace rules for data cleaning

Tips for Better Results

  • Be specific in your description - Mention the Business Central tables you want to target (e.g., "Sales Header and Sales Line" rather than just "sales orders")
  • Use meaningful column headers - Files with descriptive headers produce better results
  • Include representative data - Ensure your sample file contains realistic data values
  • Review mappings carefully - While Copilot is accurate, always verify the destination tables and fields match your requirements
  • Iterate if needed - Use the Regenerate button with a modified description if the first suggestion misses the mark

Business-Critical Field Warnings

For document imports (Sales Header/Line, Purchase Header/Line), Copilot validates that critical fields are mapped:

  • Customer/Vendor Number - Required for document creation
  • Item Number - Required for line items
  • Quantity - Required for line items

If these fields are not mapped, Copilot surfaces a warning so you can address the gap before confirming.

Troubleshooting

"Enable the Integration Framework Auto Setup capability" Error

The Copilot capability has not been enabled. Go to Copilot & AI Capabilities in Business Central and enable Integration Framework Auto Setup.

"Please upload a file before generating" Error

You must attach a file before choosing Generate. Use the Attach File action first.

"The AI could not determine appropriate mappings"

The AI could not match your file columns to BC fields. Try providing a more detailed description mentioning specific BC tables and fields.

"The AI response could not be interpreted"

The AI returned an unexpected response. Try again, or modify your description to be more specific.

Few or No Mappings Suggested

If Copilot suggests fields but no mappings, your column names may not match any known BC fields. Add a description like "Map to Customer table fields Name, Address, City, Post Code" to guide the AI.

See Also