Create the AB-620 Dataverse tables manually

Use this fallback only if you cannot import the additive solution in Lab 2, Exercise 0. Creating all four tables by hand takes substantially longer than importing the solution. Keep the agent and Customer Support Rep Assistant solution you made in Lab 1; don't create a second agent or change its publisher.

Before you start

Open the Customer Support Rep Assistant solution and confirm its publisher is Support with prefix sup. Make it your preferred solution. You need permission to create Dataverse tables in the environment. Use the exact table and column display names below so the later workflow steps can find the columns.

Create each table in the solution

  1. In Copilot Studio, select the ellipsis (...) next to your account name, select Solutions, and open Customer Support Rep Assistant.
  2. Select + New > Table > Tables, then Start from blank.
  3. On the table card, select ... > Properties. Enter the table's display name and description from the sections below. Keep the generated plural name, then select Save.
  4. On the table card, select ... > View data. Expand New column, select Edit column, set its display name to the specified primary column, and save it as text.
  5. Use + New column for the remaining columns. Leave Required off unless the interface requires a value. For single-line text, set Format to Email or Text as noted. For a choice, add the labels shown, keep the generated numeric values, leave multiple selection off, and set the default to None.
  6. After saving the columns, return to the agent solution. Confirm the table is listed. If it isn't, use + Add existing > Table and select it. Repeat for each of the four tables.

Check the column Schema name under Advanced options before saving. It must use the sup_ prefix. Lab 2's return workflow and refund filter reference the Authorization number column. If a column has a different prefix or a structurally different name, correct it before proceeding.

Returns

Display name: Returns. Description: Return authorizations created by the Customer Support Rep Assistant. Primary column: Authorization number (text).

Additional column Data type Settings
Customer email Single line of text Email format
Order ID Single line of text Text format
Reason Single line of text Text format; maximum 500 characters
Condition Single line of text Text format; maximum 200 characters
Expected refund Whole number No additional settings
Status Choice Authorized, Approved, Rejected, Escalated
Approved by Single line of text Email format

The Returns table starts empty. Do not import a schema header file as records.

Orders

Display name: Orders. Description: Customer orders. One row per order. Primary column: Order ID (text).

Additional column Data type Settings
Customer email Single line of text Email format
Customer name Single line of text Text format
Order date Date only No additional settings
SKU Single line of text Text format
Product name Single line of text Text format
Total Currency Use Decimal if Currency isn't available
Status Single line of text Text format
Ship to Multiple lines of text No additional settings

Customer Records

Display name: Customer Records. Description: Customer profile records. One row per customer. Primary column: Customer email (text).

Additional column Data type Settings
Customer name Single line of text Text format
Preferred channel Choice Email, Phone, SMS
Loyalty tier Choice Bronze, Silver, Gold
Notes Multiple lines of text No additional settings

Products

Display name: Products. Description: Product catalog. One row per SKU. Primary column: SKU (text).

Additional column Data type Settings
Product name Single line of text Text format
Category Single line of text Text format
Wattage Whole number No additional settings
Warranty months Whole number No additional settings
Compatibility notes Multiple lines of text No additional settings

Return to the sample-data import

After all four tables appear in Customer Support Rep Assistant, return to Task 3 in Lab 2, Exercise 0. Import the three CSVs separately and complete the starting-state checks. Creating the schemas alone does not populate any rows.