PL-900 Step by step · Lab 2

Let Copilot Design Dataverse Tables, Then Fix What It Misses

Let Copilot propose four Dataverse tables from a description, fix what it misses, set primary columns and check the relationships.

8 November 2026 · 5 min read

Copilot adds the missing Supplier Code column to the Supplier table — Watch the lesson: Lab 2: Dataverse tables with Copilot | PL-900 Power Platform Step by Step

Pragati Industries keeps its purchase requests in inboxes and its supplier list in three spreadsheets. In Lab 2 of the PL-900 course we replace them with four Dataverse tables. Copilot proposes them from a description; you check every column, because reviewing AI output is part of the job.

Why Dataverse, not a spreadsheet

Who runs the server? Microsoft: it is managed, backed up and scaled for you.
Security: Microsoft Entra sign-in and security roles that reach down to a single row or column.
Business logic: business rules, formula and rollup columns, and flows.
Rich types: choices, lookups, currency, files and auditing.
Apps and AI use it directly; Copilot even understands its tables.

The four tables: Supplier, Purchase Request, Request Line and Invoice. A request belongs to one supplier; lines and invoices belong to one request.

Step 1: Describe the tables to Copilot

Tables → Get started with Copilot. Describe the tables in everyday words: the four tables, their columns, the choices and how they are linked. Select Generate.

Get started with Copilot dialog: describe the tables you want Copilot to build — Describe the data in plain words; the lab guide has the full description.

Copilot proposes the four tables. The dotted lines are the relationships, and the key icon marks the primary name column.

Copilot's proposed tables Supplier, Purchase Request, Request Line and Invoice with relationships — The proposal: four tables and their relationships.

Step 2: Review it like a reviewer

Check every table against the requirement. Supplier is missing the supplier code, although the description asked for it. AI makes mistakes like this. Ask Copilot to add it.

Copilot chat: add a column Supplier Code to the Supplier table, and the column is added — "The Supplier Code column has been added." Then Save and exit creates the tables.

The solution now holds the four tables, each named with the publisher prefix pgi_.

Step 3: Primary columns and relationships

Copilot named each primary name column after its table. Change the display names to what people actually look for: Request Number, Item and Invoice Number. Only the display name changes; the logical name that apps and flows use stays the same.

Then open Purchase Request → Relationships: many-to-one to Supplier, and one-to-many to Request Line and Invoice.

Purchase Request relationships list: many-to-one to Supplier, one-to-many to Request Line and Invoice — The relationships Copilot created, checked by hand.

How you know it worked

The model is in place, and Import → Import data from Excel brings in the 20 real suppliers from the lab files: Dataverse maps the columns by name and asks you to confirm the possible matches.

What a tech lead should take away

Copilot drafts, people decide. Review AI-generated schemas column by column.
Name primary columns for humans, and never rename logical names after apps depend on them.
One shared model under every app, flow and agent is what makes the rest of the platform work.

Do it yourself

Lab guides, practice and full mock exams come with a subscription. See the plans