Create a foundation for intelligent applications
Some foundation choices become permanent the moment you save them
Some Power Platform choices can be revised at any time, and some are fixed the moment you save them. A Microsoft Dataverse table's ownership type (for example, whether its rows belong to a user or team or to the organization) can't be changed once the table is created. Solution means two things in this domain: the business solution you design, and the Power Platform solution, the container that carries its components between environments (each a container for apps, flows and at most one Dataverse database), developed as unmanaged and deployed as managed. Because you can't edit components directly within a managed solution, you change the unmanaged solution and redeploy. Each solution's publisher has a customization prefix, such as contoso, that helps avoid naming collisions between publishers' solutions. A table's schema name includes that prefix and, unlike its display name, can't be changed after the table is saved, and a saved column keeps its name, and its data type except for converting a text column to an autonumber column. The tempting mistake is to treat a foundation choice as something to tidy up later, when the only fix for a wrong ownership type is to delete the table and create it again.
The domain unfolds in two steps: design the solution, then build the data model
This domain covers the decisions an app builder makes before any screen exists, in two layers taken in order: solution design (which components a requirement calls for and how they travel from development to production), then the data model (the Dataverse tables, columns, relationships and security those components rely on), because the components decide which tables the data model must provide. First, Design Power Platform solutions with AI-enabled tools turns a requirement into components, starting with plans, the copilot-first tool in Power Apps that turns a business use case described in natural language into a proposed Power Platform solution; it then covers which built-in AI to include, which extensibility option to recommend (such as a Dataverse plug-in or a custom connector), which environment type fits, and how the result reaches production. Built-in AI keeps you involved: plans lets you review proposed technologies before saving tables and building the solution, and on model-driven app forms, form fill assistance suggestions aren't saved until the user explicitly accepts them. Second, Build Dataverse data models creates the tables those components use: table types and columns, relationships and their cascading behavior (what happens to child rows when a parent row is deleted or reassigned), prompt columns that store AI-generated output, row summaries that display a Copilot overview, the public views and main forms that apps reuse, and Dataverse security. Microsoft publishes the weight as a range, 25–30% of the exam, and the page header shows its 27.5% midpoint; Create intelligent applications shares that range, while Build business application logic and automation is larger at 40–45%.
When two answers both work, reuse what the platform already provides
When two options both meet the requirement, prefer the one that builds the least. Microsoft advises considering declarative options first, meaning configuration and low-code features such as business rules, and using code only when they don't meet the requirement; in the same spirit, check whether an out-of-box standard table such as Account or Contact can be customized before creating a new table, and write a custom connector only for a service that isn't available as a prebuilt connector. An out-of-box standard table is one Dataverse includes; a table you create can also get the Standard type listed in the table below, so Standard alone doesn't mean out-of-box. The decision tree below goes beyond reuse to place a requirement about what the solution must do, ending in three solution-design choices or the data-model layer: an extensibility option only when declarative options fall short, the data model when the rule must hold for every app that uses a table (as column-level security configurations apply to all data access requests), an environment variable when a value must differ between environments (a solution component that stores parameter keys and values, so the value can change when the solution moves to another environment), and otherwise the component (canvas or model-driven app, Power Pages site, Copilot Studio agent or dataflow) that matches the most important requirement, as Design Power Platform solutions with AI-enabled tools teaches, always built in an unmanaged solution and shipped managed.
The two foundation layers: solution design and data model
| Layer | What you decide | Main options | Can't be changed in place | Drill into |
|---|---|---|---|---|
| Solution design | Which components meet the requirement, when to extend them with code, which environment type hosts them, and how they reach production | Canvas app, model-driven app, Power Pages, Copilot Studio agent, dataflow; code component, client scripting, Dataverse plug-in, custom connector; production, sandbox, trial or developer environment; unmanaged and managed solutions, with environment variables for values that differ per environment | Components in a managed solution can't be edited directly, so changes are made in the unmanaged solution and shipped as a new managed version | Design Power Platform solutions with AI-enabled tools |
| Data model | How Dataverse stores, relates, presents and secures the data | Standard, Activity, Virtual or Elastic table type; column data types; one-to-many and many-to-many relationships; prompt columns and row summaries; public views and main forms; security roles and column-level security | A table's ownership type and schema name; a saved column's name and data type (text to autonumber is the one supported conversion) | Build Dataverse data models |