Build business application logic and automation
Name the job before you pick the tool
Take three lines from one requirements list: "email the manager when a large order arrives", "summarize each customer complaint in three bullets", and "stop the save when the discount is missing". Each names a different job: orchestrate longer-running work (a longer-running sequence of steps, often across several services, each reached through a connector; the sequence runs after a start event such as the order arriving, outside the save), generate or extract content with AI, and apply business logic in Dataverse. Each job has its own home in Microsoft Power Platform: a Power Automate cloud flow, a prompt or AI model (an AI Builder model) in AI Hub, and Dataverse business logic that runs in Microsoft Dataverse, the data platform under Power Apps. At 40–45%, this is the most heavily weighted of the three AB-410 domains, and it rewards one habit: read the requirement for its job first, and let the tool follow. The classic trap is choosing by what a tool can do: a cloud flow can update a row, but a flow started by a Dataverse row change runs in the background, not inside the database transaction, so a flow failure doesn't roll back the data change; "stop the save" therefore belongs to Dataverse business logic, not to a flow.
The domain unfolds in three steps, one page per job
Read in order, the jobs are three steps. Create cloud flows covers the first: pick the cloud flow type from how a run starts (automatically, instantly or manually, or on a schedule), then connectors, approvals, flow control, and testing. Create prompts and models in AI Hub covers the second: a prompt is a natural language instruction that tells a generative AI model to perform a task, an AI model (an AI Builder model from AI models in AI Hub, not the generative AI model a prompt instructs) is prebuilt and ready to use, or custom and built, trained, and published, and a canvas app calls either with the Power Fx Predict function (on a prompt, it returns generated text, not a prediction), a cloud flow with Run a prompt or the model's AI Builder action. Implement business and process logic covers the third, logic that must act on the form, inside the save, or in a column value, split into four finer jobs: react to data entry (business rules), guide people through stages (business process flows, which, unlike cloud flows, automate nothing beyond guided data entry and controlling entry into stages), derive a value (calculated, formula, and rollup columns), and enforce or reuse logic on the server (automated low-code plug-ins and Dataverse functions, both preview features).
When two tools both work, choose the one whose guarantee matches the requirement's wording
When more than one tool can produce the result, decide by the requirement's cue words: when the logic must run, whom it must cover, or what it must deliver. Immediately on the form points to a business rule, which can lock, hide, or require a column on a model-driven form; for every writer points to an automated low-code plug-in, which runs in the Dataverse transaction as one place for the logic of all apps, flows, and other Power Platform endpoints; guided points to a business process flow, which defines a set of steps for people to follow; derived points to a calculated or formula column; and eventually fits a rollup column or a cloud flow: a rollup column also derives a value, but its values are calculated by scheduled system jobs that run asynchronously in the background, and a cloud flow is inherently asynchronous. No tool guarantees correct generated content, so when it must be right, add a human review step, not another tool: a human should review it before it is posted online, sent to a customer, or used to inform a business decision.
The three jobs in this domain and the tool for each
| Job the requirement names | Tool | First decision | Drill into |
|---|---|---|---|
| Orchestrate longer-running work | Cloud flow (automated, instant, or scheduled) | How a run starts: a service event, a person on demand, or a timetable | Create cloud flows |
| Generate or extract content with AI | Prompt or AI model | Whether a prebuilt or custom AI model fits (such as reading receipts) or the task is an instruction to a generative AI model (a prompt) | Create prompts and models in AI Hub |
| Apply business logic in Dataverse | Dataverse business logic | Which of four finer jobs: react to data entry, guide people through stages, derive a value, or enforce or reuse logic on the server | Implement business and process logic |