6.2 Building Flows: Triggers, Actions & AI Assistance

Every flow, whether cloud or desktop, is built from the same two structural pieces: a trigger that starts the flow and one or more actions that carry out its work. The PL-900 exam tests your understanding of how connectors expose these building blocks in the Power Automate flow designer, and — increasingly important on the current exam — how AI-assisted flow creation lets makers build and modify flows using natural-language descriptions instead of manually assembling every step.

Triggers vs. Actions

Every flow has exactly one trigger — the specific event, schedule, or manual click that starts the flow running — followed by any number of actions, the sequential (or conditional/parallel) steps the flow executes once triggered.

A flow can only have one trigger, but it can chain dozens of actions — including conditions (if/else branching), loops (apply to each), and parallel branches — after that trigger fires.

Connectors

Connectors are the prebuilt bridges between Power Automate and external services. Each connector exposes a set of possible triggers (events it can watch for) and actions (operations it can perform). Connectors fall into familiar categories:

• Standard connectors — included with most licenses (Outlook, SharePoint, Teams, OneDrive)
• Premium connectors — require additional licensing (Dataverse, SQL Server, many third-party services)
• Custom connectors — built by makers or developers to reach an API that has no existing connector

Choosing the right connector trigger is the first design decision for any flow — it determines what kind of flow (automated, instant, or scheduled) is even possible, since the trigger dictates the flow type.

The Flow Designer

Flows are assembled visually in the Power Automate flow designer, where makers search for a trigger, then add and configure actions step by step. The designer supports:

• Conditions — branching logic ("If approved, do X; otherwise, do Y")
• Apply to each — looping over a collection of items (e.g., every row returned from a SharePoint list)
• Parallel branches — running multiple action sequences simultaneously
• Expressions — formulas for transforming data between steps (dates, strings, math)

Makers do not need to write code to use any of these; the designer exposes them as configurable, point-and-click elements — consistent with Power Automate's position as a low-code tool.

Building and Modifying Flows with AI

A defining capability of the modern Power Platform — and a topic the July 2026 PL-900 blueprint calls out directly — is using AI to create and modify flows from natural-language descriptions. Instead of manually searching for a trigger and adding actions one at a time, a maker can describe the desired automation in plain language (for example, "When a new file is added to this SharePoint folder, notify my team in Teams and save a copy to OneDrive"), and AI within Power Automate proposes a working flow — trigger, actions, and connections — that the maker can review, adjust, and publish.

This AI assistance applies to both cloud flows and desktop flows:

• For cloud flows, natural-language prompts can generate the trigger/action sequence, suggest connectors, and even explain what an existing flow does.
• For desktop flows, AI can help generate RPA steps from a description of the task, reducing the need to manually record every click.

Makers still review and refine AI-generated flows before publishing — AI accelerates flow creation but does not remove the maker's responsibility to validate connector selections, test conditions, and confirm the flow behaves as intended.

Testing and Managing Flows

Once built, flows can be tested manually before being turned on for production use. Flow owners can view run history showing whether each execution succeeded, failed, or is still running, which is essential for troubleshooting. Flows can be turned off, edited, or shared with co-owners, and ownership/permissions on a flow are separate from permissions on the underlying data the flow touches.

Key Takeaways

• Every flow has exactly one trigger and any number of actions; connectors supply both
• The flow designer supports conditions, loops (apply to each), parallel branches, and expressions without writing code
• AI can create and modify both cloud flows and desktop flows from natural-language descriptions
• Makers still review AI-generated flows before publishing; AI speeds up building but doesn't replace testing