Power Automate Cloud and RPA

The name covers two complementary approaches to building automations:

Cloud flows – also known as Power Automate Cloud. They connect online applications and services through triggers, actions, and connectors.

Desktop flows – created in Power Automate Desktop. They use robotic process automation (RPA) to automate tasks performed on a computer and in a web browser.

Cloud flows are primarily responsible for communication between systems, process logic, scheduling, and data processing. Desktop flows handle operations that would traditionally require a person to work directly on a computer. Combining these two layers makes it possible to automate processes that involve both modern cloud services and legacy systems without suitable integrations.

Depending on how they are launched, Microsoft distinguishes three main types of cloud flows:

Automated cloud flow – starts in response to a specific event, such as receiving an email, adding a file to SharePoint, or creating a record in a system.

Scheduled cloud flow – starts at a specified time or according to a defined schedule, such as every morning or every Friday at 3:00 PM.

Instant cloud flow – is launched on demand by a user, for example through a button in Power Automate, the mobile app, Power Apps, or a SharePoint list.

Power Automate is a low-code platform. Many processes can be built without traditional programming by using ready-made actions, conditions, loops, and data operations. More complex scenarios may still require expressions, custom connectors, an appropriate data model, and carefully designed error handling.

Connectors play a key role by enabling communication with Microsoft services and external systems such as SharePoint, Teams, Outlook, Salesforce, Jira, and Gmail. A cloud flow can retrieve and transform data, send notifications, launch additional processes, store information in a database, or trigger a desktop flow on a designated computer.

An automation can open applications, complete forms, copy data, manage files, extract information from websites, and perform operations in legacy ERP or CRM systems. However, a flow does not have to rely solely on screen coordinates. Power Automate can use UI elements and selectors to identify specific fields, buttons, and application windows. This approach is usually more reliable than automation based only on mouse movements and the on-screen position of individual elements.

Desktop flows are particularly useful when an application does not provide an API, lacks a suitable connector, or when a system integration would be disproportionately expensive. In such cases, RPA can act as a pair of “virtual hands” that performs clearly defined tasks on a computer.

How can desktop flows be launched?

Locally in attended mode – the user launches the flow on their computer, for example from the Power Automate Desktop console, through a run URL, a desktop shortcut, or Windows Task Scheduler. The automation runs in an active session, its actions are visible on the screen, and the user must remain signed in.

Remotely from a cloud flow in attended mode – a cloud flow triggers a desktop flow on a registered machine. The process uses the user’s active session and performs actions on their computer.

Remotely from a cloud flow in unattended mode – Power Automate signs in to the designated machine, creates a system session, runs the process, and signs out the technical user account after the work is completed. The automation operates without direct human involvement.

How does the process work step by step?

1. Power Automate – process initiation. Every day at 7:00 AM, a scheduled cloud flow starts the automation and triggers the desktop flow on a prepared machine.

2. Power Automate Desktop – data extraction. The bot opens a web browser, goes to a site such as Booking.com or an airline website, enters the destinations and dates selected by the company, and then extracts the available prices into a structured table.

3. Power Automate – processing and notification. The desktop flow sends the data to the cloud. The cloud flow compares the current prices with previously recorded values. If an airfare price has dropped by more than 15%, it immediately sends the manager a notification in Microsoft Teams or by email. All results are stored in a SharePoint list.

4. Power BI – data analysis. The Power BI report uses the history stored in SharePoint and displays price trends from the past month. Based on this information, the manager can make purchasing decisions, update the offer, and adjust margins accordingly.

The key principle is to minimize the number of steps and optimize licensing costs. Cloud flows should be used wherever applications provide an API or a ready-made connector. A desktop flow should handle only those steps that cannot be completed through a system-level integration. Reducing the number of operations performed through the user interface makes the process less vulnerable to application changes, improves stability, and simplifies maintenance.

Minimize the number of RPA steps and keep them separate from the business logic handled in the cloud.

Use UI elements and selectors instead of relying solely on screen coordinates.

Add error handling, retry mechanisms, logging, and notifications for failed process runs.

For unattended automation, use a dedicated machine and control system and application updates.

Choose licenses based on the actual number of users, machines, and concurrent runs.

Power Automate includes two distinct but complementary approaches to automation. Cloud flows integrate services, process data, and manage process logic, while desktop flows make it possible to work with applications and websites that do not provide suitable APIs or connectors.

With hybrid solutions, an organization is no longer limited to modern cloud systems. It can also automatically retrieve and enter data in environments where human involvement was previously required. The best results come from an architecture in which Power Automate Cloud handles as much of the process as possible, while RPA is used intentionally-only where “virtual hands” are genuinely needed.

This division improves solution stability, makes processes easier to monitor, and provides better control over licensing costs across the Microsoft ecosystem.