EXPLORE THE WORKFLOW
From a routine task to a checked result.
1. Start the task
A schedule or an authorised person starts a defined job, such as exporting the previous shift’s production log. The job records the business date and checks that another run is not already handling the same work.
2. Use the application
The automation opens the permitted application, finds the right screen, and chooses the agreed filters. A missing screen or unexpected message should stop this step for investigation.
3. Collect the export
The workflow waits for the correct file to finish downloading. It checks the filename, format, and reporting period instead of assuming that any file in the folder is the right one.
4. Validate the result
Checks compare the output against your rules, such as required columns and duplicate record IDs. Only a result that passes the agreed checks moves to the destination.
Step 1 of 4
Checks passSave the verified export to the agreed location.
A check failsHold the output and notify the person responsible.
What RPA actually does
Think of RPA as an operator following an agreed procedure. It can interact with application controls and extract information from a screen. Microsoft’s desktop automation documentation describes these interactions through windows, buttons, fields, and UI elements.
A useful starting point is a task your team can describe consistently: open a program, choose a date range, export a report, and place the file in a folder. RPA handles the application interaction; additional workflow logic decides whether the result is acceptable.
Where it can be useful
- Repeated exports: collect the same report from an older application on a schedule.
- Approved data entry: enter information that has already passed validation and any required review.
- Cross-system handoffs: move an agreed file between programs that do not have a suitable direct connection.
These are possible use cases, not a guarantee that a particular application can be automated. A pilot should use representative records and the same environment intended for the final workflow.
Where RPA needs care
Application changes, expired sessions, unexpected pop-ups, and missing data can interrupt screen-based automation. A workflow that depends only on fixed mouse coordinates can be sensitive to changes in window position or display settings.
We would first check whether the software offers a supported API or reliable data export. Those options may avoid unnecessary screen interaction. If the task depends on judgement, such as deciding how to handle a disputed quality result, keep that decision with a person or design a separate reviewed AI workflow.
Attended or unattended?
An attended workflow runs with a person involved. An unattended workflow runs in a prepared environment without someone starting or watching every step. The required machine setup, sessions, licences, and access depend on the automation platform.
For example, Microsoft documents specific machine and session requirements for unattended Power Automate flows. “Runs overnight” therefore needs an operating plan: who checks failures, how credentials are maintained, and what happens during an outage.
What to prepare before a first project
- Write the normal steps and list the exceptions your team encounters.
- Provide a sample input and an example of a correct output, with sensitive information removed.
- Identify the task owner, approval points, and permitted applications.
- Agree how to detect failure, avoid duplicate actions, and resume safely.
A good first scope is one repeatable process with an observable outcome. The aim is to reduce repetitive work while keeping responsibility for exceptions clear.
Further reading
- Microsoft: desktop automation
- Microsoft: introduction to desktop flows
- Microsoft: unattended desktop flows
Official references for the concepts in this guide. The workflow examples are illustrative; the design for your business depends on its systems, access, and operating requirements.