AI Copilot is designed to transform natural language instructions into executable RPA workflows. While clear prompts usually produce accurate results, you may sometimes encounter situations where the automation doesn’t behave exactly as expected.
This FAQ helps you handle common issues, optimize your prompt design, and troubleshoot edge cases for more reliable automation workflows.
1. Handling Ambiguous Page Elements
One of the most common challenges is instructing AI Copilot to interact with elements that are similar or visually complex. Without enough detail, the AI may select the wrong element or fail to recognize what you intended.
Buttons or links with the same name: When multiple buttons or links share the same label, specify their relative position or the surrounding structure. For example, instead of just “click the Submit button,” you could say “click the Submit button in the login form at the top-right corner of the page.”
Icons instead of text: If a clickable element is represented by an icon rather than text, describe its location in context. For instance, “click the magnifying glass icon next to the search box in the header.”
Split fields: Sometimes a single piece of information is displayed in multiple fields (e.g., first name and last name). In your prompt, instruct the AI to combine these fields during extraction to ensure you get the complete data.
Adding these details gives AI Copilot the context it needs to accurately identify and interact with page elements.
2. Navigating Pop-ups and Page Behaviors
Unexpected pop-ups, modal dialogs, or page transitions can disrupt your workflow if they are not properly addressed in your prompt.
Login prompts or modal dialogs: Include clear instructions on how the AI should handle them, such as closing the pop-up or skipping the step.
Page transitions: Make it explicit on which page each action occurs. For example, “On the search results page, click the first item to open its detail page.”
Search fields without a button: If a page requires pressing Enter instead of clicking a button, include this in your prompt to ensure the AI knows how to submit the search.
Specifying these behaviors prevents the AI from stalling or performing actions on the wrong page.
3. Complex Data Structures
When extracting data from structured sources like tables, nested fields, or Excel sheets, clarity is essential.
Nested fields: Clearly describe each layer and the specific fields to extract. For example, “From each listing’s details page, collect the building name, unit number, and owner information.”
Excel or table structures: Explain the data layout so AI Copilot can map fields correctly. For example, “Extract the column labeled ‘Price’ and save it under a field named ‘Property Price’ in the output Excel file.”
Clear data instructions ensure that AI organizes the results as you intend.
4. Conditional Logic Limitations
AI Copilot can handle conditional logic, but there are some constraints to be aware of.
If/Else branches: During a trial run, AI Copilot typically validates only one branch. This means that complex workflows with multiple branches may not be fully tested in a single generation.
Multi-branch workflows: Plan to create separate prompts or run iterative tests for each branch to verify that all logic paths function as expected.
Understanding these limitations helps you design prompts that reduce unexpected behavior.
5. Platform-Specific Examples
When using real web pages for testing, be mindful of privacy and copyright:
Provide generic, safe examples that illustrate challenges without using sensitive URLs.
Focus on the structure and types of elements rather than specific live sites.
This approach ensures your examples are instructional without introducing legal or privacy risks.
6. Summary & Best Practices
To get the most out of AI Copilot, follow these guiding principles:
Keep prompts structured, precise, and step-by-step. Avoid vague language such as “maybe” or “some.”
Break tasks into clear actions with exact wording where relevant.
Reference the main tutorial for beginner guidance on prompt writing.
Combine these advanced tips with iterative testing to refine your workflows.
By applying these strategies, you can significantly increase the reliability of your AI-generated workflows.
💬 Not sure how to write a good prompt? Explore ready-to-use templates and learn from examples at the Octoparse AI Prompt Hub. You can also join our Discord community to see real-world examples, ask questions, and share your own AI prompts with other users.