Building Steps
Build steps by combining business intent, stable data inputs, deterministic browser actions, and reviewable assertions.
Recommended Workflow
Start with the smallest meaningful behavior, validate it, then compose it into a case or workflow. This makes failures easier to diagnose and improves reuse.
Write the step goal in plain language.
Identify required input data fields.
Generate or write the implementation.
Add assertions for the expected page state or API result.
Run with one row and inspect logs before scaling up.
Maintenance
Workday UI changes, tenant configuration, and permissions can all affect steps. Keep steps short, review failed screenshots, and use AutoFix or AI assistance as a starting point rather than an unreviewed patch.
Practical Tips
- Prefer semantic actions and visible labels over brittle DOM paths.
- Capture enough logs to understand where failure happened.
- When a step fails only for one row, check row data before editing code.
Data Binding
Data binding lets one step pull values from the current row, environment variables, or workflow parameters. Clear bindings make generated and hand-written steps easier to reuse.
Name each data column for its business meaning.
Use the same column name across related suites when possible.
Keep secrets in environment configuration, not row data.
Run one row first and confirm each value reaches the intended field.
Review Before Save
AI-generated code should be treated like a draft. Review selectors, waits, data usage, logs, and assertions before saving it into a shared suite.
Practical Tips
- Remove debug logs that expose sensitive data.
- Make error messages explain the business failure.
- Prefer one clear assertion over several vague checks.