20.1 Autopilot for Developers: AI Assistance in Studio and Studio Web
Key Takeaways
Autopilot for developers generates and fixes expressions, summarizes activities, generates sequences, coded workflows, and apps, and refactors coded automations.
In Studio Web, Autopilot helps build API workflows with context-aware JavaScript.
Testing-focused Autopilot features generate tests and test data and import manual test cases.
Specific prompts that name data and expected results produce better output; generated work must still be reviewed and tested.
Autopilot runs through the AI Trust Layer, and governance settings can enable or restrict its features.
20.1 Autopilot for Developers: AI Assistance in Studio and Studio Web
Core Concept: Autopilot brings generative AI into UiPath's design tools. For developers it generates and fixes expressions, builds sequences from a description, writes coded workflows, summarizes existing logic, and creates apps and tests. The exam description lists "Autopilot for Developers" alongside Autopilot for Everyone. Treat Autopilot as an assistant: it drafts, you review.
Autopilot Actions in Studio
| Action | What it does |
|---|---|
| Generate and fix expressions | Writes a VB or C# expression from a description in the expression editor, and proposes fixes for invalid expressions |
| Summarize activities and sequences | Explains what selected activities or a sequence do, useful for reviewing inherited workflows |
| Generate sequence | Builds a sequence of activities from a natural-language description |
| Generate coded workflows | Writes C# code for coded workflows from a prompt (in the code editor, Ctrl+Shift+G opens code generation) |
| Refactor coded automations | Suggests improvements to existing C# code |
| Generate apps | Creates app pages from images or PDFs, or from Data Service entities |
In Studio Web, Autopilot also helps build API workflows: it suggests activities, generates JavaScript expressions and code with context of the workflow's data, and helps with testing and errors.
Autopilot for Testing
Test-focused Autopilot capabilities in Test Manager and Studio include:
- Generate tests for a requirement and generate coded test cases.
- Import manual test cases and turn them into automation candidates.
- Generate synthetic test data.
- Find obsolete tests and generate test reports.
- Generate tests for SAP transactions.
Writing Good Prompts
Autopilot works best with specific prompts that name the data and the expected result:
| Weak prompt | Better prompt |
|---|---|
| "Filter the table" | "From dt_Invoices, keep rows where Status is Pending and Amount is greater than 1000, sorted by DueDate ascending" |
| "Make a date" | "Return the first working day of next month as a DateTime, skipping Saturday and Sunday" |
| "Email workflow" | "Read unread Outlook emails with 'Invoice' in the subject, save PDF attachments to the Invoices folder, and mark the emails as read" |
After generation:
- Read the result. Check variable names, types, and edge cases such as empty tables or null values.
- Run it. Test with realistic data, including cases that should fail.
- Keep standards. Apply naming conventions and error handling; Workflow Analyzer rules still apply to generated work.
Governance
Autopilot uses models through the AI Trust Layer, where administrators govern which AI features and models are available and how data is handled. Organizations can enable or restrict Autopilot features for developers through their governance settings, so some actions may be missing in a given tenant.
Worked Example
A developer must parse a vendor reference like ACME-2026-00042 into its parts.
- In the Assign activity's expression editor, they ask Autopilot: "Split str_Ref on '-' and return the year part as an Int32."
- Autopilot proposes
CInt(str_Ref.Split("-"c)(1)). - The developer notices the expression fails when the reference has no year and asks Autopilot to "return 0 when the reference has fewer than three parts". The revised expression checks the length first.
- They use Summarize on the surrounding sequence to write the workflow annotation, then run a test with valid and invalid references.
Exam-Style Scenarios
Scenario 1: A developer needs a C# helper that normalizes phone numbers inside a coded workflow. Use Generate coded workflows in the code editor, then review the code, add tests, and handle invalid input.
Scenario 2: An expression in an If activity shows a validation error after a variable was renamed. Use the Autopilot fix action in the expression editor, then confirm the corrected expression uses the right variable and type.
Scenario 3: A team inherits a large workflow with no annotations. Use Summarize activities and sequences to understand each part before refactoring, and keep the summaries as annotations after checking them.
Scenario 4: A tester has a list of manual test cases in Test Manager and wants automation candidates. Use the Autopilot testing features that import manual test cases and generate coded test cases, then review each generated test against the requirement.
In every scenario the developer stays responsible for correctness: Autopilot shortens the drafting step, not the reviewing step.
Autopilot Versus Other AI Features
| Feature | Audience | Purpose |
|---|---|---|
| Autopilot for Developers | Automation developers | Build and test automations faster |
| Autopilot for Everyone | Business users | Chat, get answers, and run automations |
| GenAI Activities | Automations at run time | Summarize, classify, translate, extract, and generate text inside workflows |
| Agents | Agentic automation | LLM-driven decisions that call tools such as API workflows and processes |
Common Traps
- Accepting generated expressions without testing null or empty inputs.
- Assuming every Autopilot action is available; governance can disable features.
- Forgetting that generated code still needs error handling and logging.
A developer wants Autopilot to explain an inherited sequence before changing it. Which Autopilot action fits?
Generate sequence
Refactor coded automations
Summarize activities and sequences
Generate apps
What expression language does Autopilot generate for API workflows in Studio Web?
VB.NET
C# only
Python
JavaScript
Autopilot generated a LINQ expression that works for the sample data. What should the developer do before using it in production?
Test it with realistic and edge-case data, such as empty tables and null values, and apply the project's standards.
Nothing; generated expressions are validated by the AI Trust Layer.
Convert it to C#, because generated VB is not supported.
Disable Workflow Analyzer for that file.
Sections you finish are checked off in the contents.