Use AI tools in your engineering workflow
Use coding assistants to move faster while staying responsible for the result.
AI coding tools can help explain unfamiliar code, suggest tests, draft a small change, or compare approaches. You remain responsible for requirements, correctness, security, and the final code.
Use a reviewable workflow
Start by defining the task and constraints. Share only the relevant, approved context. Ask for one bounded change instead of handing over an unclear goal. Inspect every changed line and run the checks that apply to the code.
Use AI output as a proposal. Check claims against official documentation or source code. Reject code you cannot explain. Keep a note of what the tool missed or where you had to correct it.
GitHub’s Copilot responsible-use guide warns that generated answers and code can be wrong or insecure. Its agent guidance calls for human review and testing before merging.
Protect code and data
Follow the company’s approved-tool policy. Do not paste customer data, private code, credentials, incident logs, or unreleased plans into a tool unless the organization explicitly permits that use. Check which files and tools an agent can access before running it.
Keep agents on the least access needed for the task. Run generated code in a safe environment. Review commands before they run, especially commands that delete files, change permissions, send data, or affect production systems.
Explain how you use AI
Use a real example and cover four points:
- The task you needed to finish.
- What you asked the tool to do and what context you provided.
- How you checked its output.
- What it missed, and what you changed yourself.
For example, describe how you used an assistant to suggest edge cases for a parser, then wrote tests and checked the implementation yourself. Use this only if it matches something you actually did. Do not invent time savings or claim that generated code was your unaided work.
Follow interview rules
Interview formats may differ on whether AI tools are allowed. OpenAI says this explicitly in its interview guide. That is evidence about OpenAI’s process, not every employer’s process. Check the instructions for your interview and ask the recruiter if they are unclear.
When an interview does not allow AI, do not use it. When tools are allowed, explain how you are using them and continue to reason through the solution yourself. Practise both ways so you can show your fundamentals without relying on a tool.
Practice prompt
Try this with a small, familiar code change: ask an AI assistant to suggest missing tests. Review every suggestion, add the useful tests, and run them. Then explain which suggestion caught a real gap and which one you rejected.