AI use is expected. In an AI-Assisted round you work the way engineers work now: with a coding agent. Nobody counts how much you used the AI. What the hiring team looks at is how you directed it and how you checked what it did.
How the round runs
- Setup (a few minutes). Confirm your name, check your camera, microphone and speakers, and share your entire screen.
- The onboarding call (up to 15 minutes), if your round has one. A voice call with the task's point of contact: a colleague played by AI, or an AI interviewer. They introduce themselves and the team, explain the task, the deadline and how the round runs, and answer your questions. During the call the project's files are read-only and the coding agents are off; your build time starts when it ends. The call is recorded and transcribed, and what you ask in it is part of the assessment.
- The build (30–45 minutes, depending on the round). A workspace opens in your browser: VS Code with the project's repository, a terminal, and the coding agents the employer allows (Claude Code, Codex or both). Read the brief, then work as you would on the job. Some tasks have colleagues you can message in an Ask or Team panel: they're played by AI and know things that aren't written down. Stuck? Tell a colleague what you've tried and ask for help: they may give you a hint (a few per round). Hints are part of the assessment. Some are set in a simulated team, where Cohera may make scripted changes on a colleague's behalf during the build, such as a commit to the code you're working on, a message or an incident report. You're told when each one happens, and they're scripted the same way for every candidate on the task.
- Checks. When you finish, or time runs out, your work is saved and automated checks run on it, including some you don't see while you work.
- The debrief (up to 15 minutes). A voice conversation about your work with an AI interviewer, or with the same point of contact if your round began with an onboarding call: what you changed and why, what you checked, and what you'd do next. Your work stays open read-only while you talk.
A WorkVerse mission runs about 80 minutes: the onboarding call, then four timed parts and a handoff, then the debrief. You build a change (about 20 minutes). Your pull request is reviewed by a colleague: answer each comment, push back on one you disagree with, and make one round of changes (about 10). An incident comes up after a deploy: investigate from the logs, fix it safely and mark it resolved with a short summary (about 10). A teammate asks you to review their pull request: comment on lines and approve it or request changes (about 7). Then you write a short handoff note (what's done, what's risky, what's next) before the debrief. A part ends when its time runs out, and what you've done so far counts. Ending a part early gives the next one more time.
In a design round there's no onboarding call or separate debrief: you talk your design through with the interviewer while you work on it.
What the hiring team looks at
- The outcome. Does the change do what the task needs, including what the brief doesn't spell out: the project's documents, its tickets, compatibility with what's already there?
- Verification. Did you check the agent's work: run the tests, read the diff, try the edge cases?
- Judgment with the AI. Did you give it clear direction, and catch it when it went the wrong way?
- Ownership. Can you explain and defend the result, including the parts the agent wrote?
Tips
- Read the brief and the documents it points to before your first prompt.
- Ask what you'd ask a colleague before you start building: in the onboarding call, or by message if the task has someone to ask.
- Run the tests yourself, and read what the agent changed before you finish.
- A careful change you can explain beats a large one you haven't checked.
- Don't paste personal information, passwords or other secrets into the AI.
What you need
- A laptop or desktop with Chrome or Edge. Phones, tablets and Safari can't run the round.
- A camera, a microphone, and speakers or headphones.
- Your entire screen shared for the whole round.
- A stable internet connection and a quiet place to talk.
What's captured
Your screen, camera and microphone are recorded for the hiring team, and the voice calls are transcribed. We also capture how you work with the AI:
- your prompts to it, and a summary of its responses and of what its commands returned;
- the commands you run and the files you open;
- the changes you make, including which of them the AI made;
- the results of the checks;
- your questions to anyone you can ask in the task;
- in a task with scripted changes, when each one's notice was shown to you.
How the AI Interview Works explains what is and isn't analysed, how the hiring team reviews the results, and how to ask for a human review.
Accommodations
Need extra time, a format without AI coding tools, or anything else? Ask before you start, at support@coheraai.com or through the help link on your interview page. We pass your request to the employer, who decides on it.