Guide for hiring teams
Let candidates use AI under clear rules you state up front, give them realistic engineering work instead of puzzles, and review how they got to the result: whether they directed the AI with clear constraints, caught incorrect or incomplete output, verified the work with tests, and can explain what they chose to ship. Keep the hiring decision with people.
Six steps
Decide whether AI is allowed for this role and this task, which assistants candidates may use, and what is recorded. Tell candidates in the invitation, not at the start of the session.
Clear rules remove the guesswork that makes candidates nervous and makes results hard to compare. When everyone works under the same conditions, the evidence means the same thing for each person.
Modern engineers use AI, so a puzzle an assistant can solve in seconds tells you little. Give candidates the kind of work the job involves: a small service with tests and logs, a bug to fix without hiding failures, or a change to review.
Set a time box and ask for an output you can review, such as a patch plus a short decision note explaining what they changed and why.
Access is not the compromise. Uncritical use is the risk. The useful signal is whether a candidate can direct AI, question it, verify it, and take responsibility for what ships.
A single score hides how the work was done. Keep the code, the tests, the prompts, a replay of the session, and the candidate’s own explanation together so reviewers can check the reasoning, not just the outcome.
Treat integrity events, such as unusual pasting or leaving the tab, as context for a reviewer to investigate, never as a cheating verdict on their own.
Use a short live interview to explore the parts of the take-home work that were unclear: a decision that was not explained, a test that was skipped, or an AI suggestion that was accepted without checking.
A focused follow-up is quicker than repeating the whole assessment, and it gives candidates a chance to show depth.
Assessment evidence is one reviewable input, not a complete verdict on an engineer. Your team reviews the work, decides who advances, and communicates the outcome.
DevEval runs assessments and live interviews this way. See how it works for hiring teams or pricing.
See it in your hiring context
A 15-minute walkthrough with your open role: the task, the evidence packet, and the reviewer workflow your team would use.
