Signature feature
AI Sandbox
See how candidates actually work with AI
Every knowledge worker now has an AI assistant on the other side of the screen. So the question that actually predicts performance in 2026 isn't whether a candidate can solve a problem without AI — it's whether they can solve it well with AI: directing the tool effectively, catching it when it's wrong, and getting to a result they can stand behind.
The AI Sandbox is a live, hands-on task — closer to a real coding or work check than a quiz — that hands candidates the tools they'd use on the job and evaluates the collaboration itself. Traditional assessments ban AI and test a world that no longer exists. The AI Sandbox tests the one your hires will actually work in. It is a live job simulation and the candidate evaluation signal that predicts AI-era performance — one a recycled test library cannot give you.
What it measures
Prompting & direction
How clearly and efficiently a candidate steers an AI tool toward a real, specific outcome — not vague requests, but effective instruction.
Verification instincts
Whether they check AI output instead of trusting it — spotting a wrong answer, a subtle bug, or a confidently-stated fact that isn't true.
Judgment & correction
How they respond when the first result is flawed: refining the prompt, fixing it by hand, or deciding the AI is the wrong tool for the job.
Quality of the finished work
Whether the end result actually clears the bar, produced in the realistic, tool-assisted workflow the role runs on.
Why it matters
- ✓Banning AI tests a scenario your hires will never face — they'll be using these tools on day one.
- ✓AI raises the floor and the ceiling at once: it makes weak work look passable and lets strong people move much faster. Only a hands-on check tells the two apart.
- ✓Verification is the new core skill. The real risk isn't that people use AI — it's that they ship its mistakes without noticing.
See it in a real assessment
Watch how H-Evaluate builds a role-tuned assessment — then see a real one end to end.
The other half of the pair
AI Fluency
Measure whether candidates can get real value from AI
Related reading
AI Sandbox assessment: how to run one and score it
Your hires will use AI from day one, so assess them with it. What a good AI Sandbox task looks like, what to measure, and how to read the signals fairly.
AI Sandbox Assessment: Test How Candidates Work With AI
An AI Sandbox assessment measures the real 2026 skill: working with AI well. See how our live approach evaluates prompt craft, verification and judgement.
The five pillars of hiring: what assessments measure
The five pillars of hiring — cognitive, situational, behavioural, domain and AI fluency — predict who can do the job. Why a single score hides the picture.
Frequently asked questions
What is an AI Sandbox assessment?
An AI sandbox assessment is a live, hands-on task where a candidate works on a realistic problem with AI tools available, and is evaluated on the collaboration — how they prompt, verify, correct, and whether the finished work is actually good. It measures working with AI rather than testing in an artificial, AI-free bubble.
Why let candidates use AI during an assessment?
Because they'll use it on the job. Banning AI tests an unrealistic scenario and tells you nothing about the skill that now matters most: getting a strong, trustworthy result with AI in the loop. The AI Sandbox makes that skill visible and measurable.
What does the AI Sandbox actually measure?
Effective prompting, verification instincts (catching when the AI is wrong), judgment about when to trust or override it, and the quality of the finished work. Together these separate candidates who ship AI's mistakes from those who use it to do genuinely better work.
Can candidates work with AI — and how do you test it?
Yes — the AI Sandbox hands candidates the AI tools they'd use on the job and evaluates the collaboration itself. Rather than banning AI or checking whether they can avoid it, it observes how they prompt, verify and correct, and whether the finished work clears the bar. You see working-with-AI as a measurable skill, on a realistic task, instead of guessing at it from a CV line.