For teams hiring in the age of AI

Great work is now human, augmented by AI.

Most interviews still test for a world without AI. But real work now happens with it — so that’s what we should measure. aptlab drops a candidate into a real, messy case and lets them work it the way they actually do: directing an AI co-pilot, with the tools they’d reach for any day. Every session is graded against a rubric built with domain experts.

Try a case yourselfSee how it works

Runs in the browser — the same environment your candidates get.

What you're really measuring

See how candidates reason with AI — not just where they land.

A candidate directs an AI co-pilot that writes and runs real code on real data. You see the whole path — the questions they pose, the query they take or rewrite, the bad join they catch, the save when the data surprises them. It's the part a résumé can't show, and the part the job now depends on.

A live session — the candidate asks, the co-pilot proposes a query, and nothing runs until they hit Run.
A live session — the candidate asks, the co-pilot proposes a query, and nothing runs until they hit Run.

Grading that holds up

Graded the way your best analyst would — every time.

Every transcript and deliverable is scored out of band against a rubric built with domain experts. The same dimensions, the same bar, for everyone — no reviewer fatigue, no gut calls, no pile of take-homes to wade through. You open a scorecard, not an inbox.

A score per dimension, each point traced back to what the candidate actually did.
A score per dimension, each point traced back to what the candidate actually did.

Also built in

We sweated the details.

Realistic cases, built to produce signal

Every case is a genuine, open-ended business problem pulled from real analyst/DS/DE work — a flat-looking A/B test, a metric nobody's agreed on, a number that doesn't add up. There's no single gotcha to spot; a candidate has to frame the problem and defend a call themselves, the way the job actually works.

See every case
Every case, what it tests for, and who it's a good fit for.
Every case, what it tests for, and who it's a good fit for.

Bring your own case

Upload your own data and prompt in Case Lab, try it live against the co-pilot, and interview on the exact problem your team is hiring for — not a stock template.

Upload data, write the brief, and test-drive the co-pilot before you send it out.
Upload data, write the brief, and test-drive the co-pilot before you send it out.

Integrity, built in

Quiet behavioral signals — paste bursts, tab-switches, an inhuman work rate — surface the sessions that don't look like real work, so a strong score is one you can trust.

A confidence read, with the signals that back it.
A confidence read, with the signals that back it.

The fastest way to get it is to do it.

Take the exact case a candidate would — in the browser, in a few minutes.

Try the demo