Skip to content
Get started
Tasks

AI assessment tasks, taken from real work.

Each task is real work from the role. The candidate opens a link, reads a short brief, gets the files, and does the work in their own AI tool, on their computer or in the browser. You set the time.

Works in Claude CodeCodexClaude CoworkChatGPT

The task library

A full assessment for each role. When you set up a role you see every task for it, and you can bring your own.

Full assessment

One real task from the job, with a first read from Lemonsignal that your team checks. You get back a level for each skill, with the moment it came from.

Engineering

Fix a double charge

Customers are being charged twice. Find out why and fix it.

See the task
Product

Decide an onboarding release

New users drop off in onboarding. Decide what goes into the next release.

See the task
Design

Redesign an ID check

40% of new users drop off at the ID check. Redesign that step.

See the task
Marketing

Plan a launch budget

Split a €50k launch budget for Germany, and write the launch email.

See the task
Data

Explain a revenue drop

Revenue is down 12% in March. Explain why to the CFO in five lines.

See the task
Finance

Find the overbilling

Check 400 supplier invoices against six contracts for overbilling.

See the task

What’s in every task

The candidate sees the first two. We write the other three before anyone takes the task, so every candidate is judged the same way.

What the candidate gets

  1. 1
    The brief and the files. What someone in the role would be handed, and nothing more.
  2. 2
    A part that repeats. In role tasks, the same work comes round a second time. You see whether they reuse what they built, or start again by hand.

What we write first

  1. 3
    A right answer. So every candidate is held to the same thing.
  2. 4
    Places where the work needs checking. Real work has them too. They show who checks the AI’s output before using it.
  3. 5
    A scoring sheet. What a candidate has to do for each skill.
How we score

Bring your own task

Want to see how candidates handle the work your team actually does? Use it as the task.

Get started
  1. 1
    You send us a real piece of work and tell us what a good result looks like.
  2. 2
    We mark the places where the work needs checking and write the scoring sheet.
  3. 3
    You choose the tools and the time: a deadline, a time limit, or both.
  4. 4
    Lemonsignal gives a first read of every session, for your team to check.

See how candidates really work with AI.

Get started

Try it free on 3 candidates.