An evidence-led approach
Open the work.
Follow the judgement.
This is what inspectable evidence can look like. Explore the work, the checks and the limits behind a recommendation.
This fictional customer-support task explains our approach. It is not a real candidate assessment or a claim about a candidate’s performance.
AI Adoption Lead
Design an AI-assisted workflow for customer queries. Use a supplied returns policy and synthetic enquiries. Include source checks and a human approval step.
A workflow you can inspect.
See how the candidate turns a business task into a usable AI workflow.
EXAMPLE WORKFLOW
- Read the customer query
- Retrieve the relevant policy
- Draft a source-backed reply
- Send for human approval
You receive the task, prompts, workflow and output.
What the task reveals
The assessment asks the candidate to design a customer-support workflow using a supplied policy and synthetic enquiries. The deliverable includes a worked example and the points where a person reviews the result.
What we look for
Can they explain their choices, produce a useful result and make the process repeatable?
Judgement you can follow.
See how the candidate tests the output and corrects a plausible error.
EXAMPLE QUALITY CHECK
- Policy: returns within 30 days
- Customer: purchase was 42 days ago
- AI draft: “You qualify for a refund”
- Correction: flag the conflict for review
You receive their checks, corrections and reasoning.
What the task reveals
This example deliberately includes an AI draft that contradicts the supplied policy. The evidence records whether the candidate identifies the contradiction, checks the source and explains the correction.
What we look for
Do they verify the answer against the source, or accept an AI response because it sounds confident?
The gaps, made visible.
See where a candidate sets boundaries and what still needs testing.
EXAMPLE REVIEW BOUNDARIES
- Use synthetic customer information
- Flag uncertain policy matches
- Require human approval for refunds
- Test exceptions before wider rollout
You receive clear limitations and interview follow-ups.
What the task reveals
A small work sample cannot prove performance at scale. The assessment notes distinguish what was demonstrated from what remains untested, including unusual cases, real system integrations and team adoption.
What we look for
Do they recognise uncertainty, protect information and identify where human responsibility remains?
How this helps you hire
A useful starting point
for a better interview.
For your role, we agree the work and the assessment criteria with you. Your shortlist explains the evidence, the strengths and the areas still to explore.