Evidence / 01
Sets up the task
The first prompt includes the brief and asks for options, trade-offs and missing information.
What context did the candidate give AI?
Read the excerptA worked example · Operations & project delivery
Follow a candidate as they question an AI answer, receive an urgent message and change their recommendation. Then see the evidence your hiring panel can review.
Prefer a quick introduction? Watch the 70-second explainer.
Friday · 09:00
You are an operations project lead. Recommend whether to launch a customer service on Monday or use a staged rollout. Explain the risks and what needs checking. You may use the assessment’s AI assistant, but you are responsible for your recommendation.
The evidence pack contains three conflicting signals:
01 / Project report
The summary recommends proceeding on Monday.
02 / Test log
There is no evidence that fixes have passed re-testing.
03 / Support rota
A weekend release or recovery cannot assume support is available.
View 01 / Candidate work
These excerpts show the candidate’s prompts, the AI’s responses and the decisions that follow. All messages and emails below are simulated inside the assessment.
09:02
Candidate prompt
“Use the brief, project report, test log and support rota provided. Compare a Monday launch with a staged rollout. Set out the trade-offs and unknowns. Separate evidence from assumptions.”
AI response · excerpt
“The project report confirms readiness, so a Monday launch is the strongest option. A staged rollout would delay customer benefits.”
The answer sounds decisive. It overlooks the critical issues in the test log.
09:06
Candidate follow-up
“The test log shows three unresolved critical issues. That conflicts with the readiness claim. Revise the comparison using all three sources. Do not assume the issues are fixed or that weekend support exists.”
Revised AI response · excerpt
“A full Monday launch is not supported by the available evidence. A staged rollout also needs defined scope, verified fixes and confirmed support. Ask for issue owners, re-test results and a support plan before recommending a release.”
The panel can inspect the original answer and the candidate’s challenge, alongside their final work.
09:12–09:14
09:12 · Simulated team message
From: Project manager
“The sponsor meeting has moved forward. Please send your recommendation by 09:25, with the main risk and the action you need from me.”
09:14 · Simulated priority email
Subject: Supplier fix delayed
“The supplier now expects to deliver the fix on Tuesday at the earliest. Delivery timing is not yet confirmed. Re-testing will still be required.”
The candidate now needs to prioritise a short decision update and account for the new dependency.
09:17
Candidate prompt
“Prioritise the 09:25 update. Draft a short recommendation for the manager using the supplier email. Include the decision, uncertainty, requested owners and next checkpoint. Do not promise a release date.”
AI draft · a claim to check
“The supplier will resolve the issues on Tuesday, allowing us to proceed after that.”
Candidate’s correction
“Tuesday is the earliest expected delivery, not confirmed resolution. The fix still needs re-testing. Remove the promise that we can proceed.”
A polished draft still needs judgement. The candidate checks what the supplier actually said before writing the final recommendation.
09:23
Candidate’s final written update
Recommendation: pause the full Monday launch. Three critical issues remain open, weekend support is unavailable and the supplier fix is expected no earlier than Tuesday.
Prepare a staged rollout plan, but do not release until the relevant fixes have passed re-testing and support is confirmed.
Action requested: please confirm an engineering owner for the fix and re-test, and a support lead for coverage. I will update the plan and prepare the customer communication.
Next checkpoint: Tuesday at 12:00, or sooner if the supplier position changes. This is a review point, not a release commitment.
This is one illustrative response, not a prescribed answer for every role. The scenario and review criteria are agreed for each pilot.
View 02 / Human assessor review
Your panel reviews the AI conversation and written work against the criteria agreed for the role. The links below connect each discussion point to the evidence in this example.
Your assessors make the judgement. Your team makes the hiring decision.
Evidence / 01
The first prompt includes the brief and asks for options, trade-offs and missing information.
What context did the candidate give AI?
Read the excerptEvidence / 02
The candidate challenges the launch recommendation against the unresolved issues in the test log.
Did they verify claims against the source material?
Read the excerptEvidence / 03
A manager’s deadline and a supplier delay change what the candidate needs to address next.
How did their next actions reflect the new information?
Read the excerptEvidence / 04
The candidate edits a draft that wrongly treats a possible Tuesday delivery as a confirmed fix.
What did they accept, change or reject?
Read the excerptEvidence / 05
The written update states a decision, the remaining uncertainty, requested owners and a checkpoint.
Can they explain the trade-offs and next steps?
Read the excerptThe approach in 70 seconds
Candidates can use AI under clear rules. Simulated messages introduce changing priorities. Your panel reviews the work behind the recommendation.
The team messages and emails are part of the simulation. They do not require a live Teams or Outlook integration.
70-second explainer with fictional examples. English (UK) captions are available in the player. You can also watch on YouTube.
Start with one role
Bring a professional role you are hiring for. We’ll discuss the candidate experience, what your assessors need to review and the scope of a focused paid pilot.