Scenario Simulation: assess judgement in realistic work

What it is for

Scenario Simulation turns a job description or brief into a realistic situation from the work. It can explore judgement, prioritisation, communication and trade-offs across a broad range of roles. The immediate purpose is to automate the running of an assessment and give recruiters detailed evidence they can inspect.

Decision Agent can generate the scenario, objectives and suggested behaviours. The recruiter can define and refine every element, with the operator helping to research the role and adjust the assessment. The behaviours describe what to look for, rather than prescribing one exact answer.

What candidates experience

Candidates work through a text-based situation over multiple turns. A scenario can contain persistent people, systems and facts, with information that is either revealed or still hidden. What happens earlier can affect what happens later: a colleague may remember an unreliable promise, or an unresolved concern may become important again.

The work can centre on an artefact such as a contract, document or negotiating brief. An optional AI assistant can provide help or information. When that assistant is included, what the candidate asks and how they use the help are recorded for assessment against the agreed behaviours.

What recruiters receive

Recruiters receive a report linking behaviour findings to the candidate’s actual responses. The AI is asked a narrow question: was a predefined behaviour fully or partially demonstrated in this evidence? The report explains the finding and points to the material supporting it.

Where a score is shown, fixed rules combine the behaviour findings using weights the recruiter can adjust. This is what deterministic scoring means here. The calculation follows the chosen rules; the underlying AI interpretation can still be challenged. The score is guidance and is not a validated prediction of overall suitability or future job performance.

Worked example: competing priorities

Illustrative example: a candidate acting as a team lead receives a customer escalation while a colleague asks for help with a deadline. They can clarify urgency, identify missing information and explain a workable order of action. A later turn can show what happens when someone relies on the commitment they made.

Different actions can demonstrate the same behaviour. One candidate might gather more context before deciding; another might explain a justified provisional decision and check it promptly. Review focuses on the defined behaviour and the evidence, rather than whether the candidate guessed one preferred script.

How AI use can be assessed

When AI assistance is part of the scenario, the recruiter can include behaviours about how it should be used. Looking up knowledge, exploring an unfamiliar concept or checking an approach can be useful. The question is whether the candidate retains the understanding and judgement needed to apply the help.

The report can bring the candidate’s assistant use together with the work they completed. Asking for a recommendation is not automatically a weakness; accepting it without checking its assumptions may provide different evidence from evaluating and adapting it.

Limitations and human review

A scenario samples behaviour in a designed work situation. It does not establish that a candidate will succeed in the role. Recruiters can inspect and correct findings in the UI, affecting that report. Future work may use a sufficient body of corrections and outcome data to improve assessment, but that should not be confused with an automatic change to every future candidate’s scoring today.

Candidates can complete the assessment within the campaign’s deadline and email for technical help. Assessment exceptions are decided by the recruiter. Support does not change the scenario for an individual candidate, because the exercise is intended to simulate the work.

Future direction: linking assessment to outcomes

With enough data, the aim is to study how assessment evidence relates to later outcomes, including performance in the role. That future analysis could revisit which behaviours and weights are useful. A later performance review would offer a different outcome to learn from than the fact that someone was hired.

This is future work once the necessary data exists. Decision Agent’s current claim is about evidence of predefined behaviours in particular responses, rather than a demonstrated ability to predict good hires or good employees.

Read how to interpret an assessment report, explore the interactive assessment and report examples, or check module credit costs.