How to design an assessment around useful hiring evidence

Start with the decision

Write down the next decision the assessment should inform. A recruitment agency screening field staff may need eligibility answers first. A team assessing legal counsel may need to see a connected conversation, document redline and negotiation. The platform can support both, but the evidence required is different.

Define observable behaviours relevant to the work. A behaviour should describe what you want to see, without prescribing one exact response. “Clarifies the requirements” allows different useful questions; a rule requiring one particular phrase would narrow the assessment unnecessarily.

Choose a suitable experience

An Intelligent Questionnaire collects screening answers and uses CV information to avoid repeated questions. A text-based Scenario explores choices and consequences over several turns. A Conversation explores a fully voiced interaction with an AI roleplayer.

Combine stages when they answer distinct questions. Specialist Legal, Academia and AI Aptitude offerings are configurations of these shared modules, available today. The recruiter can define the assessment, while generation and the operator help with the initial draft, research and refinement.

Make the task clear

Explain the candidate’s role, the situation and what they are asked to do. Include the information needed for the task, while keeping any intentionally hidden information consistent with the work being simulated. If an AI assistant is included, make its permitted use clear.

Invitations are sent by email, and candidates can complete their configured stages within the campaign’s deadline. Reminders can reflect the work remaining. Candidates can email for technical help; the recruiter decides exceptions. Support does not change a scenario for an individual candidate.

Work through an example before launch

Illustrative example: to assess prioritisation, present two competing requests and enough context to question their urgency. Look for how the candidate gathers relevant information and explains a sequence of action, allowing more than one defensible approach.

Review plausible responses against the behaviours, including an incomplete answer and a good alternative. If reviewers disagree, examine the criterion and the source evidence. A generated draft saves setup work, but the hiring team still owns what the assessment is intended to establish.

Choose what can be automated

For questionnaire automation, recruiters label answers and agree a validation target in advance. The team selects and tunes a suitable machine-learning approach for each question and currently offers to enable automation after that target is met.

Configured knockout, progression and rejection steps can be handled on the platform and passed back through the live Pinpoint integration. Review is normally done in Decision Agent so the team can inspect the evidence and reports. Other integrations and review arrangements can be scoped with the team.

Keep the interpretation bounded

In Scenarios and Conversations, the AI assesses whether agreed behaviours were fully or partially demonstrated in the response or transcript. A displayed score is a fixed-rule combination of those findings, with adjustable weights. Neither the finding nor its score is an established prediction of future performance.

Outcome-linked predictive work is a future direction once the required data exists. The immediate benefit is automating the creation and running of assessments while giving recruiters evidence they can inspect, correct and use in the next decision.

Explore the assessment and report examples and how to read the resulting evidence.