How to read an assessment report without overreading it

Separate the layers

A candidate response is the source evidence. In a Scenario or Conversation, the AI interprets whether a predefined behaviour was fully or partially demonstrated in that evidence. A score, where shown, combines findings according to fixed rules and recruiter-adjustable weights.

These are different questions. The AI is not asked to decide overall suitability. A deterministic score calculation means the same findings and weights produce the same calculation; it does not mean every behaviour judgement is correct or that the score predicts performance in the role.

Return to the source

Read the relevant response and the explanation supporting a material finding. Conversation reports include the transcript and playable evidence clips, so the reviewer can inspect the exchange rather than relying only on the summary.

Check whether the candidate had the information and opportunity needed to show the behaviour. A transcript can contain errors or omit context, and missing evidence in one short response does not necessarily establish that someone cannot perform the behaviour.

Compare against the same criteria

The recruiter defines the behaviours and objectives, with assistance from Decision Agent if useful. They should reflect the work and allow different defensible approaches. A response should not need to reproduce one preferred script to demonstrate a behaviour.

Illustrative example: one candidate makes a quick recommendation; another first checks a constraint. If the criterion concerns gathering relevant information, inspect that action and its context. Speed or polished wording alone does not establish better judgement.

Inspect AI use in context

Where a Scenario includes an AI assistant, the report can include how the candidate asked for and used its help. The relevant behaviours are agreed with the recruiter in advance.

Using AI to obtain knowledge can support good work. Asking for a recommendation is not itself a failure. Review whether the candidate understood the suggestion, checked its assumptions and applied it appropriately, rather than treating the presence of AI assistance as the result.

Use uncertainty to guide follow-up

Convert a material gap into a question about the actual response: which assumption did the candidate make, why did they choose an approach, or what would they do if a constraint changed? Keep the follow-up grounded in the evidence.

Recruiters can correct findings in the UI, affecting the individual report. The reason for a correction should be understandable against the source and the agreed criterion. Corrections are not a claim that the system automatically updates every future candidate’s assessment today.

Keep the final decision with people

Use the evidence to support the hiring team’s process. Configured questionnaire and ATS automation can handle defined steps, while recruiters remain responsible for the rules, review arrangements and exceptions. Current human review normally takes place in Decision Agent, with outcomes sent back to Pinpoint where that integration is used.

Decision Agent does not currently claim a validated prediction of suitability or job performance. Future analysis aims to examine links between granular assessment evidence and later outcomes, including performance reviews, once the necessary data exists. That is a future direction, separate from what today’s report establishes.

See the interactive report examples and the published perspective on what validation can establish.