BPaaS candidate screening¶

Status: Current | Last reviewed: 2026-10-19

The BPaaS screener collects five spoken customer-service experiences for a recruitment reviewer. AI transcribes and organizes the evidence. A person makes the hiring decision.

Open https://assessments.dev.trueserve.deloitte.com and enter your assigned Name, Email and Code, or select Log in with Deloitte SSO. See sign-in and invitations.
On My assessments, choose Get started under Behavioral Assessment. English Proficiency is available alongside it.
Read the introduction, Terms and Privacy notice. Responses should describe real experiences in your own words. STAR means situation, task, action and result.
Agree to recording, transcription and AI-assisted human review. Every signed-in Deloitte user admitted to this demo has reviewer access to submitted evidence. Optional model-improvement reuse is separate and starts off.
Select Run ten-second microphone check, speak naturally, play the check, then confirm that you can hear yourself clearly. The check is not submitted.
Select Start assessment when you are ready. Allow approximately 30 minutes in a quiet place.

Contact an assessment administrator before starting if you need an adjustment, cannot consent or have questions about the demo. For SSO users, Start new assessment creates a separate attempt after submission. Code users get one attempt per assigned type; an administrator must issue a new invitation for another attempt. Continue assessment resumes an unfinished attempt with its original answers and timers; submitted answers remain locked.

For a presentation, use View AI review on the Behavioral card to open your latest completed AI result, then select Reveal AI-assisted assessment. This link stays available while you work on a fresh attempt. Earlier submissions and their review links remain in Your behavioral assessments, with dates to distinguish them. See the demo walkthrough.

Answer five questions¶

Questions appear one at a time, with a counter and visible timer. You have 30 seconds to start recording and up to five minutes to answer.

Select Submit answer when you finish early. Recording stops at the time limit. If you do not start in 30 seconds, that question is marked No response and the assessment advances. Submitted answers cannot be re-recorded or replaced.

The questions cover handling an upset customer, taking ownership, spotting a problem proactively, resolving complexity, and learning from negative feedback. Use specific actions and outcomes, rather than repeating the question or describing only what you would do.

When all five questions close, the confirmation page shows your assessment reference and says You’ll receive feedback later. The candidate page does not display AI scores. For Deloitte SSO reviewers, Open reviewer workspace lets you explicitly reveal your AI assessment when analysis has finished. A different person must save your human review.

Recover after a technical interruption¶

Reopen the same browser page first. Audio already uploaded is retained. If the original recording timer is still running, Resume recording continues that answer. Timers do not reset. If time has expired, saved audio is finalized and the interruption is visible to the reviewer.

If your session has ended or you need a different browser, sign in with the same invitation details or Deloitte account and open the saved Behavioral attempt from My assessments. An administrator can replace a lost invitation code without resetting the assessment. Recovery returns to the same attempt and cannot replace submitted answers.

Review submitted evidence¶

Every admitted Deloitte user can open Behavioral reviews after SSO sign-in. Assessment configuration remains restricted to administrators.

Open the candidate's Review workspace after submission. You can also open your own submission to inspect AI evidence, but cannot save a human review of yourself.
Expand each question to listen to the recording. You may record an independent judgement before revealing AI.
Select Reveal AI-assisted assessment to view the provisional suitability score, meaningful-response count, competency breakdown, transcripts, supporting evidence quotes, STAR coverage and flags.
For another person's assessment, enter reviewer notes, an optional human content score, and a final disposition. Select Save human review. Earlier reviews remain in the history.

An overall score is shown only when the configured evidence threshold is met. Insufficient evidence differs from demonstrated low capability. Review missing questions and technical interruptions before deciding whether to request more evidence. Speech intelligibility, pace and quiet audio are shown separately from customer-service competence.

AI reveal is audited. A rating made after the same reviewer has seen AI is not counted as independent, even in a new browser session. No AI score automatically accepts, rejects or disqualifies a candidate.

Retry AI analysis queues another run using the original frozen configuration and preserves human review history. Candidates recover interrupted attempts by signing back in; administrators manage invitation codes in Invitations.

Administrators: questions, rubric and models¶

Administrators, including erigolden@deloitte.com, can open Behavioral controls or Admin console. Ordinary reviewers do not see these controls and cannot call their APIs:

Questions & rubric: edit the target role, industry boundaries, recruitment contact, five questions, criterion weights, 0–4 anchors, minimum evidence threshold, retention and model selection. Weights must total 100. Save a new draft; a different administrator must publish it. Existing attempts keep their original version.
Models & performance: inspect actual provider token counts and comparisons against independent human scores. Missing observations display “—”; small demo samples do not establish model accuracy.
Audit trail: inspect SSO enrollment, consent, recording locks, resumed recordings, evidence access, model runs, publication and review actions.

The initial EDJ configuration is a general customer-service demo. Confirm the role criteria and replace the placeholder recruitment contact before collecting real candidate evidence.

Model improvement¶

The console exports an eligibility manifest for retained attempts with optional reuse consent and an independent human rating. It does not upload candidate data or train a model. An approved training workflow must recheck consent, prepare de-identified data and evaluate a model before its deployed ID is entered in a new reviewed configuration.

Use Withdraw optional reuse permission in the candidate's review workspace when handling a withdrawal request.

Demo measurements¶

The workspace shows candidate completion rate, mean time from submission to first AI result, and mean active reviewer time per candidate. Review time is measured while the review tab is visible. Synthetic demos and time spent viewing one’s own submission are excluded from live metrics.

The manual-review baseline is a configurable assumption. Measure the current manual process before claiming a time saving. For the demonstration, include substantive answers, one response that only repeats the question, and one missed question.