AI-Assisted Candidate Evaluation for Government Hiring
Employment & Work
What it collects that can identify you
- Audio recordings of each candidate's spoken responses to structured behavioural interview questions, submitted asynchronously via the Knockri platform. These recordings are tied to identified candidates and include voice data.
- Candidate identity information including email addresses used to invite candidates, communicate results, and link assessment records to specific individuals within the TBS staffing process.
- Run by
- Treasury Board of Canada Secretariat (TBS)
- Where
- No fixed location
- Kept
- Retained 5 years after the last administrative action
- Shared with
- Accountable organization, Vendor
What it is for
This system uses artificial intelligence to automatically administer and score structured behavioural interviews for job candidates applying to the Treasury Board of Canada Secretariat. Candidates record audio responses to interview questions, which are transcribed and analyzed by the Knockri platform to generate behavioural attribute scores. The scores are one of several inputs reviewed by human HR advisors, who retain the final hiring decision. Candidates may opt out and request an in-person alternative assessment.
What it collects and what happens to it
Data taken in
- Audio recordings of each candidate's spoken responses to structured behavioural interview questions, submitted asynchronously via the Knockri platform. These recordings are tied to identified candidates and include voice data.
- Candidate identity information including email addresses used to invite candidates, communicate results, and link assessment records to specific individuals within the TBS staffing process.
Processing
- The system uses speech-to-text transcription to convert audio interview responses into text, which is then analyzed by a specialized natural language model. The transcription has been validated to meet industry word error rate guidelines across demographic groups.
- A specialized natural language model finely tuned to identify work-related behaviours in transcribed interview text. The model compares identified behaviours against behavioural rubrics to score each candidate's response. The algorithms are static — they do not change once in production.
What it does
- The system scores candidates' behavioural interview responses, producing question-level attribute scores and an overall assessment score. These scores rank and classify candidates' demonstrated behavioural attributes but are advisory — HR advisors and hiring managers retain the final hiring decision.
- The system automatically transcribes audio recordings of candidate interview responses into text. This transcription step operates without human involvement and feeds the downstream natural language analysis pipeline.
Outputs
- The system produces question-level behavioural attribute scores, an overall assessment score, a Behaviour Insight report (visual breakdown of behaviours demonstrated), full audio recordings, text summaries of responses, and shareable candidate feedback reports. These outputs are advisory inputs to human hiring decisions.
Run by
- The Treasury Board of Canada Secretariat (TBS), Human Resources Division, is the deploying government institution accountable for the Knockri pilot project and its use in TBS staffing processes.
Built by
- Knockri is a software company that specializes in AI-driven hiring assessments. They provide the AI-based skills assessment platform used in this pilot, utilizing natural language processing and machine learning to analyze candidates' behavioural interview responses.
Kept for
- Personal information collected during the staffing process (including audio recordings and email addresses) is retained for five years after the last administrative action and then deleted. It may also be deleted earlier at the request of the HR Division or the individual candidate.
- Duration: 5 years after the last administrative action
Shared with
- All system outputs — including question-level scores, overall assessment scores, Behaviour Insight reports, audio recordings, text summaries, and candidate feedback reports — are available to the TBS HR team via a Knockri dashboard. Access is controlled with grant, monitor, and revoke permissions.
- Knockri, as the platform provider, processes and has access to candidate audio recordings, email addresses, and assessment data as part of delivering the service. A data sharing agreement with appropriate safeguards has been established between TBS and Knockri.
Stored
- Candidate data (audio recordings, assessment scores, email addresses) is processed and stored on the Knockri platform, which is operated by a foreign third-party (Canadian company Knockri). The system interfaces with external IT systems. A data sharing agreement with appropriate safeguards is in place.
- Duration: 5 years after the last administrative action
How to read the colours
Can it identify you?
- Anonymized data
- Data about people with the link to who is broken. Stripped of identifiers, blurred, aggregated, or noised so this system can’t reasonably tie a record back to an individual.
- Pseudonymous data
- Each person’s data is tied to a token (hash, ID, template) that lets this system recognise the same person across events, but the token itself doesn’t reveal a name. Reidentification is possible with extra information.
- Identifiable data
- The data either contains a direct identifier (name, address, account name, recognisable face or voice, plate number) or carries a token this system uses to look up legal identity during processing.
Who completes the loop?
- Human decides
- This mode suggests; a person decides what to do next. The AI is always advisory — a human is in the loop on every decision. Example: a triage tool ranks cases for a clinician who chooses which to see first.
- Human executes
- This mode decides; a person carries out the result. Example: an optimizer plans the day’s trash-collection routes, and drivers run them.
- Autonomous
- This mode decides and acts on its own. No person reviews each decision or carries out the resulting action.
Definitions from the DTPR standard. Amber is about your data, violet about who decides. The fuller the shape and the deeper the colour, the more identifying the data or the less a person is involved.
- AI registerGovernment of Canada Algorithmic Impact Assessment Registry — Knockri Pilot Project
- Policy documentAlgorithmic Impact Assessment: Using AI to Automate Candidate Evaluations in the Staffing Process's Assessment Phase
- Policy documentAIA — Project Details, Questions 3–4
- Policy documentAIA — Section 3.2, Privacy Q33
- Policy documentAIA — Section 3.1, Q22 (Employment domain)
- Policy documentAIA — Section 3.1, Q23–24
- Policy documentAIA — Section 3.1, Q26 (Audio Transcription and Analysis)
- Policy documentAIA — Section 3.1, Q52
- Policy documentAIA — Section 3.2, Privacy Q33
- Policy documentAIA — Section 3.1, Q26
- Policy documentAIA — Section 3.1, Q26
- Policy documentAIA — Section 3.1, Q27
- Policy documentAIA — Section 3.1, Q8 and Section 3.2, Q9
- Policy documentAIA — Project Description (opt-out and alternative assessment)
- Policy documentAIA — Project Description (right to challenge, appeal, inquire)
- Policy documentAIA — Section 2 Requirements (Notice) and Project Description
- Policy documentAIA — Section 3.1, Q21 and Section 3.2, Q28–29
- Policy documentAIA — Project Description (right to challenge or appeal) and Section 3.2, Q27
- Policy documentAIA — Section 2 Requirements (Explanation)
- Policy documentAIA — Section 3.2, Privacy Q33 and Q38
- Policy documentAIA — Section 3.1, Q27 and Section 3.2, Q25
- Policy documentAIA — Section 3.2, Privacy Q33 and Q36
- Policy documentAIA — Section 3.2, Privacy Q38
- Policy documentAIA — Section 3.1, Q49 and Section 3.2, Q36
- Register entryPublished by the Helpful Places. Reference b734be58. This disclosure was drafted with AI assistance.Schema: ai@2026-05-06-beta
What you can do
Ask about this system
Questions go to the Helpful Places, not the vendor.
Your rights
- Right to Be Informed of AI UseCandidates are informed before the assessment that an automated AI system (Knockri) will be used to evaluate their interview responses. This choice and its implications are clearly communicated before candidates opt into the automated assessment. A meaningful explanation of the system's role, input data, criteria, and outputs must be published per the Directive on Automated Decision-Making requirements at Impact Level 1.
- Right to a Human ReviewDuring the pilot, all candidates assessed by the Knockri system will also have their results reviewed by an HR advisor. Hiring managers and HR advisors retain the final decision authority — automated scores are advisory inputs only. The system enables human override of automated assessments, and instances of override are logged.
- Right to ContestCandidates retain the right to challenge or appeal their assessment results. After completing an assessment, candidates may inquire about their score. The Knockri system provides a detailed Behaviour Insight report that documents which behaviours were and were not demonstrated, supporting a meaningful basis for any appeal. A recourse process is established for candidates wishing to challenge decisions.
- Right to Algorithmic TransparencyPer the Directive on Automated Decision-Making (Impact Level 1), TBS must publish a meaningful plain-language explanation of the system covering: its role in decision-making, input data and collection method, criteria used to evaluate data, system outputs and how to interpret them, and the principal factors behind a decision. This explanation must be available via a departmental website. The AIA itself is publicly accessible via the Open Government Portal.
- Right to Be ForgottenCandidates are notified that they have the option to request the deletion of their personal information (including email addresses and audio recordings) after the assessment. Personal information collected during the staffing process will be retained for five years after the last administrative action and then deleted, or earlier upon request by the HR Division or the individual.
Risks and safeguards
- Civil liberties harmAlgorithmic bias may produce discriminatory assessment scores for candidates from protected groups (race, gender, language, disability), undermining the fairness goals of the pilot and disproportionately excluding underrepresented candidates.Safeguard: Knockri uses behavioural annotation rather than human evaluation scores as training targets to minimize bias in training data. Technical documentation provided to TBS demonstrates equal algorithmic performance across demographic groups. Gender-based Analysis Plus (GBA+) of the data was conducted. When technically feasible, studies will assess differences in assessment scores across protected groups. Algorithms are static once deployed to prevent data drift. Candidates may opt out and request an in-person alternative at any time.
- Loss of autonomyCandidates may feel coerced into using an automated interview system that records and algorithmically judges their voice, limiting their ability to present themselves as they would in person, or may face barriers if the system misinterprets nuanced responses.Safeguard: Participation in the Knockri assessment is explicitly optional — candidates may opt out and receive an equivalent in-person behavioural assessment following the same rubric. Candidates are informed before assessment about how the platform operates, data collected, and their right to request data deletion. The choice and its implications are clearly communicated prior to consent.
- Reputational harmAn incorrect or unfair automated behavioural score could unjustly exclude a qualified candidate from a hiring pool, causing professional reputational harm and potentially stigmatizing them for future applications.Safeguard: For the pilot, all candidates assessed by Knockri are also reviewed by an HR advisor regardless of their automated score. The system produces a detailed audit trail and Behaviour Insight report that can be used to explain and, if necessary, contest results. Candidates retain the right to challenge or appeal assessment results and may inquire about their score after assessment.