Knockri and Plum — Canada's AI-Screened Diversity Recruitment Pilot That Skipped Required Privacy Reviews

In September 2020, Canada's Department of National Defence (DND) ran an external recruitment campaign for EX-01 (entry-level executive) positions with an explicit goal of increasing the representation of visible minorities in its executive cadre. Of 471 applicants, 422 completed an AI-driven video-based assessment built by vendor Knockri, which scored six leadership competencies from transcribed text answers only, without using visual or vocal cues, and processed the full cohort in about a week. The 74 candidates who advanced then took a 25-minute assessment on Plum, a second AI vendor platform measuring personality, cognitive ability and social acumen, which produced match scores and identified 36 fully-qualified and 38 partially-qualified candidates for a diverse executive talent pool. Applicants could consent to AI screening or request an alternative human-led process. According to DND's own report, published 5 May 2021, seven appointments had been made by 1 May 2021, with five more candidates recommended to the Deputy Minister.
The campaign became a cautionary example of AI governance. On 7 February 2021, the Office of the Privacy Commissioner of Canada confirmed to The Globe and Mail that DND had deployed Knockri and Plum without completing either of two assessments mandatory under Treasury Board Secretariat rules: a Privacy Impact Assessment and an Algorithmic Impact Assessment measuring bias and other risks before an automated tool is used in decision-making. A DND spokesperson argued no algorithmic assessment was required because the AI tools did not make "final decisions", since human panels still reviewed shortlists — a defence the Privacy Commissioner's office rejected, noting the assessments "should have been completed...prior to the use" of the tools. The campaign cost roughly CAD 179,000.
The case illustrates a recurring pattern in public-sector AI adoption: a legitimate equity goal, delivered through commercially available screening tools, executed without the ex-ante compliance steps that Canada's own federal directive requires. It is a useful reference precisely because the outcome (a shortlist of diverse, qualified candidates) and the process failure (skipped mandatory reviews) are both well documented by independent sources, letting other public bodies weigh the efficiency case against the governance risk before replicating it.

Read the full analysis: https://thelogic.co/briefing/defence-department-didnt-follow-government-ai-rules-with-hiring-tool-test/