Screening Assist Rule: Score the Rubric Before You Trust the Shortlist
Can we trust an async screening tool's ranked shortlist enough to put it in front of a client without a human re-scoring pass? Inspect and rewrite the scoring rubric before any shortlist reaches a client, and keep a human re-score step on every batch until the rubric has been validated against hires the client already approved.
By InnovaAI ResearchPublished
“Can we trust an async screening tool's ranked shortlist enough to put it in front of a client without a human re-scoring pass?”
Inspect and rewrite the scoring rubric before any shortlist reaches a client, and keep a human re-score step on every batch until the rubric has been validated against hires the client already approved.
Agencies forward the vendor's top-five list to the client as the shortlist, then discover the rubric over-weighted keyword matches or penalized non-standard career paths after the client has already interviewed a weak slate. The fix is not a better vendor; it is owning the rubric and sampling rejected candidates before the client sees anything.
White-label screening platforms such as Raffi and HyRiX score applicants against role-specific rubrics and return a ranked shortlist within 24 hours, which is exactly the layer where an agency's judgment gets outsourced to a vendor's weighting. Forrester's warning that AI vendors now control the meter, the gate, and the referee applies directly: the same platform that sets the score also decides which candidates clear it and arbitrates disputes, leaving the agency with no independent check. MIT Technology Review and Snowflake found fewer than half of agent-generated insights lead to meaningful action, so a ranked list is a draft, not a decision.
- •The tool returns a ranked top 3-5 shortlist rather than raw recordings for the agency to review
- •Screening output will be shared with a client under the agency's brand or a white-labeled portal
- •The role is high-volume or junior, where async video replaces the first recruiter call
- •The vendor sets its own scoring rubric and the agency cannot inspect or edit the weighting