AI CV Screening in Recruitment: What Is Working and What Is Not

The volume problem recruitment firms face

Recruitment consultants spend a significant portion of their week on CV screening. For high-volume roles, a single job listing can generate hundreds of applications within days. Each one needs to be read, assessed against the role requirements, and either progressed or rejected. The time this consumes is time not spent on the activities that actually generate revenue: client relationships, candidate engagement, and closing placements.

The economics are straightforward. A consultant earning fees based on placements is spending hours on administrative screening that follows a repeatable pattern. Every hour spent reading unsuitable CVs is an hour not spent on the candidates and clients that drive the business forward. This is why recruitment firms have been among the earliest adopters of AI screening tools.

Where AI screening is delivering results

The firms seeing real outcomes from AI CV screening have one thing in common: they scope the AI’s role narrowly. The agent receives applications for a specific role, screens them against defined criteria (skills, experience, qualifications, location), and produces a shortlist with a structured summary of each candidate. The consultant then reviews the shortlist and makes the judgment calls about cultural fit, career trajectory, and client suitability.

This approach works because it plays to AI’s strengths without asking it to do what it cannot. Pattern matching against defined criteria at volume is something AI does reliably. Assessing whether a candidate will work well with a specific client’s team is not. The division of labour is clear: the AI handles volume reduction, the consultant handles selection.

The time savings are significant. A screening task that previously took a consultant two to three hours can be reduced to twenty minutes of reviewing a pre-filtered, pre-summarised shortlist. Multiply that across multiple live roles and the capacity gain is substantial.

Where it is going wrong

The firms that are disappointed with AI screening have typically made one of two mistakes. The first is deploying a generic screening tool without configuring it for their specific role requirements. A general-purpose AI that scores CVs on broad criteria produces a shortlist that is too vague to be useful. The consultant ends up re-screening the shortlist manually, which defeats the purpose.

The second mistake is using AI to make the final selection decision rather than the initial filter. When firms try to automate the entire screening-to-shortlist process without human review, they encounter problems with edge cases. Candidates with non-traditional backgrounds, career changers, or those whose experience does not fit neatly into keyword-based criteria get filtered out incorrectly. The firm misses good candidates and the consultants lose trust in the tool.

Both failures stem from the same root cause: asking the AI to do too much. The technology is effective at narrowing a large pool against defined criteria. It is not effective at replacing the judgment that experienced recruiters bring to candidate assessment.

What to consider before deploying

The starting point for any recruitment firm considering AI screening is identifying the specific roles or role types where volume is highest and the screening criteria are most clearly defined. These are the use cases where the time saving will be greatest and the AI’s output will be most reliable.

The screening criteria need to be explicit before the tool is built. Not “good candidates for this type of role” but specific, measurable requirements that the AI can assess against. The more precisely the criteria are defined, the more useful the output will be. Vague criteria produce vague results.

Finally, the AI should feed into the existing workflow, not replace it. The consultant receives a filtered, summarised shortlist and continues from there. The recruitment process itself does not change. The only difference is that the first two hours of manual screening have been handled before the consultant opens the list.

---

Delancy builds standalone AI agents for CV screening, candidate triage, and recruitment workflow automation. Each one is configured for specific role types and screening criteria.