Your job post received 300 applications. At 15 minutes per phone screen, screening all of them takes 75 hours. That is nearly two full work weeks, before you have talked to a single candidate you would actually consider hiring. This is not a productivity problem. It is a structural one - and automation is the structural fix.
The volume problem is getting worse
According to Ashby's 2026 Talent Trends report - covering 1,200 startups, 32,000 hires, and 11 million applications - remote roles receive 42% more applications than on-site equivalents. The average hiring process involves 15 interview touchpoints per hire.
Candidates are also using AI tools now. AI-assisted resumes take minutes to customize. This means more applications per posting, higher variance in resume quality, and less signal per resume. The phone screen is often the first real differentiator - the point where a human conversation reveals whether the polished application reflects the actual candidate.
The problem is that you cannot conduct 300 real conversations with one recruiter. Something gets cut: either you screen a fraction of applicants (and miss qualified candidates), or you compress the conversation to the point where it stops being useful, or you spend the better part of two weeks on phone screens before any evaluation actually begins.
The cost you are already paying
300 candidates at 15 minutes each = 75 hours of recruiter time. At a blended recruiter rate of $29/hour, that is $2,175 in labor before you have moved a single candidate forward. That is the manual screening cost per hire cycle, per role, every time you post.
What AI voice screening actually does
AI voice screening is not a chatbot asking yes/no questions. It is a conversational AI - running on ElevenLabs in WorkSignal's case - that conducts a real phone interview with the candidate. It asks the questions you designed for the role, listens to the responses, asks follow-up questions when appropriate, and produces a full transcript and LLM-scored evaluation when the call ends.
The AI does not make decisions. It does not advance or reject candidates. It surfaces information: a score (0-100), a recommendation (strong_proceed / proceed / consider / reject), the full transcript, and flags for anything that warrants closer review. Your recruiter then reviews that information and decides what happens next.
The shift is from "recruiter spends 15 minutes on a call to gather information" to "recruiter spends 3 minutes reviewing information the AI already gathered." The information quality is comparable. The time cost is not.
You design the questions
Write the screening questions that actually matter for this role. Not a generic template - the specific questions that tell you whether a candidate is worth a human conversation.
The AI conducts the call
Candidates receive an invitation to complete a voice screen. The AI calls them (or they call in), conducts the interview, and handles the full conversation including follow-up questions.
You review the output
Full transcript, score, recommendation, and authenticity signals. Three minutes per candidate instead of fifteen. You decide who moves forward.
The human-in-the-loop design
"Human-in-the-loop" is not a marketing phrase at WorkSignal. It is a product design constraint and a legal protection. No candidate advances based on AI score alone. A human must take explicit action to move a candidate to the next stage.
This matters for two reasons. First, it keeps the judgment where it belongs - with the people who understand the role, the team, and the context that no AI evaluation captures. Second, it is the architecture that protects you legally. The Mobley v. Workday case established that AI vendors can be sued as agents of the employer when AI decisions are effectively final. Human review before advancement is the design pattern that prevents AI scores from becoming final decisions.
The recruiter's job does not go away. It changes. Less time gathering basic information. More time making decisions with better information. That is the right trade.
What "human-in-the-loop" means in practice
In WorkSignal, a candidate with a strong_proceed recommendation still sits in your review queue. You look at the transcript, review the score, and click Advance. The AI did not advance them. You did - with the AI having done the information-gathering work that would otherwise have taken 15 minutes of your time.
20 minutes to set up
Try it on a real role
WorkSignal integrates with Greenhouse, Ashby, and Lever. Set up your first voice screening pipeline in 20 minutes. No sales call required.
The ROI math
The numbers are not complicated.
| Cost | Manual Screening | AI Voice Screening |
|---|---|---|
| Time per candidate | 15 min (recruiter on call) | 3 min (recruiter reviewing) |
| 300 candidates | 75 hours / ~$2,175 | 15 hours / ~$435 |
| Availability | Business hours only | 24/7 - candidates screen on their schedule |
| Consistency | Varies by recruiter, time of day | Same questions, same rubric, every time |
WorkSignal starts at $49/month. One agency placement fee ($10,000-$25,000) covers several years of the platform. If you are using agencies for any of your roles, the math is not close. For a side-by-side cost breakdown against the ATS you are already paying for, see WorkSignal vs Greenhouse.
Compliance considerations
AI voice screening creates compliance obligations. Ontario Bill 149 requires disclosure that AI is used in screening. Quebec Law 25 requires notifying candidates of automated employment decisions. Illinois BIPA covers voice recordings as biometric data.
WorkSignal handles disclosure language automatically - compliant AI disclosure text is generated for each job posting. Candidate consent is tracked with version control. The audit log records 44 action types with checksum integrity.
The compliance obligation is on you, not your vendor. But the right platform reduces the manual compliance work to near zero by building it into the workflow. If you are evaluating any AI screening platform, ask them directly: how does your platform handle Ontario Bill 149 disclosure? The answer will tell you whether compliance is architecture or a checkbox.
When not to automate
Not every role benefits from AI voice screening. Being honest about this matters.
- Executive and senior leadership roles. The initial conversation is relationship-building as much as evaluation. A human call signals respect for the candidate's seniority. The screening function is less relevant when you are evaluating 10 candidates, not 300.
- Roles where volume is not the problem. If you are hiring for one highly specialized position and expect 15 qualified applicants, manual screening is fine. Automation earns its value at scale.
- Relationship-dependent sales or partnership roles. Some roles require the hiring manager to form an initial read on the candidate's interpersonal style before deciding whether to invest more time. A human conversation provides that. AI screening provides competency signals - which is different.
The correct use of automation is not to remove humans from hiring. It is to remove humans from the parts of hiring where they add the least value - and concentrate their time where they add the most.
Your job got 300 applications. Let's find the 8.
WorkSignal AI voice screening with built-in compliance. Set up in 20 minutes, integrates with your ATS. No per-seat pricing. No sales call required.
Related comparisons
