We ran 1,200 phone screens — 600 by a senior recruiter, 600 by Raffi (AI conversational screening). Same roles, same JDs, same hiring bar. AI screened all 600 in 4 days; the recruiter took 11 weeks. Candidate completion was 71% for AI vs 48% for recruiter (the gap is almost entirely self-scheduling — candidates pick AI slots at 8pm and weekends). Hire rate of screened candidates was 9.1% AI vs 7.4% recruiter (within noise). Quality of the top 3 (rated by hiring managers) was 4.2/5 AI vs 4.4/5 recruiter — slight edge to humans on intangibles. The honest answer: AI doesn't replace a senior recruiter at the top, but it makes the funnel work at a scale and speed no human can match.
What is AI phone screening?
AI phone screening replaces the recruiter's first phone conversation with an AI conversational interviewer that calls every applicant within minutes of applying. Candidates self-schedule, answer the same structured questions a recruiter would ask (role fit, salary expectations, notice period, must-have skills), and the AI scores transcripts against the rubric the hiring team set up. The recruiter then reviews only the top-rated 10-20% — turning a 200-applicant funnel into a 20-finalist shortlist within 48 hours instead of 4-6 weeks. The rest of this guide compares 1,200 head-to-head screens and shows where AI phone screening wins, where it loses, and which funnel stages should use which.
See Raffi in 72 seconds
How Raffi runs the conversational AI interview — end to end. Same loop the article above describes.
Where AI wins clearly
Volume. A recruiter doing thoughtful phone screens does 6-8 per day, max. After that the calls all blur together and quality drops. AI does unlimited parallel calls. For roles getting 200+ applicants, this is the dominant factor — and per SHRM's 2025 benchmarking the US average time-to-fill is 44 days, the majority of that lost in the screening backlog.
Speed. First call within 24 hours vs first call within 2-3 weeks. Time-to-shortlist drops from 6 weeks to 6 days. Applicant ghosting plummets — LinkedIn's Global Talent Trends report flags "slow first conversation" as the leading abandonment trigger.
Consistency. Every candidate gets the same questions in the same order. Comparing them is fair. No recall bias on candidate #18 vs candidate #3.
Self-scheduling. Candidates pick a slot 24/7 in any timezone. No coordinator emails, no missed-call ping-pong.
Cost. ~$5 per completed screening on AI vs ~$50-80 of senior recruiter time per call. At 100 applicants per role, that's $500 vs $5,000-8,000 — a 90% cost reduction even before you count the recruiter time freed for higher-value work.
Where recruiters win clearly
Read between the lines. A great recruiter picks up on a candidate's tone — they sound checked out, they got laid off recently and are anxious, they're using language that suggests they're underqualified but smart. AI can score on transcript but it can't fully replicate that pattern-match.
Selling the role. Phone screens aren't one-way. A recruiter can sell — make a candidate fall in love with the company in 10 minutes, close the candidate who was about to take a competing offer. AI is okay at this but not great.
Edge cases. The candidate who's a brilliant atypical fit. The career-changer who looks weak on paper. The senior person taking a perceived demotion to get into your industry. Recruiters catch these. AI scores them down because the rubric doesn't have a column for "exception."
What candidates actually say about the difference
Greenhouse's 2026 candidate report found that candidate satisfaction with AI phone screening is comparable to human phone screening when AI use is disclosed up front. When AI is used without disclosure, satisfaction collapses. Translation: candidates don't reject AI — they reject feeling tricked. Gartner found 26% of candidates trust AI to evaluate them fairly; that rises sharply with clear disclosure and a published opt-out for a human interview.
The pattern that actually works
Don't pick one. Use both, with AI at the top and recruiter at the strategic candidates layer.
- AI screens every applicant. Top of funnel. 200 → ~80 who pass.
- Recruiter reviews the AI shortlist. Spends 15 minutes per top-10 candidate looking at transcripts, salary signals, anti-cheat flags.
- Recruiter does a 20-minute "deep" call with the top 5. This is where they sell, dig into edge cases, get the candidate excited.
- Hiring manager meets the top 3. Standard final.
This stack does what neither does alone: covers the volume problem AND keeps the human judgement at the moments it matters.
When you should skip AI screening
- High-volume manual labor roles where applicants don't have email
- Senior executive roles (VP+) where the recruiter relationship is the product
- Tiny funnels (<20 applicants) where a recruiter doing each by hand is faster than configuring the AI
- Regulated markets where you can't yet show an NYC Local Law 144 compliant bias audit (NYC specifically; consult counsel for your jurisdiction)
What to look for in an AI phone screening tool
The short version:
- Voice-first, real-time conversational AI (NOT chatbot-with-voice)
- CV-aware before the call starts
- Role-specific behavioral questions (not generic "tell me about yourself")
- Full transcript + recording + structured scorecard
- Anti-cheat scoring on every transcript — see the HBR ethics primer on AI in hiring for why this matters
- 30+ language support
- Self-scheduling 24/7
Want to try one on a fictional role? Step into a room with Raffi — 5 minutes, no signup, no card.