Sapia.ai

What it is

Sapia.ai (formerly PredictiveHire) is a structured AI interview platform built around one idea: interview every applicant by text chat instead of screening resumes. A candidate answers the same set of role-specific, free-text questions at their own pace — no time limit, no video, no multiple choice — and Sapia’s scoring engine (marketed as the Smart Interviewer / SAIGE) rates the answers against a validated competency model and returns a ranked shortlist with written, explainable feedback for every person. Because only the text is ingested, the screen never sees a face, an accent, or a name-implied background. The company is Australian (Melbourne), led by CEO Barb Hyman since 2018, rebranded from PredictiveHire in 2022, and backed by roughly $17M in Series A funding (Macquarie, Woolworths’ W23). Sapia reports 10M+ candidates interviewed to date.

It is not a video interviewer and not a coding test. The category-correct frame is the text-interview pole of structured screening, opposite the video-first incumbents like HireVue and the live AI interviewers like Apriora.

Why it shows up in Recruiting/TA stacks

• High-volume frontline and graduate hiring. Sapia’s anchor customers — Qantas, Woolworths, Starbucks, Holland & Barrett, BT Group, Costa Coffee — run retail, hospitality, contact-centre, and grad funnels where you genuinely want to interview everyone and a human can’t. This is the scoped use case where it earns its keep: thousands of applicants per req, where a 10-minute chat that screens and scores 24/7 collapses a multi-week phone-screen queue.
• Candidate experience as the differentiator. Every applicant gets the same questions and a personalised written feedback report, untimed, screen-reader and speech-to-text compatible. For employer-brand-sensitive volume hiring, that “everyone gets a fair go and a reply” property is the reason teams pick it over a resume-keyword filter.
• Bias-defensibility paperwork. Sapia’s FAIR™ framework, pre- and post-deployment bias testing, and an ICO (UK regulator) audit give TA and Legal a documentation trail when an AEDT (automated employment decision tool) lands in front of compliance. Blind text removes the visual/audio signals that make video assessment harder to defend.
• ATS-native delivery. Native Workday integration, plus SmartRecruiters, Greenhouse, Avature, PageUp, eArcu, and SAP SuccessFactors (via middleware), and a listing on the SAP Store — so the chat fires as a stage in your existing pipeline rather than a separate portal.

Pricing

Sapia does not publish a price list; deals are negotiated per account and priced on interview volume, not per seat — which is the right model for an “interview everyone” tool. Third-party trackers put the entry point near $1 per completed interview with a 100-interview/month floor ($100/mo), a Growth tier around $500/mo, and Enterprise commonly $2,000/mo and up for the integrated ATS, SSO, and audit support. Treat those as triangulated indications from resellers and buyer reports, not vendor-published numbers, and expect geography and industry to move them. The economic test is simple: per-interview pricing rewards you when applicant volume is high and punishes you when it’s low — model your real annual applicant count, not a headline rate.

Best for

High-volume Recruiting/TA teams hiring frontline, hourly, retail, contact-centre, or early-career roles at scale — the kind of funnel where every req draws hundreds to thousands of applicants, candidate experience is a measured employer-brand metric, and you need a defensible, consistent first-round screen that runs around the clock. English plus multilingual text coverage; strongest in markets where Sapia already has volume reference customers (UK, ANZ, EU retail).

Do not buy Sapia.ai if your hiring is low-volume senior or technical work, or if your screen needs to test a hard skill (the text interview measures communication, drive, and integrity signals, not whether someone can write SQL), or if your stakeholders specifically want a video or live-conversation interview. Under low req volume the “interview everyone” model and per-interview floor don’t pay back.

Versus the alternatives

• HireVue — the market incumbent and the default reason a buyer is in this category at all. Pick HireVue when stakeholders insist on video assessment, game-based ability tests, or a broader enterprise assessment suite. Pick Sapia when you want the opposite: a blind, text-only screen that’s easier to defend on bias and that candidates consistently rate higher.
• Paradox — the conversational-hiring leader for high-volume hourly. Paradox (Olivia) owns the scheduling-and-conversational-apply motion; Sapia owns the structured-scored-interview motion. Many high-volume teams want both — Paradox to capture and schedule, Sapia to assess. If you can only buy one and your bottleneck is no-shows and scheduling, that’s Paradox; if it’s “which of these 4,000 people do we talk to,” that’s Sapia.
• Apriora — the fast-growing real-time AI interviewer (agent “Alex”). Pick Apriora when you want a live, voice/video conversation with follow-up probing and cheat detection. Sapia is the deliberately lower-tech, blind, asynchronous counterpoint — less impressive in a demo, easier to stand behind in an audit.
• Maki People — the conversational candidate-assessment alternative if you want a broader assessment-builder rather than a fixed chat-interview format.

If none fit and your volume is modest, the honest answer is a structured human phone screen against a scorecard — Sapia’s value is entirely a function of applicant volume you can’t staff against.

Watch-outs

• It measures signal, not skill. The chat scores communication, conscientiousness, and integrity-type traits — not technical ability. Guard: for any role with a hard-skill gate, pair Sapia with a real skills test (CodeSignal, HackerRank) and treat the chat as the behavioural/screening layer, not the competency proof.
• It’s an AEDT — regulators are watching. A scored AI interview is an automated employment decision tool under NYC Local Law 144 and similar emerging US/EU rules. Guard: before go-live, get the latest independent bias audit and per-role adverse-impact data in writing, publish candidate notice/consent, and keep a human reviewing the shortlist — wire this into your recruiting AI policy.
• Vendor outcome stats are case studies, not your baseline. Headline figures (time-to-hire and turnover reductions in the 50–90% range) come from reference customers with very specific funnels. Guard: run a back-test on your own historical reqs — does the score actually correlate with quality-of-hire and retention in your population — before you let it gate candidates.
• The model fits volume, not breadth. Sapia is sharp on one motion (first-round structured screening at scale) and isn’t an ATS, a sourcer, or a scheduler on its own. Guard: keep your system of record (Greenhouse, Workday) as the source of truth, confirm the stage-status sync both ways, and don’t expect it to replace Paradox-style scheduling.