Amazon Connect Talent

What it is

Amazon Connect Talent is AWS’s AI hiring product: agents that run structured voice interviews, administer assessments, and score candidates, with recruiters making the final hiring call. AWS announced it in Preview on 2026-04-28 as one of four agentic Connect solutions, alongside Connect Customer, Connect Health, and Connect Decisions.

It is still in Preview as of 2026-08-14 — available in US East (N. Virginia) and US West (Oregon) only, free to try, with no announced GA date and no pricing page. That is why this entry is a watch item rather than a buying recommendation, and why its score sits where it does: you cannot shortlist a product whose price nobody has published.

The mechanism is worth understanding because it differs from one-way video screening. Connect Talent reads a job description, builds an interview plan from it, invites candidates on a flexible schedule, and runs an adaptive voice interview around the clock through a mobile-first, brand-customizable candidate portal. Recruiters then get competency scores with names and identifying details stripped out, plus full transcripts and notes. Preview supports English, Portuguese, French, Italian, German, and Spanish. ATS connection is via bi-directional APIs, and AWS has named no ATS partners — a gap that matters more than the feature list.

Why TA teams are watching it

• Surge capacity bought as infrastructure. The scoped case is high-volume hourly and frontline hiring — warehouse, retail, contact center, healthcare staffing — where hundreds of applicants per req arrive in bursts and the constraint is first-round screen throughput, not sourcing. AWS’s pitch is evaluating hundreds of candidates simultaneously during a hiring surge, which is the one thing a recruiter headcount plan cannot flex.
• Anonymized scoring as a designed control. Names and identifying information are stripped from the recruiter dashboard before review. That is a structural control rather than a policy promise, and it is the most defensible thing in the product today.
• Procurement through an account you already have. For an org on an AWS enterprise agreement, this arrives under an existing vendor, existing DPA, and existing security review. That shortens the buying cycle — it does not shorten the AEDT compliance work, which stays with you either way.

What it costs

Nothing published. It is free during Preview, AWS has issued no price sheet, and there is no pricing page for the product. Plan against the Connect family’s billing convention — consumption metering rather than per-seat — which means a per-seat line in next year’s TA budget is the wrong planning shape even before the number exists.

For sizing the decision, the anchors are what you would otherwise pay: HireVue enterprise programs land in a $35–100K+/year band and Apriora in roughly $10–35K/year. Both are triangulated third-party bands, not vendor-published numbers. Use them to decide whether the problem is worth solving at all, then revisit Connect Talent when AWS publishes a meter.

Best for

Nobody, as a purchase, today. As a pilot: TA teams at AWS-committed enterprises running high-volume hourly or frontline funnels in US regions, who have recruiter capacity as a named constraint and can run a pilot without displacing an incumbent contract.

Do not put it on a shortlist if you need a GA product with a price this fiscal year, you hire outside the US, you run low-volume or senior roles where an AI-only first round costs you candidates, or you need published validation evidence to clear your legal review.

Versus the alternatives

• HireVue — the incumbent to beat. Pick it when you need assessment science, published validation studies, and the bias-audit paper trail an enterprise legal team expects. It is GA, priced, and audited; Connect Talent is none of those yet.
• Paradox — the high-volume conversational leader, now inside Workday after the October 2025 acquisition. Pick it for hourly and frontline hiring at scale, especially if you already run Workday; it addresses the same funnel Connect Talent targets, with a decade of deployments behind it.
• Apriora — the fastest-growing entrant in AI-led interviewing. Pick it when you want live adaptive interviews now, with ATS breadth and integrity detection Connect Talent has not demonstrated.
• Sapia AI — pick text over voice when candidate accessibility and a bias-controlled, structured chat format matter more than conversational realism.
• Fountain — pick it if the real bottleneck is hourly funnel logistics and scheduling rather than the interview itself.

Watch-outs

• Preview with no GA date and no price is not a budget line. Guard: fund it as a pilot only, cap it to one req family, keep the incumbent contract live through GA, and require a written GA date and price sheet before it enters a fiscal plan.
• US-only regions decide your candidate data residency for you. AWS says more regions will follow customer demand, which is not a commitment. Guard: do not route EU or UK candidates through it until an in-region option ships, and get your DPO’s position in writing before any pilot touches non-US applicants.
• Amazon’s own recruiting-AI history is part of your legal review. Reuters reported in 2018 that Amazon scrapped an internal recruiting tool after it learned to downgrade resumes containing “women’s”. Connect Talent is a different product built by a different team, but AWS has published no validation or adverse-impact studies for it. Guard: demand that evidence before the tool gates anyone, and run your own adverse-impact analysis against the four-fifths rule on pilot data.
• AEDT compliance sits with you, not AWS. NYC Local Law 144 and Illinois HB 3773 are in force now; the Colorado AI Act applies from January 2027 and the EU AI Act’s Annex III high-risk hiring obligations from December 2027. Guard: commission the independent bias audit, publish the candidate notices, and offer a documented alternative path before the first candidate is screened.
• The voice agent is still being refined. AWS has acknowledged the agent does not yet sound convincingly human, and candidate hostility to AI interviewers is well documented. Guard: disclose the AI interviewer up front, track completion and drop-off by candidate segment against your human-screened baseline, and pull the pilot if either moves the wrong way.