AI VOICE AGENTS BY ROLE
AI voice agents for recruiting calls
In hiring, a voice agent is most useful — and least risky — doing logistics: calling candidates to book or move interviews, reminding them the day before, answering questions about the process and telling them where their application stands. Judging candidates is a different job, and the law treats it very differently.
Every figure below links to its source. This is a summary, not legal advice.
The problem it solves: silence
- In Talent Board's 2023 research (nearly 240,000 candidate surveys), 36% of North American candidates had not heard back 1–2 months after applying, and only 5% remembered being told they didn't get the job.
- SHRM's benchmark puts time to fill at a median of 44 days for non-executive roles and 60 for executive roles (2021 data).
Weeks of waiting, scheduling back-and-forth and unanswered questions are where a phone agent saves time without deciding anything about anyone.
Scheduling and questions vs screening
| The agent… | Examples | Risk profile |
|---|---|---|
| Handles logistics | Books, confirms or moves interviews; reminders; directions; "what happens next?" | A procedural task; no decision about the candidate |
| Collects facts the candidate states | Availability, notice period, work authorisation as stated | Lower, if nothing is scored or filtered automatically |
| Screens or evaluates | Scores answers, ranks or rejects candidates, judges tone or enthusiasm | High-risk in the EU; audits and notices in New York City and Illinois |
The rules
- European Union. The AI Act lists AI used "to analyse and filter job applications, and to evaluate candidates" as high-risk. An Annex III system can fall outside high-risk when it only performs a narrow procedural task — but never when it profiles people, and the provider must document that assessment. The high-risk rules now apply from 2 December 2027 (moved from August 2026). Using AI to infer emotions in the workplace is banned already, and every AI must say it is an AI from August 2026.
- New York City. Local Law 144 requires a yearly bias audit, a published summary and notice to candidates before using an automated employment decision tool that substantially assists or replaces hiring decisions.
- Illinois. Since 1 January 2026 the Human Rights Act prohibits AI that has a discriminatory effect in hiring and requires notice when AI is used (HB 3773); a separate law covers AI analysis of video interviews.
- Colorado. SB 26-189 replaces the earlier AI Act and covers automated decisions about employment from 1 January 2027.
- US federal. The EEOC removed its AI guidance in January 2025, but anti-discrimination law still applies to hiring tools.
How to run it in SAGOT
- Give the agent a logistics job: booking and confirming interviews, answering process questions from approved knowledge, and handing anything else to a recruiter.
- Don't ask it to judge. SAGOT records what was said and the outcome of the call; it doesn't score candidates or read emotions, and you shouldn't use call outcomes to rank people without the assessments the rules above require.
- Connect your scheduling tool so agreed interview times arrive as actions, and set calling hours and attempt limits candidates will find reasonable.
- Every call says it is an AI and that it is recorded or transcribed; if a candidate objects, the agent ends the call, and it offers a recruiter when they ask for a person.
SAGOT is in a closed alpha; new phone lines start in test mode, where only the line's test numbers can be called. Join the waitlist.
Frequently asked questions
Is an AI recruiting call high-risk under the EU AI Act?
AI that filters applications or evaluates candidates is high-risk, with the rules applying from 2 December 2027. An agent limited to scheduling and process questions is a narrow procedural task and likely outside it — document that assessment.
Can an AI agent screen candidates by phone?
It can ask questions, but scoring or filtering candidates triggers the EU high-risk rules, New York City's bias audits and Illinois's notice and anti-discrimination rules. Most teams should start with logistics.
Do I have to tell candidates it's an AI?
Yes: the EU requires it from August 2026, Illinois requires notice when AI is used in hiring decisions, and SAGOT says it on every call.
Can the agent judge a candidate's tone or enthusiasm?
Not in the EU, where inferring emotions in the workplace is banned, and it's a bad idea anywhere. SAGOT doesn't offer it.
Updated 2026-10-16