The platform
Five agents. One candidate record.
Each agent owns a stage of the funnel and writes back to the same profile, so nothing is re-asked and nothing is lost between channels. Below is what each one actually does, in the order a live requisition hits them.
The run
What happens when a requisition opens.
Rediscovery first, because the cheapest qualified candidate is usually one who already applied to you.
- 01
Rediscover
TalentGraph
Before a dollar goes to job boards, TalentGraph scores every profile you already own against the new requisition. Recruiters get a ranked internal shortlist, with availability re-confirmed by an automated check-in, usually on the day the role opens.
- 02
Source
Beacon Sourcing
Beacon turns the requisition into live queries and works job boards, public profiles, and community sites continuously. New matches are deduplicated against your database, then queued for outreach with a first message written from their actual background.
- 03
Qualify
Signal Screening
Everyone who applies or replies enters a WhatsApp or SMS thread that works the knockout criteria in their own language. Qualified candidates self-schedule on the spot, and the rest get a clear reason plus an opt-in to stay in TalentGraph.
- 04
Connect
Voice Outreach
Candidates who prefer a phone call, or who go quiet on chat, get a live AI call that confirms interest, compensation, and timing. Anything outside the script transfers warm to a human recruiter on the same call.
- 05
Interview and score
Nova Recruiter
The structured interview runs on the candidate's schedule and produces a transcript scored against the requisition rubric. Recruiters review ranked results with the supporting evidence attached, then make the hiring call themselves.
01 — Interviewing
Nova Recruiter
Structured AI interviews, generated from any job description.
Paste a job description and Nova Recruiter parses it into weighted competencies, then writes a question set mapped to each one with follow-up branches for thin answers. Candidates join a browser-based session with no install, and the agent probes in real time when a response lacks evidence. Every answer is transcribed, scored against a fixed rubric, and stored next to the timestamped clip that produced the score. Recruiters review a ranked shortlist with the rubric visible, and any score can be overridden with a written reason that lands in the audit log.
- Parses job descriptions into weighted competency rubrics automatically
- Asks adaptive follow-ups when answers lack concrete evidence
- Timestamps every score to the clip that justifies it
- Runs unattended interviews in 32 languages, day or night
02 — Conversational Screening
Signal Screening
Conversational pre-qualification at volume, on WhatsApp and SMS.
Every requisition carries a knockout set: work authorization, shift availability, location, licenses, compensation band. Signal Screening asks them one at a time in the candidate's own language over WhatsApp or SMS, reads free-text replies instead of forcing menu choices, and re-asks when an answer is ambiguous. Qualified candidates get a self-scheduling link on the spot and land directly in a recruiter's calendar, while disqualified candidates get a clear reason and an opt-in to stay in TalentGraph. Median first response is under 30 seconds, which is what stops hourly and high-volume pipelines from leaking overnight.
- Reads free-text replies, no rigid menu trees
- Handles knockouts: authorization, shift, license, pay band
- Books interviews into recruiter calendars through self-scheduling links
- Re-engages silent applicants on a timed nudge sequence
03 — Voice Automation
Voice Outreach
An AI phone agent that calls and answers.
The agent dials from a local area code, opens by disclosing that it is an AI recruiter, and holds a full duplex conversation with barge-in, so candidates can interrupt the way they would with a person. It works a call list against time zone and recency rules, retries no-answers on a decaying schedule, and leaves a voicemail with a callback number that returns to the same conversation state. Inbound calls reach the same agent, which answers requisition questions, confirms or moves an interview, and transfers warm to a recruiter mid-call when the candidate asks for one. Every call is recorded with consent captured up front, then transcribed and summarized into the candidate record.
- Full duplex speech with barge-in and sub-second latency
- Discloses AI status and captures consent before recording
- Warm transfers to a live recruiter mid-conversation
- Retries no-answers on decaying, time zone aware schedules
04 — Candidate Database
TalentGraph
Every candidate you have touched, made searchable again.
TalentGraph ingests resumes, interview transcripts, call summaries, and chat threads, then normalizes skills and titles into a shared taxonomy so RN and registered nurse resolve to the same node. Profiles are stored as embeddings alongside structured fields, so a search returns semantic matches rather than keyword hits alone. When a requisition opens, the rediscovery engine scores your existing database against it first and hands recruiters a ranked internal shortlist within minutes. Profiles refresh themselves: the platform re-contacts quiet candidates through Signal Screening or Voice Outreach to confirm availability, title, and compensation before anyone reaches out cold.
- Normalizes skills and titles into one shared taxonomy
- Semantic search over embeddings, not keyword matching alone
- Scores the existing database before any new sourcing spend
- Auto-refreshes stale profiles through chat and voice check-ins
05 — Outbound Sourcing
Beacon Sourcing
Outbound sourcing across job boards and the open web.
Beacon reads the requisition, infers the boolean and semantic queries a senior sourcer would write, and runs them continuously across licensed job board feeds, public profiles, portfolios, and community sites. Candidates are deduplicated against TalentGraph before contact, so nobody already in your pipeline gets a cold message about a role they applied for last month. Outreach is drafted per candidate from specifics in their profile, sent by email or InMail, and followed up twice on a cadence tuned to reply behavior in that market. Replies route into Signal Screening automatically, so a sourced candidate can be qualified and booked without a recruiter opening the thread.
- Generates boolean and semantic queries from the requisition
- Dedupes against TalentGraph before any message is sent
- Drafts outreach from specifics in each candidate profile
- Hands replies straight to Signal Screening for qualification
The output
What a recruiter actually opens.
Not a black-box score. A rank, the rubric behind it, the clip that produced each line, and an override button that writes a reason into the audit log.
Thirty-two languages, one pipeline
Every agent screens, calls, and interviews in 32 languages, including regional Spanish, Tagalog, and Portuguese variants. Scores stay comparable across languages because the rubric is applied to the meaning of an answer, not its wording.
Rubric scoring on every answer
Each response is scored against a fixed, requisition-specific rubric with the supporting clip or transcript line attached. Two candidates for the same role are always evaluated on the same criteria, in the same order.
Bias controls built in
Demographic proxies are stripped before scoring, models are audited for adverse impact, and every automated decision carries a written reason. The platform produces the disclosure and audit artifacts expected under NYC Local Law 144 and the EU AI Act.
Omnichannel by default
Chat, SMS, voice, email, and video run off one candidate record and one shared conversation state. A candidate can start on WhatsApp, finish on a phone call, and never repeat themselves.
Rediscovery before net-new sourcing
TalentGraph searches the candidates you already paid to acquire before Beacon spends anything on new sourcing. Across our high-volume clients, roughly two in five hires come from profiles that were already in the system.
Automated pipeline, human decision
The five agents run sourcing, screening, outreach, interviewing, and scoring without a recruiter working the queue. The hiring decision is never automated: a Novexhire recruiter reviews the ranked shortlist and signs off before a candidate moves forward.
Rank 1 of 214 screened
Service Desk Analyst, Tier 2
Bilingual EN/ES · Nights · Available in 2 weeks
89
rubric score
- Technical depth
- 92
- Communication
- 88
- Role fit
- 79
- Availability
- 96
Trail
- WhatsAppSignal ScreeningPre-qualified in 4 messages09:12
- VoiceVoice OutreachConfirmed shift and rate09:31
- VideoNova RecruiterStructured interview, 11 questions10:04
Awaiting recruiter sign-off. No candidate advances on a model score alone.
Integrations
Your ATS stays the system of record.
Two-way sync on candidates, stages and notes. Novexhire does the screening work and writes the result back where your team already looks.
Workday
HRIS
Greenhouse
ATS
Lever
ATS
SAP SuccessFactors
HCM
BambooHR
HRIS
Slack
Team messaging
Microsoft Teams
Team messaging
WhatsApp Business Platform
Candidate messaging
Twilio
SMS and voice
Okta
Identity and SSO
Zapier
Automation
Google Calendar
Scheduling
Indeed
Job boards
LinkedIn
Sourcing
Named third-party products describe interoperability only, not endorsement. Anything not listed connects through the REST API, webhooks or Zapier.
Ask about your stackStart with one role
Your first shortlist in 72 hours
Send us one open requisition. We run sourcing, screening, and AI interviews on it, then hand back a ranked shortlist with scorecards and interview recordings.
Or call +1 (512) 555-0142 — Austin, TX