Every candidate call answered, every answer recorded, every hiring decision left to a recruiter
A staffing agency's phone rings at the worst possible times. A candidate calls about a job posting while every recruiter is in interviews. A client calls in a job order at 6:45 a.m. because two people quit overnight. A worker calls out sick ninety minutes before a shift, and someone now has to work down a call list before the client notices the gap.
ThunderPhone gives a staffing or recruiting agency an AI phone agent for that call volume. It answers immediately, identifies itself as an AI assistant, asks the exact questions the agency configured, records the answers, and routes the call under the agency's rules. It collects and routes; recruiters decide.
That boundary is the design principle of this entire page. Screening questions—and how their answers are used—are the agency's responsibility, with all the equal-opportunity and applicable-law sensitivity that implies. The agent asks only the question set the agency or its client defined, stores the answers for human review, and never evaluates, scores, ranks, or rejects anyone.
Give every call a consistent, recruiter-defined path
A staffing-agency agent can be configured to:
- identify the agency and disclose its role at the start of the call;
- capture applicant contact details and the position or posting they are calling about;
- ask the agency- or client-defined screening question set—only that set—and record answers verbatim;
- check approved calendar availability and book or confirm interviews;
- run outbound campaigns—shift-fill, confirmations, reminders—under the agency's configured consent rules;
- route existing-worker calls by topic: payroll questions to payroll, site issues to the account manager;
- transfer urgent calls to a recruiter or on-call coordinator, warm or cold; and
- deliver every captured record to the agency's own ATS or systems through signed webhooks, connected APIs, or MCP tools.
Sample application-intake call flow
Agent: Thank you for calling Meridian Staffing. I am the agency's AI assistant. I can take your application details and screening answers for a recruiter to review, and help schedule next steps. Which position are you calling about?
Caller: The warehouse associate job in the Riverside listing.
Agent: Great. May I get your full name and the best phone number and email to reach you?
Caller: Dana Reyes, this number is fine, and my email is dana at example dot com.
Agent: Thank you, Dana. I have a few questions the agency asks every applicant for this role. First: are you able to work an 8-hour standing shift, with or without a reasonable accommodation?
Caller: Yes.
Agent: Do you have reliable transportation to the Riverside area?
Caller: Yes, I drive.
Agent: Last one: which shifts are you available for—day, swing, or overnight?
Caller: Days and swing.
Agent: I've recorded your answers for the recruiting team to review. A recruiter makes all decisions about next steps. Would you like me to book a phone interview from the currently open times?
The agent asked exactly the configured set, confirmed the answers, and handed the record to humans. It offered no opinion on Dana's chances, because it does not have one.
Ten phone workflows a staffing agency can configure
Each workflow is defined by the agency's own script, question set, routing rules, and integrations. The agent executes the steps; the agency keeps every hiring judgment.
Application intake and screening-question collection
When a candidate calls about a posting, the agent captures identity and contact fields, then works through the screening questions the agency or its client defined for that role—availability, shift preference, transportation, certifications held, license numbers to verify, or any custom list. It asks only the configured set, asks it the same way on every call, records answers verbatim, and posts the completed record to the agency's ATS through a signed webhook or connected API. Recruiters review the answers and make every decision about who advances. The agent never screens anyone out.
Interview scheduling and confirmations
Connected to an approved calendar or scheduling API, the agent offers only the interview slots the agency exposes—by recruiter, branch, or role—books the appointment, and reads the confirmed date, time, and format back to the candidate. Outbound confirmation calls the day before can reconfirm attendance, capture a reschedule request, or record a cancellation, so recruiters walk into a day of interviews that will actually happen.
Shift-fill outreach (outbound)
When a client shift opens—a call-out, a surge, a new order—an outbound campaign can call the agency's available-worker list for that role and site, describe the shift using the agency's approved wording, and record each worker's answer. Assignment follows the agency's rules: first confirmed worker wins, the fill is written back to the agency's system immediately so later calls in the campaign are told the shift is taken, and every decline, no-answer, and voicemail outcome is logged for the coordinator. The agent works the list; the agency's rules and system of record decide who gets the shift.
No-show and check-in calls for placed workers
First-day and first-week check-in calls can confirm that a placed worker knows the site address, start time, dress code, and check-in procedure, and capture any problem—transportation, childcare, paperwork—early enough for a coordinator to act. Morning-of confirmation calls surface a brewing no-show while there is still time to backfill; when a worker reports they cannot make it, the agent records the reason, flags the shift, and can trigger the shift-fill workflow above.
Availability updates
Workers between assignments can call any hour to update their availability—days, shifts, sites they can reach, dates away—and the agent writes the update to the agency's system in the agency's defined structure. A periodic outbound refresh campaign can call the bench to keep that data current, so shift-fill campaigns dial people who can actually say yes.
Client job-order intake
When a client calls in an order, the agent collects the fields the agency's order form requires: role, headcount, site, shift pattern, start date, duration, pay rate, and any role-specific requirements the client specifies. Urgent orders—same-day starts, safety-critical coverage—can trigger a warm transfer to the on-call account manager under the agency's escalation rules. Everything else lands in the order queue as a complete, structured record instead of a voicemail that has to be called back and re-asked.
Timesheet-reminder calls (outbound)
Unsubmitted timesheets stall payroll, invoicing, and worker pay. Before each payroll cutoff, an outbound campaign can call workers with missing timesheets, state the deadline, capture a committed submission time or a stated problem, and offer a transfer to payroll when the script cannot answer a question. Outcomes post back to the agency's system, shrinking the coordinator's chase list to the calls that genuinely need a human.
Candidate-status questions routed
"Did I get it?" calls deserve a respectful answer, not an accidental one. The agent shares only the status information the agency has approved for self-service—application received, interview scheduled, decision pending—and routes anything beyond that to the responsible recruiter as a message or callback task. It never speculates about outcomes, timelines, or reasons.
Multilingual candidate communication
Staffing runs on workforces that speak many languages, and an application that dies at a language barrier is a placement lost. The agent supports 47 languages and can switch mid-call when a caller uses one of the additional languages the agency has configured, so a candidate who opens in English and continues in Spanish, Tagalog, or Vietnamese completes the same configured question set without being asked to call back. Answers arrive in the agency's system in the structure the agency defined regardless of the conversation's language. Premium voice or language paths can add up to 3¢ per minute.
After-hours application capture
Candidates apply when their current shift ends—evenings, overnights, weekends. Instead of voicemail, the after-hours agent runs the same intake and screening-question collection the daytime flow uses, so recruiters open the morning with complete, reviewable applications instead of a callback list of half-captured names.
Guardrails a staffing agency should configure
The question set is the boundary
Employment screening is regulated territory. Which questions are asked, how they are worded, and how answers are used are the agency's responsibility under equal-opportunity and other applicable law. The agent's role is narrow by design: ask the configured questions—identically, on every call—record the answers for human review, and take no action on their content. Prompted boundaries can additionally instruct it to decline off-script topics, never characterize a candidate's prospects, and never improvise a new question.
Consent rules for outbound
Shift-fill, confirmation, and reminder campaigns run under the consent rules the agency configures for its worker and candidate lists, using approved wording. Voicemail detection can let the prompt decide, hang up, or leave a predefined message, and campaigns can retry voicemail outcomes under their retry policy.
Structured delivery, verified origin
Captured records reach the agency's ATS or back office through connected APIs, MCP tools, or webhooks signed with HMAC-SHA256, so the receiving system can verify each record genuinely came from the agency's agent. The receiving system determines which fields and actions are available.
Transfers when a human is needed
Cold transfer routes a call immediately. On eligible production VoIP or SIP numbers, a warm transfer lets the agent place the caller on hold, privately brief the recruiter or coordinator, and connect the parties only after acceptance—useful for urgent client orders and escalated worker issues.
Knowledge the agent may use
Attach approved content—branch locations, onboarding requirements, benefits summaries, FAQ documents—as text, Markdown, CSV, PDF, or DOCX files, or import from a URL. The agent searches those sources during calls and escalates when the material does not answer the question.
Test the boundaries before candidates hit them
Before launch, simulate the calls that matter: a candidate who volunteers protected-characteristic information the script must not pursue, a worker disputing pay, a caller pressing for a hiring verdict. Graded scenarios can run as a regression suite so a script change that weakens a boundary fails the test before it reaches a real caller, and A/B experiments can compare script variants on live traffic. Simulations are billable real calls, with the charge shown before a run.
What would this cost each month?
Suppose an agency handles 1,500 AI-assisted calls per month, averaging 4 minutes each:
1,500 calls × 4 minutes = 6,000 minutes per month
- Spark:
6,000 × $0.02 = $120/month - Bolt:
6,000 × $0.05 = $300/month - Storm:
6,000 × $0.09 = $540/month
That is base engine usage, not a quote. Premium voice or language paths can add up to 3¢ per minute, and phone-number or carrier charges depend on the telephony setup. Hold time bills a flat 2¢ per minute; voicemail and phone-menu minutes bill at the engine rate. Optional add-ons such as verbal acknowledgements, call supervision, and long prompts carry their own published surcharges. The agent builder shows the current all-in rate before calls are placed.
Frequently asked questions
Does the AI agent make hiring or screening decisions?
No. The agent asks only the screening questions the agency or its client configured, asks every candidate the same set, and records answers for human review. Evaluating those answers—and every hiring decision—remains with the agency's recruiters.
Who is responsible for what the screening questions ask?
The agency. Screening questions and their use carry equal-opportunity and applicable-law obligations that belong to the employer and agency, not to the phone channel. ThunderPhone executes the configured script; the agency owns its content and its compliance review.
How does shift-fill calling avoid double-booking a shift?
Assignment follows the agency's configured rules—typically first confirmed wins—and each confirmation is written back to the agency's system immediately, so subsequent calls in the campaign are informed the shift is filled. The agency's system of record, not the agent, is the authority on who holds the shift.
Can the agent call candidates and workers proactively?
Yes. Outbound campaigns can place shift-fill, interview-confirmation, check-in, availability-refresh, and timesheet-reminder calls under the consent rules the agency configures, with voicemail handling and retries following the campaign's policy.
What happens when a caller asks something off-script?
Prompted boundaries tell the agent what it may discuss and what it must route. Off-script questions—pay disputes, status verdicts, anything the agency did not approve for self-service—become a transfer, a message, or a callback task for the responsible human.
Can it talk to candidates in languages other than English?
Yes. The agent supports 47 languages and can switch mid-call when a caller uses one of the additional languages the agency has configured. Captured answers arrive in the agency's defined structure regardless of the conversation's language. Premium voice or language paths can add up to 3¢ per minute.
How do captured applications reach our ATS?
Through the integrations the agency approves: signed webhooks (HMAC-SHA256), connected APIs, or MCP servers the agent can call during the conversation. The receiving system verifies the signature and determines which fields and actions are available.
Is ThunderPhone appropriate for handling candidate data?
ThunderPhone is GDPR and HIPAA compliant. The agency remains responsible for configuring its workflows, recordings, retention, integrations, consent rules, and notices to meet its own legal and regulatory obligations, including employment-law obligations attached to screening.