TL;DR: A standard AI interviewer RFP assumes one employer and one ATS. An agency has several clients, a submittal clock, client-facing evidence and per-req work, so add questions for each. Below are 31 to copy, plus red flags to watch for.
Updated 18 November 2026. ZeaHire is our product; it appears once, near the end.
When you need this
Use this page after you have the AI interviewer RFP template and before you send it to vendors. It suits agency owners and TA leads about to run a vendor call. Add these sections to the base template, then check answers against the red flags at the end.
1. Several clients on one platform
Your recruiters move between clients all day, and each client defines a good candidate differently. A single shared template is the usual failure, along with a permission setting that lets one client's user see another client's people; tools built for one employer often treat multiple clients as an afterthought. Ask the vendor to show the separation in the product, with a restricted user, instead of describing it on a slide, and to put data ownership in the contract.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| Can each client have its own question set and scorecard? | Clients define "good" differently. | Per-client or per-job templates, shown live. |
| How is access separated between clients, recruiters and client users? | One wrong permission exposes another client's candidates. | An editable role matrix and a restricted-user demo. |
| Can a recruiter work across client organizations, and move jobs or candidates between them? | Recruiters rarely serve one client. | The exact steps, and what is kept or removed. |
| What happens when one candidate is screened for two clients? | Duplicate interviews annoy candidates and blur consent. | Clear reuse rules and consent per client. |
| Who owns the candidate data, and can you contact our candidates or offer them to other customers? | Your candidate pool is your business. | A contract clause: your data, no vendor outreach, no reuse. |
2. Submittal speed and the hours candidates apply
Submittal speed is how you win reqs, and candidates often apply evenings, weekends and between shifts. Be wary of a speed claim that quietly depends on a person, on the vendor's side or yours, clicking something first; invites that expire too soon also cost you candidates. Ask for a step list from application to submittal that says which steps run unattended and which wait for a human.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| How do interviews run outside office hours, and does anyone on your side need to be online? | Applications arrive when your team is off. | What runs unattended and what waits for a person. |
| How long does a candidate have to start after an invite, and can you set it per job? | Expiring links decide completion. | Configurable scheduling and expiry. |
| Which steps need a recruiter before a candidate reaches the client? | Hidden manual steps cap your speed. | A step list from application to submittal. |
3. Client-ready evidence
Your client does not want a number, they want a reason to say yes, and your reputation rides on what you send. What goes wrong is a bare score nobody can defend, branding that points at the tool instead of you, or a report that reaches the client before a recruiter has looked at it. The same report also decides whether you can reuse good candidates on the next req. Ask for a real sample, written reasoning per question and a clear approval step.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| What does the client see: scorecard, transcript, recording, summary? | The client judges you on it. | A real sample. |
| Can invites and client reports carry our brand, or our client's? | Candidates and clients should see you, not a tool. | What can be branded, shown on a sample. |
| Can a past interview be reused when the same candidate fits a new req? | Redeployment is where agencies earn back the work. | The reuse rule, how old a result can be, and whose consent is needed. |
| Does each score show written reasoning per question? | A bare score cannot be defended. | Reasoning tied to what a strong answer covers. |
| Does anything reach the client without a recruiter approving it? Is each submission logged? | The agency owns the relationship. | Human approval and a per-job log. |
4. Staffing ATSs and duplicate records
Staffing ATSs are often niche, and a recruiter will not check two systems for every candidate. Watch for a vendor saying yes to your ATS when they mean a different one, or a connection that leaves duplicate candidate and job records behind. Look for a plain yes or no on your system, a named owner and written timeline if it is not supported, and a request to see it demonstrated on your own account.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| Is our ATS supported today? If not, who builds the connection and when? | Many staffing ATSs are niche. | Yes or no, a named owner and a written timeline. |
| What goes back to our ATS after an interview? | Recruiters won't check two systems. | A live demo with your own account. |
| How do you avoid duplicate candidate or job records? | Duplicates damage your database. | The matching rule and what happens on a mismatch. |
5. Recruiter workload and review
Once interviews are running, your recruiters still review evidence, question a ranking and make the call. Per-req economics depend on recruiter hours, and a tool that saves interview time but adds review time just moves the cost. The warning signs are rankings nobody can explain, or a score treated as a decision. Ask for review time measured in a pilot on live reqs rather than an estimate, and for a vendor who treats your recruiter's override as normal and keeps a record of it.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| What does a recruiter review per candidate, and how long does it take in a pilot? | Review time is the real cost. | Measured times, not estimates. |
| Can a recruiter see why a candidate ranked where they did, and override it? | Judgment must stay with your team. | Visible reasoning and a logged override. |
6. Candidate experience at volume
Your candidates are often applying from a phone, between jobs, and a clumsy experience costs you completions, and with them submittals. The candidate also thinks of the interview as your process, so it reflects on you. Trouble looks like an interview that needs an app, reminders you cannot edit, or a dropped connection with no retake. Ask for a mobile flow the vendor has tested, controls you hold yourself, and a plain statement of which languages the interview itself supports.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| Does the interview work in a phone browser, with no app? | Many candidates only have a phone. | A tested mobile flow. |
| Which reminders go out, and can you edit them? | Reminders drive completion. | Configurable wording and timing. |
| Can a candidate retake an interview, and who decides? | Connections fail. | A retake rule you control. |
| Which languages does the interview itself support? | A translated site is not a translated interview. | A language list for the interview. |
7. Compliance for agencies
Agencies sit between employer and candidate, so legal duties when screening with an automated tool can land on you, on your client, on the vendor, or on all three. It matters because a vendor that assumes the buyer will sort it out leaves you holding the risk and without the documents you need. Ask for a written split of duties that matches the contract, and for specific documents rather than reassurance. Treat a general promise of being compliant as no answer.
New York City's Local Law 144, for example, applies to "employers and employment agencies" (NYC DCWP, read 18 Nov 2026); our NYC Local Law 144 page has the detail. Whether you are covered depends on your facts, so check with your legal adviser.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| Which duties does the vendor take, and which stay with us or our client? | Vendors often leave the risk with the buyer. | A written split matching the contract. |
| Can you supply what we need for our own duties: audit data, notice text, retention statements? | You may have to publish or pass these on. | A document list and who owns each. |
| Does any audit or testing cover our mix of clients and roles? | Someone else's data may not fit your pool. | Data used, date and auditor independence. |
The full list, law by law, is in AI hiring compliance questions for vendors.
8. Pricing that fits agency work
Agency demand follows reqs, which open, pause and cancel without warning, and your margin on each req is thin. A pricing unit that sounds simple can hide edge cases, such as retakes or abandoned invites, and a minimum commitment may not flex when a big client pauses. Ask for one named unit, a worked example using your own volume, and terms for scaling down and leaving, with every figure in writing.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| Is pricing per interview, per seat, per req or a mix? | The unit decides who absorbs a spike. | One named unit and a worked example of your volume. |
| What counts as one unit: a started interview, a completed one, a retake, an abandoned invite? | Edge cases change the bill. | A written definition of each. |
| What do we pay when a req is cancelled or a client pauses? | Clients cancel reqs. | Terms for pausing, scaling down and exit. |
9. Pilot on live reqs
A demo is built to impress, while your reqs, clients and ATS are what decide whether the tool fits. What goes wrong is a pilot with vague scope, no baseline and no exit, so you cannot tell whether anything improved. Agree the measures before you start; they are things to measure, not results to expect. Ask for scope, length, support and exit terms in writing, a willingness to use live reqs, and a clear plan for pilot data and candidates if you stop.
| Question | Why it matters | A strong answer includes |
|---|---|---|
| Will you pilot on our live reqs, and on what terms? | A demo is not your workload. | Scope, length, support and exit in writing. |
| Can we track time to submittal, submittal-to-interview ratio and recruiter hours per req? | You need a baseline. | Your current figures first, then the same measures in the pilot. |
| What happens to pilot data and candidates if we don't go ahead? | You still owe those candidates care. | Export, deletion and a date. |
Red flags in vendor answers
- "We can do that" with no live demo on your ATS.
- One shared scorecard for every client.
- A speed claim with no method.
- Scores with no reasons, or a score that sends a candidate to a client with no human in the loop.
- "Compliance is the client's problem" with nothing in writing.
- A vague pricing unit.
- No answer on duplicate records.
How ZeaHire answers these
ZeaHire is our product. Here is how we answer, and where we point you to a demo:
- Several clients: a user can work across more than one organization, and jobs or candidates can move between them. Access is role-based, with an editable permission matrix.
- Evidence: every candidate on a job gets the same questions. Recruiters set what a strong answer covers, and each answer is scored against it with written reasoning. The scorecard has an overall score and recommendation, five skill dimensions with evidence, per-question reasoning and a transcript.
- Candidates: they take the interview from an emailed link scheduled with a date, time and link expiry, with no recruiter on the call. Interviews run 10 to 60 minutes, in English. Candidates can rate the interview afterwards.
- Human decision: recruiters approve, reject and submit to the client. Nothing goes to a client on its own, and submissions are logged per job.
- Panel interviews: a human panel can run the interview in ZeaHire or in Microsoft Teams, with an AI note-taker that suggests follow-up questions.
- ATS: ZeaHire connects with nearly a dozen ATS and HR systems, including Workday, SAP SuccessFactors, Greenhouse, JazzHR and JobDiva. If yours isn't on the list, we'll build the connection. Ask to see it with your own system in a demo. Bullhorn isn't connected today; ask us about building it.
FAQ
Can AI conduct interviews for a staffing agency?
Yes. An AI interviewer asks structured questions and records the answers, with no recruiter on the call. Your recruiters still review the evidence and decide. See how AI interviews work.
Does AI review job applications?
Some tools score each resume against the job. A person should still approve or reject. Ask what the score is based on; see explainable AI hiring.
How do I write an RFP for AI interview software as an agency?
Start from the base template, then add the nine sections above.
Should an RFP include a budget?
Share your expected volume and ask each vendor to price it by their own unit. Add a budget only if procurement requires one.
Should AI be used in the hiring process?
That is your call, and the law may apply to you. Keep a person in every decision and ask for written reasoning on scores. See our volume hiring playbook.
Related reading
- AI interviewer RFP template (the hub)
- AI interviewer security questionnaire
- AI for staffing agencies, JobDiva AI, Bullhorn AI
- Which AI recruiting platform to buy
Comments
No comments yet — start the conversation.