AI interview agents: sixteen you can buy, five that can act without a recruiter

Sixteen AI interview agents run a first-round interview today. Ribbon and HeyMilo lead volume hiring. CodeSignal leads technical roles. Published entry prices run from $79 a month to $499 a month. Five of the sixteen can move or reject a candidate with no recruiter looking first. One only does this for a missed deadline. A sixth vendor’s own pages disagree. Check which setting is on before you sign.

Each of the sixteen gives recruiters a score and a transcript. Every vendor in this market says a person still decides who moves a candidate forward.

16
you can buy today
5
can act alone
0
vendors carry your audit duty

HeyMilo, Humanly, Maki, Apriora and Metaview can act on their own

Apriora puts the usual promise plainly. “No candidate is auto-rejected by AI. Hiring teams can challenge, override, or disregard AI outputs at any point.” HeyMilo says the final decision always stays with a recruiter, not the AI.

Each of these five also has a setting that acts with no recruiter looking first. Someone on your team has to switch it on.

What each one can be set to do

Seven vendors. Indeed’s is still in beta.
  1. HeyMilo A score that can move or archive a candidate in Workday

    “Configure score thresholds in the agent and HeyMilo will request the next-stage move on the Application when a candidate clears the bar, route ambiguous scores to recruiter review, and hold low scores for archive or alternate routing.”

    In plain terms, a high score moves the candidate on and a low score archives them. Its rule builder does the same job in Bullhorn and Avionté. In iCIMS, SAP SuccessFactors and SmartRecruiters, a score only moves a candidate forward. That is six systems in total. UKG Pro does not move candidates.

  2. Humanly A score threshold that decides who goes further

    “Optional score threshold (0–100) applied at enrollment to gate candidates before the first step.”

    In plain terms, you set a cut-off score. A candidate who scores under it hears nothing more from Humanly. No recruiter has to look at the file first.

  3. Maki (Mochi) A filter that can reject a candidate

    “Mochi may automatically exclude candidates only in very specific and objective cases, such as when a work permit is explicitly required and you confirm that you do not hold one.”

    Maki’s product page says no candidate is ever automatically rejected. Both statements are on Maki’s own site. Treat the filter as real. It is a setting you choose.

  4. Apriora A step that books the next interview once a candidate clears your bar

    “When a candidate clears your bar, Alex books them on your team’s calendar instantly, packages a complete briefing with scores and insights, and handles confirmations, reminders, and rescheduling automatically.”

    Apriora says elsewhere that no candidate is ever auto-rejected by AI. That is true of rejection. Booking the next interview is the part that runs with no recruiter involved.

  5. Metaview Screening A setting, off by default, for a missed deadline only

    “You always make the final call — Metaview never auto-rejects anyone unless you’ve explicitly turned on rejection at the deadline, and that’s triggered by a missed deadline, never by a rating.”

    This setting is not about what a candidate said. A recruiter still reads and rates every finished interview. It only fires when a candidate never finishes one at all, and only once someone turns it on.

  6. Tengai One page says it decides alone, others say a person does

    “Psychometric experts tested our framework and validated that Tengai independently can screen and recommend job-seekers, without any human interference.”

    Tengai’s other pages say a person always makes the final call. Get Tengai to state in the contract which one is true.

  7. Indeed Smart Screening The same kind of action, still in beta

    “Employers, and not Indeed, decide their own job requirements, screening and scoring instructions, which tools or features to use (including score-based automations and ATS delivery settings), and which jobs to enroll for Smart Screening.”

    In plain terms, Indeed builds the tool, but every automatic action is one an employer switches on. It is not for sale yet, so it does not count among the sixteen you can buy.

The law looks at how the tool is actually set up. In the UK, the rules on decisions made without a person changed on 5 February 2026. In the EU, the duties for high-risk hiring systems start on 2 December 2027. Reading a candidate’s emotions from their face or voice has been banned in the EU since 2 February 2025.

Three questions for the first demo. Which actions can run without a recruiter looking at that candidate first? Which of them are switched on the day it arrives? Who on your side can switch the others on? Ask to see the settings screen itself.

Two of six hiring jobs are not this market

There are six common reasons to buy an interview agent. For two of them, the better answer is to spend the money elsewhere.

Screen thousands of applicants for hourly roles

Volume product

Most of this market is built for this job. These tools charge by the interview or by the month, not by how many people use them. They run at any hour. They connect to applicant tracking systems such as Workday and Greenhouse. Ribbon, HeyMilo, Apriora, Humanly and Employment Hero all sell for this job.

Run the first round for professional or early-career roles

Enterprise platform

This job has fewer candidates, and each one gets a closer look. Buying one of these usually means a procurement process. HireVue, Eightfold and Talview sell for this job, each as part of a bigger hiring platform. None of the three publishes a price.

Interview for technical roles

Testing platform

CodeSignal, HackerRank, Glider AI and Talview Ivy started as testing tools and now interview candidates too. Talview built exam checks before it built Ivy. If your engineers already run a coding test, the interview can sit next to it. One of the four publishes a price. CodeSignal’s own contract defines the service as screening software developers. Its Grow plan adds interviewers for sales, customer success and marketing roles. Read the contract before you use it outside engineering.

Add a screen inside your tracking or HR system

Sold inside other tools

Greenhouse, Employment Hero and Metaview each sell an interviewer inside a bigger product. Check whether you already pay for one of them. If you do, avoid a second tool that sends its results to the same tracking system.

Make your own interviewers more consistent

Not this market

If eight of your managers each interview a different way, an interview agent will not fix that. It interviews in place of them, before your own people are even involved. What helps instead is a different kind of product. It records the interviews your own people run, then scores them the same way each time.

Ask knockout questions and book the ones who qualify

Not this market

You may only need five fixed questions and a call booked with whoever passes. If so, a tool that holds a full conversation is more than you need. Paradox and Sapia both do this well. Your recruiters write the questions and set the rules. Neither one responds to what a candidate actually says.

Four sell to you this month with no procurement

Two names buyers are often shown cannot be bought yet. Ashby’s interviewer sits in a private beta behind a waitlist. Ashby’s own terms make no promise that it will ever go on sale. They also let Ashby pull it at any time. Indeed calls its version a beta too. It sells it, in its own words, “in limited quantities and locations.”

Ribbon starts at $499 a month for two recruiters and 100 interviews. Employment Hero starts at A$199 a month for 50 interviews. You do not have to move your HR or payroll onto it. CodeSignal starts at $79 a month, for technical roles only. Metaview starts you with 25 free interviews, then sells packs that expire.

Vendors aimed at larger companies price from your own numbers. Eightfold says its interviewer is built for companies of 1,000 to 5,000 people with small hiring teams. Talview usually wants about 150 to 200 interviews committed before it starts. Maki counts more than 80 global enterprises among its customers.

Before you price anything, check what you already pay for. Almost half of these tools are sold inside another product, such as Greenhouse, Employment Hero or a testing platform. If you already run one of those, the interviewer inside it is the cheapest thing to try. It needs no new contract. A couple of weeks will show whether recruiters use it.

The sixteen you can license today · read 19 Sep 2026
Product Sold as Price Independent bias audit What it does with its result
HireVue Platform Ask DCI Consulting, 2023 Scores only
Eightfold Platform Ask NYC Local Law 144, passed Scores only
Talview Ivy Platform Ask · needs about 150–200 interviews to start None published Scores only
Apriora On its own Ask · by headcount Warden AI, monthly Can book the next step alone
HeyMilo On its own Shown at signup · per interview Warden AI, monthly Can act in six systems
Humanly On its own Ask None published Can act on a score
Ribbon On its own Published Holistic AI, annual Scores only
Greenhouse Voice AI Inside Greenhouse Ask · by headcount Warden AI, monthly Scores only
Metaview Screening Inside Metaview Ask · per interview Warden AI, Aug 2026 Only for a missed deadline
Employment Hero Inside its HR system, or alone Published Self-reported only Scores only
CodeSignal Testing platform Published None published Scores only
HackerRank Chakra Testing platform Ask · per interview BABL AI, other HackerRank tools only Scores only
Glider AI Testing platform Ask None published Scores only
Maki Mochi On its own Ask · per interview Holistic AI, Mar 2026 Can act through a filter
Spotted Zebra Inside its assessment platform Ask VerifyWise, Sep 2026 Scores only
Tengai On its own Ask None published Its own pages disagree

Every “Can” or “Only” verdict in that column means someone on your team can switch on an action. The software then takes that action by itself. “Scores only” means no such switch is described in that vendor’s own documents. “Its own pages disagree” is Tengai. Check your own contract either way. A new switch can arrive in a future update.

Ribbon for volume, HeyMilo for hands-off screening, CodeSignal for engineers

A product built for 40,000 hourly applicants and a product built for engineering interviews do different jobs. So the sixteen are grouped here by what they are, not ranked.

Volume hiring, sold on its own

Four tools sell straight into high-volume hiring

Each one connects to a tracking system and charges by the interview or by the month. Ribbon publishes its price, and what it charges once you go over it. Apriora, HeyMilo and Humanly give you a quote based on your headcount or your hiring volume. Apriora, HeyMilo and Humanly can also act on a candidate’s score by themselves, once that setting is on. On Ribbon, a recruiter makes every call.

Sold inside a bigger product

Three interviewers come inside products sold for other jobs

Greenhouse Voice AI works deepest inside Greenhouse Recruiting. It is also sold on its own for companies on other tracking systems. Turning it on starts with a support ticket, not a purchase order. Employment Hero sells its interviewer both inside its HR and payroll subscription and on its own. Metaview sells Screening by the interview, alongside its notetaker.

Ashby AI Interviewer and Indeed Smart Screening also fit this group, but neither is for sale yet.

Enterprise hiring platforms

Bought as part of something larger

HireVue and Eightfold each sell interviewing as part of a bigger platform. Their buyers are companies that already run a procurement process. Expect a scoping call, a security review and a quote, not a price on a page.

Testing platforms that now interview

Four testing tools now interview candidates too

CodeSignal and HackerRank both take the interview out of the same allowance as their tests. If you already hold one of those contracts, a trial can run inside it. Glider sells its interviewer inside a wider skills-testing platform. Talview built exam checks before it built Ivy.

Assessment-led, and European

Three European sellers come at interviews from assessment

Tengai works from Sweden, Spotted Zebra from the UK and Maki from France. All three started in assessment, not in recruiting software. Spotted Zebra publishes a bias audit of its interviewer, done under New York City’s rules. Tengai publishes no audit of its own.

Three of sixteen publish a price

Ribbon, CodeSignal and Employment Hero show their prices. The rest work out a number from your headcount or your hiring volume. Some instead sell you a block of interviews up front, which you then use up over time.

Ribbon $499 a month · 100 interviews this is the entry plan, billed yearly, for two recruiters and two open roles. After the first 100 interviews, each extra one costs $4.00 an interview. The next plans up cost $999 and $1,999 a month · ribbon.ai/pricing · checked 19 Sep 2026
CodeSignal $79 a month · 60 a year, shared with its tests this is the Build plan, billed yearly, for product, design and engineering roles only. The Grow plan costs $479 a month and adds sales, customer success and marketing roles. Interviews come out of the same allowance as its tests · codesignal.com/pricing · checked 19 Sep 2026
Employment Hero A$199 a month · 50 interviews this price excludes GST, the Australian sales tax. It comes to A$2,388 for the year, or A$249 a month if you pay monthly. Every interview past 50 is billed automatically, with no upper limit · employmenthero.com/pricing · checked 19 Sep 2026
Greenhouse Voice AI By headcount this is a yearly subscription priced on your headcount. Your number of interviews is set from what you expect to use · Greenhouse · checked 19 Sep 2026
Apriora By headcount Apriora has no pricing page. Its own ROI calculator says Alex is “priced against headcount, not software” · alex.com · checked 19 Sep 2026
Metaview Screening Per interview, in packs a new account starts with 25 free interviews. After that you buy packs, which expire rather than carry over. You pay once an interview is assessed, not when you test it · Metaview · checked 19 Sep 2026
HeyMilo Monthly plan + interviews you pay a monthly plan, plus interviews, résumé screens and text messages. You buy these up front and use them up over time. Plan prices only appear once you open an account · docs.heymilo.ai · checked 19 Sep 2026
Maki Mochi Per candidate you are charged by the candidate, not by the recruiter. Each of Maki’s agents is switched on and billed separately · makipeople.com · checked 19 Sep 2026
HackerRank Chakra Per interview taken HackerRank explains how it charges but not how much. You pay for each interview a candidate starts · hackerrank.com · checked 19 Sep 2026
Talview Ivy Not published its pricing page is a form that ends in a sales quote. Talview’s own pricing questions describe charging by the unit. They name a minimum of roughly 150 to 200 interviews · talview.com/en/pricing · checked 19 Sep 2026
HireVue · Eightfold · Humanly · Glider · Spotted Zebra · Tengai Not published none of these six publish a price or a rate · checked 19 Sep 2026
Do not fill the gap with a number from a review site. G2 and Capterra show no price for these products. Figures in comparison articles trace back to those same review sites. Take your own monthly applicant count to the vendor. Ask for a quote against that number, and ask what happens in the month you go over it.

Where a tool charges by the interview, the quote shows a rate, not a bill. A role that attracts 2,000 applicants means paying for 2,000 interviews. Your busiest month sets your bill, not your average one.

The audit duty is yours once you sign

The contract sets the audit you publish, how long your data is kept and when it renews.

AuditsEight of the sixteen publish a current bias audit from an outside firm. Apriora, HeyMilo, Greenhouse Voice AI and Metaview Screening all use Warden AI. Ribbon and Maki use Holistic AI. Spotted Zebra uses VerifyWise. Eightfold’s audit runs under New York City’s Local Law 144, and Eightfold says it passed. HireVue says its models get an audit every year, but the one HireVue has published dates from 2023. HackerRank has audited its plagiarism checker and its webcam checker, not Chakra, its interview tool. Employment Hero reports its own testing, not an outside audit. CodeSignal offers a study of its impact during a trial, not a published audit. Humanly, Talview, Tengai and Glider AI publish no audit at all.
The dutyA vendor’s own bias audit helps you meet New York City’s Local Law 144. Under that law, the audit and the candidate notices fall on the employer, not the vendor.
KeepingHow long a vendor keeps your data can differ by years. The contract decides that, not the product. Spotted Zebra’s privacy policy typically keeps a candidate’s data for up to seven years after the assessment. Metaview keeps application data for the length of the hiring process, plus up to 24 months. HeyMilo waits 60 days after you confirm a deletion before the data actually goes. That wait can be set anywhere from 30 to 180 days.
LeavingAfter a Humanly contract ends, you have two days to download what its screens show. You have to ask for it. Ask every vendor what you can take out, in what form, and how many days you get to do it.
RenewalHumanly’s contract renews for another year unless either side gives 90 days’ notice. It also allows a yearly price rise of up to 3% plus inflation. Talview’s terms block any claim brought more than a year after it happened. They also let Talview hand your contract to a buyer of its business without asking you first.
OwnersThree of the sixteen companies have been through a takeover in the past year, as buyer or target. Metaview bought Reval. Reval’s site now sends new customers to a different Metaview product. Humanly bought Qualifi, Sprockets and HourWork in October 2025. Several of the customer results on Humanly’s pages actually belong to those products, not to its interviewer. By March 2026 Findem had said it would buy Glider. Glider’s own site still calls the two companies partners. Get written confirmation of who owns Glider before you sign.

Before you sign, take the interview yourself on a phone with a poor connection. Read what applicants are asked to agree to before it starts.

The hard parts stay yours whether you build or buy

Building this yourself looks possible. You would still buy the technical parts. These are a voice that sounds like a real person, a transcript, and a score set against a scoring guide. The questions you ask and the scoring guide are yours either way. So is what happens to a candidate at the end, and who checks the results a year later.

A smaller, more realistic build sits on top of your own systems. It would be a structured screen that asks your own questions, plus a transcript scored against your own guide. A summary would then go into the candidate’s record.

One thing changes if you build it yourself. The maker’s duties that sit with a vendor today would move to you. Under the EU AI Act, that happens from 2 December 2027. Your duties as the employer stay the same either way.

Here is what building it looks like, agent by agent, with the data each one needs and who decides. The HR Agent Map

Ribbon, HeyMilo or CodeSignal, or nothing at forty screens a month

Sources and method We tested 116 products against one question, between 17 and 19 September 2026: does the software itself hold the interview, two-way, asking open questions and responding within the same session? 38 passed. We researched nineteen of them in full, straight from each vendor's own site, documentation, help centre and legal terms, read directly on 19 September 2026. A separate reader then checked every claim a second time, against the same sources. Prices come from the vendors' own pricing pages. Company facts come from public registers and court records. The law is read on the legislature's or regulator's own text: for New York City, Local Law 144 and the city's own guidance; for the EU, the AI Act and its 2026 amending regulation; for the United Kingdom, the statutory instrument that commenced section 80 of the Data (Use and Access) Act 2025.

Two products a buyer will meet in this market are described here but have no page of their own, because neither can be licensed yet: Ashby's AI Interviewer and Indeed Smart Screening. Paradox and Sapia are named as neighbours rather than as members of this market: both screen with questions your recruiters write, and neither responds to what a candidate says in the moment.