Maki (Mochi)

Mochi is Maki People's screening agent, built by the French company MAKIPEOPLE SAS. It holds a candidate's first interview by phone or browser, then hands your recruiters a transcript and a score. A recruiter decides almost every outcome. The exception is a set of filters your team can switch on, which reject a candidate alone. Maki publishes no price for Mochi. You pay per candidate interaction.

In one line

Mochi holds your first interview call, by phone or browser. It leaves almost every decision to your recruiter. The exception is a set of named filters. Your own team can switch these on to reject a candidate with nobody checking.

Buy it if

You hire in several European languages. You want the first call held without a recruiter picking up the phone. You are also willing to share legal responsibility for candidate data with Maki, not just be its customer.

Do not buy it if

You need a price before you talk to anyone. You need a final answer today on the EU AI Act. You need a fixed rule for how long this product keeps its recordings and transcripts.

Price

Not published. Maki bills you per candidate interaction, not per seat, and each agent, including Mochi, is priced and switched on separately.

Mochi holds your first call. Its filters can reject a candidate alone.

Mochi holds the first call your recruiters would make on every application. It holds that call by phone or in a browser. A candidate picks up the phone or clicks a link. The call runs about ten minutes. It runs longer if you add a skill-based question. Maki recommends no more than two. Mochi asks your questions and follows up when an answer is vague. It does not read from a script. It also adjusts its pace to match the candidate.

Your recruiters pick the questions from a set Maki has already built, called "activities." Your team cannot yet write a new question from scratch. Mochi's AI cannot write one either, though Maki says it is exploring this. You can add your own eligibility and motivation questions. You can also write the answers Mochi gives if a candidate asks about the role. Mochi can hold up to 30 calls at once. Its default languages are English, French, Spanish, German and Italian. Maki adds others on request.

After the call, your recruiters get a score, a transcript and a status. If you added a skill-based question, the score is set against the scoring guide you approved. The transcript can be read or listened back to in full. A separate fraud check also runs underneath the score. It scans the transcript for a scripted or AI-assisted answer. It also looks for a mismatch with the profile the candidate applied under. Your recruiter sees the result as a risk level, with the reasoning behind it. Maki says a flag like this is a reason to look again. It is never, on its own, a reason to reject someone.

Filters your team switches on can reject a candidate. Maki's product page says "no candidate is ever automatically rejected." A score alone never turns anyone away. But Mochi also has "objective filters." Your recruiters switch these on when they build the screening. Examples are a missing work permit, a required certificate or a language level. If a candidate fails one, Mochi excludes them and nobody reviews that decision. Confirm which filters are on before you run a single interview.

Maki also sells a separate feature it calls "Proctoring," across its whole platform. It watches the webcam for a blocked or missing face. It tracks whether a candidate leaves the browser tab. It also listens to gaze and voice for a scripted or fed answer. These checks belong to Maki's longer Assessments product, not to Mochi. Get it in writing whether they also run inside Mochi.

Mochi reads a candidate from a chosen stage in the tracking system you connect. Once your recruiter decides, it sends the result back into SAP SuccessFactors, SmartRecruiters, Workday and iCIMS. Once a candidate clears your bar, Maki's separate scheduling agent, Kumi, takes over. It books the next round straight into Google Calendar or Outlook. Our guide to AI interview agents lists which other products can act on a candidate alone. It also says which ones leave every decision to a person.

Scores read words, candidates consent, and an outside firm audits yearly

The score reads what a candidate said, not how they said it

Mochi scores what a candidate said against a fixed guide for what the role needs. Speech is turned into text before scoring starts. Delivery, pace, accent and speech patterns play no part in the result.

Maki tells recruiters plainly not to rely on the score alone

Maki tells recruiters to "never reject an application based solely on an automated score." It also says a low score "does not automatically disqualify a candidate." Its help centre goes further. It says Mochi "does not make any hiring decisions." It adds that "its sole role is to conduct the interview and collect your answers."

Candidates are told it is AI and give consent first

Mochi tells every candidate, by text or email, that they are talking to AI. It gets their consent first. It avoids sensitive topics such as health and religion. It drops a topic if one comes up by mistake. A candidate who is deaf or hard of hearing can request a text-based interview instead. Anyone can request extra time or another adjustment before the call starts.

An outside firm checks Mochi for bias every year

Holistic AI audits Mochi for bias once a year. It also audits Shiro, Maki’s other screening agent. The audit follows New York City’s Local Law 144. That law requires a bias audit for any tool like this used to screen candidates. The audit is dated March 2026. It covers gender bias, and race and ethnicity bias, both on their own and combined.

Maki shares the legal data duty, and its contract renews itself

Your contract will be with MAKIPEOPLE SAS, a French company registered in Paris. Three co-founders run it, Maxime Legardez, Paul-Louis Caylar and Benjamin Chino. Maki says it raised $28.6 million in a Series A round led by Blossom Capital. It announced that in January 2025. That was on top of an earlier €11 million. No acquisition, no lawsuit and no court case names the company.

Maki and your company are both data controllers

Your data agreement makes Maki and your company two separate "data controllers" over the same candidates. Usually a vendor only processes data for its customer, which carries less duty. A controller carries its own legal duty for how the data is handled. Each side gets its own legal basis for holding a candidate’s data. Each side also agrees to cover the other if something goes wrong. Put this to your own lawyer before you sign.

Its EU AI Act work is not finished yet

Maki says its "conformity assessment is being finalised with an external partner." That is the check that a high-risk AI tool follows the EU AI Act. Maki says it will publish the result. The Act’s high-risk duties for hiring start to apply on 2 December 2027.

The contract renews itself, with no availability promise and a late-payment fee

Your subscription runs for one month or one year. It renews itself unless you give notice. For a monthly plan, give notice ten days before it ends. For a yearly plan, give notice a month before. Maki does not promise the platform will be available. If Maki owes you damages, the most it pays is the fees you paid in the twelve months before. Late payment triggers three things automatically. You are charged interest at three times the French legal rate. Your account is suspended immediately, with no refund. You are also charged a flat fee of 40 euros.

The link to its bias audit results loads empty

Maki’s security page links to a results page for Mochi’s own audit. That page currently loads with nothing on it.

It fits high-volume, multilingual hiring across Europe

It fits roles with more first calls than recruiters can make

One of Maki’s own examples has more than 100,000 applicants a year. A person there could never answer every first call.

It offers named accents within each language

English comes as British, American, Australian or Canadian. French, Portuguese, Arabic and Spanish also come in named forms. Portuguese and Arabic are among the languages Maki adds on request.

Only two customer stories cover Mochi, and neither isolates its result

Capgemini has more than 200,000 employees. Its case study reports ten fewer days to hire and a 96% completion rate. But the case study names both Mochi and Shiro as the agents used. Its results cover both agents together. The Restaurant Group Company has more than 15,000 employees across 300 or more locations. Its case study describes manual screening removed and time given back to hiring managers. No figure is attached to either agent by name. Mochi’s own product page shows two bigger numbers too, a 49% drop in turnover and a cut in screening time. Those numbers come from the Nespresso and Forvis Mazars case studies. Both case studies give the result to Shiro, Maki’s other agent, not to Mochi.

It suits a hiring process you have to defend

A regulator, or a candidate who complains, may ask to see that a person looked before anyone was turned away. Outside its filters, Mochi leaves that decision to a person.

Maki bills per candidate interaction, and names only large customers

Maki asks buyers to request a quote. You pay for each candidate interaction, not for a seat. Each of its agents, Shiro, Mochi, Ken and Kumi, is switched on and billed separately. Mochi itself is sold as one line, called "Mochi Web & Phone." Maki's contract also describes smaller, packaged offers under its negotiated Enterprise tier. Its marketing names only large customers.

Maki states no minimum company size, headcount or hiring volume. It names its own market as "80+ global enterprises, incl. Fortune 500." Every company named on its site has 1,000 employees or more. The smallest named for Mochi itself is The Restaurant Group Company.

Maki's general promise is to keep a candidate's past results for six months. That way, they are not asked to repeat themselves. But its recruiter guide calls that promise "planned but currently not available" for a Mochi phone screening. A candidate can write to Maki and ask it to delete what it holds. Maki can only delete data still inside its own system. A copy in your tracking system is yours to delete, not Maki's. Maki says it does not use candidate interviews to train its models. It otherwise keeps personal data only as long as its purpose needs. It keeps nothing longer than legal, accounting and tax rules require.

You could also build a screen, an interview and a debrief on your own tracking system. We have mapped what that looks like. See it as an agent

Eight questions worth asking before you sign

  1. Which filters can end an application alone, and which are on when Mochi arrives? Who on our side can switch them on? Good answer: None on by default, one named admin, and a record of every exclusion.
  2. Which languages can Mochi hold a live interview in today, not just on request? Good answer: A written list that covers every language we hire in.
  3. Does the link into iCIMS carry Mochi’s own results, or only results from Maki’s other agents? Good answer: Mochi’s own results, on the candidate’s record.
  4. The page you link to for Mochi’s bias audit holds nothing to read. Will you send us the findings? Good answer: Yes, the full Holistic AI report, before we sign.
  5. What happens to Mochi recordings and transcripts after six months? Your guide calls that retention period "planned but currently not available". Good answer: A fixed period in writing, and deletion after it.
  6. Are both the phone and the browser version of Mochi included in what we sign? Good answer: Yes, both, in one agreement.
  7. When will you publish your EU AI Act conformity assessment? Can we see it before we sign? Good answer: A publish date, and the result shared before we sign.
  8. You describe smaller, packaged offers under your Enterprise tier. What sits under the tier you are quoting us? Good answer: A written list of what the tier includes, with its price.

Alternatives to Maki Mochi, including buying nothing

The nearest alternatives to Mochi are HeyMilo, Tengai, Spotted Zebra, Apriora and Metaview Screening. Each holds a first interview with the candidate. They differ in how much of the outcome a setting can decide alone. They also differ in the company behind the product.

Apriora covers the most steps. It screens a résumé, holds the interview, checks for fraud, watches for help and books the next round. Several of these steps can act on a candidate without anyone reviewing that case. Metaview Screening covers the fewest. Its score comes from a transcript alone. Its only automatic step rejects a candidate who misses the deadline. That setting is off by default. Mochi sits between the two. Most of what Mochi produces waits for a recruiter. But its named filters can end an application on their own.

Like Mochi, HeyMilo lets your team set rules that move a candidate on or out, once you switch them on. Tengai and Spotted Zebra both start from an assessment, not from a call inside your tracking system. They are closer to Maki's own Shiro than to Mochi.

Buying nothing is a sixth option. It makes sense if your recruiters already keep pace with first calls. It also makes sense if you hire mostly in one language and one country. If you hire in New York City, the bias audit and notice duty fall on the employer, not the vendor. Maki's own yearly audit covers part of that.

Sources and method Everything said about Mochi here comes from Maki’s own website, its product and integration pages, its Security and Science pages, its help centre articles for Mochi’s candidate and recruiter FAQs and its Phone and Web Screening guides, its legal pages, its funding coverage in TechCrunch, and the French national business register. What Maki publishes about its own product is taken as its description of that product. Every source was read on 19 September 2026, and quotations are verbatim. Where two of Maki’s own pages give different figures for the same thing, the more specific one is used here. The languages are named rather than counted, and the EU AI Act position is the more specific of the two. Both differences are covered in the questions section. Where a difference changes what a recruiter can actually do, both positions are stated in full, because a buyer has to act on both. Two figures on Mochi’s own product page, Nespresso’s 49% lower turnover and Forvis Mazars’ cut in screening time, are drawn from case studies that give them to Shiro, Maki’s separate agent, and are named as Shiro’s here rather than used as evidence for Mochi.

https://www.makipeople.com/agents/mochi · Mochi, the conversational screening agent · read 19 Sep 2026
https://help.makipeople.com/en/articles/414768-recruiter-faq-mochi-ai-screening-agent · Recruiter FAQ, Mochi AI-screening agent, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/414764-candidate-faq-mochi-ai-screening-agent · Candidate FAQ, Mochi AI-screening agent, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/455684-mochi-phone-screening-a-guide · Mochi Phone Screening, a guide, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/456346-mochi-web-screening-a-guide · Mochi Web Screening, a guide, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/659390-how-is-ai-used-in-my-maki-assessment · How is AI used in my Maki assessment, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/654154-responding-to-candidates-asking-about-ai-and-recruitment-decisions · Responding to candidates asking about AI and recruitment decisions, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/227005-how-can-i-request-an-accommodation-for-my-assessment · Requesting an accommodation for an assessment, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/152270-how-long-does-maki-keep-my-assessment-results · How long Maki keeps assessment results, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/167717-how-do-i-request-the-deletion-of-my-personal-data · Requesting deletion of personal data, Maki Help Center · read 19 Sep 2026
https://help.makipeople.com/en/articles/70275-how-long-is-my-data-stored · How long data is stored, Maki Help Center · read 19 Sep 2026
https://www.makipeople.com/security · Maki security and trust page · read 19 Sep 2026
https://www.makipeople.com/science · The science behind Maki’s hiring products · read 19 Sep 2026
https://www.makipeople.com/proctoring · Maki Proctoring · read 19 Sep 2026
https://www.makipeople.com/product-overview · The Maki agent stack, product overview · read 19 Sep 2026
https://www.makipeople.com/integrations · Maki integrations overview · read 19 Sep 2026
https://www.makipeople.com/integrations/sap-successfactors · Maki for SAP SuccessFactors · read 19 Sep 2026
https://www.makipeople.com/integrations/icims · Maki for iCIMS · read 19 Sep 2026
https://www.makipeople.com/integrations/workday · Maki for Workday · read 19 Sep 2026
https://www.makipeople.com/agents/kumi · Kumi, Maki’s interview scheduling agent · read 19 Sep 2026
https://www.makipeople.com/pricing · Maki pricing page · read 19 Sep 2026
https://www.makipeople.com/agents/ken · Ken, Maki’s structured assessment agent · read 19 Sep 2026
https://www.makipeople.com/customers/asos · ASOS case study · read 19 Sep 2026
https://www.makipeople.com/use-cases/frontline · Frontline hiring use case · read 19 Sep 2026
https://www.makipeople.com/customers/capgemini · Capgemini case study · read 19 Sep 2026
https://www.makipeople.com/customers/the-restaurant-group-company · The Restaurant Group Company case study · read 19 Sep 2026
https://www.makipeople.com/customers/nespresso · Nespresso case study · read 19 Sep 2026
https://www.makipeople.com/customers/forvis-mazars · Forvis Mazars case study · read 19 Sep 2026
https://techcrunch.com/2025/01/15/as-gen-z-job-applicants-balloon-companies-are-turning-to-ai-agent-recruiters/ · TechCrunch, on Maki’s funding and founding, 15 Jan 2025 · read 19 Sep 2026
https://www.makipeople.com/company/team · Maki team page · read 19 Sep 2026
https://recherche-entreprises.api.gouv.fr/search?q=899882989 · French national business register, SIREN 899882989 · read 19 Sep 2026
https://www.makipeople.com/legal/terms-of-sale · Maki terms of sale · read 19 Sep 2026
https://www.makipeople.com/legal/terms-of-use · Maki terms of use · read 19 Sep 2026
https://www.makipeople.com/legal/privacy-policy · Maki privacy policy · read 19 Sep 2026
https://www.makipeople.com/legal/transfer-agreement · Maki data transfer agreement · read 19 Sep 2026
https://api.usaspending.gov/api/v2/search/spending_by_award/ · USAspending.gov award search, no contract naming Maki People · read 19 Sep 2026
https://www.courtlistener.com/api/rest/v4/search/?q=%22Maki%20People%22&type=r · CourtListener search, no case naming Maki People · read 19 Sep 2026
https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32026R1744 · When the EU AI Act’s high-risk duties start to apply to employment systems · read 19 Sep 2026
https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page · New York City Department of Consumer and Worker Protection, automated employment decision tools · read 19 Sep 2026