> This is the markdown version of https://www.classet.ai/blog/classet-vs-upwage > Learn more at https://www.classet.ai ![Classet vs Upwage comparison for high-volume frontline hiring](/_next/image?url=%2Fimages%2Fblog%2Fclasset-vs-upwage.png&w=3840&q=75) # Upwage vs Classet: Typed Screening or Phone Interviews? Upwage's AI interview has candidates type their answers in a browser. Classet's Joy calls their cell and talks. What typing costs you in frontline and rural hiring. [![Paul Jones](/_next/image?url=https%3A%2F%2Fassets.basehub.com%2Fe0b5701f%2F6599306507912123f90f150a8bfaaf6c%2Fscreenshot-2026-01-28-at-10.53.16-am.png%3Fwidth%3D100%26height%3D100%26quality%3D100&w=96&q=75) Paul JonesHead of Growth at Classet ](/blog/authors/paul-jones) July 16, 2026 AI Recruiting Ask a warehouse picker to type three paragraphs about their forklift experience on a cracked phone screen during a fifteen-minute break and you'll learn something, but it won't be about forklifts. It'll be about who's willing to type. That's the quiet filter sitting inside every text-based screening tool, and it's the thing that separates Upwage from Classet. Both are AI interviewers built for high-volume frontline hiring. Upwage has the candidate type answers into a browser. Joy calls their cell and has a conversation. Everything else follows from that. **TLDR:** - Both are AI interviewers for high-volume frontline hiring, and both start the interview the moment someone applies - Upwage's interview is text-based: candidates open a link and type their answers, at their own pace, in a browser - Joy texts on apply, asks if now is a good time, and calls the candidate's cell when they say yes. It's a spoken conversation, about five minutes - Typing filters out the people you most want in frontline roles, and a browser needs internet where a call only needs signal - Upwage's AI Analyst scores candidates into a scorecard. Joy writes a summary and stops. No score, no ranking, and a human recruiter makes every call ## What is Upwage? Upwage is an AI interviewing platform for high-volume hourly hiring, founded by Diana Tsai and Greg Call. It pairs two agents: an AI Interviewer that runs structured, STAR-format interviews, and an AI Analyst that scores competencies and produces consistent scorecards. Upwage says the interviewer is trained to be EEOC compliant and can be tuned per role by the recruiting team. The interview is attached to your apply button. A Gopuff recruiter describes it in Upwage's own case study: "The AI interviewer is attached right to our apply button, and it automatically screens everyone who applies." Candidates get a link, work through a structured interview on their own device, and can pause and resume. It fires immediately, which is genuinely fast, and then it waits on the candidate. And it's typed. Every description of the Upwage interview we could find has candidates writing rather than speaking, and their own candidate testimonial sells the typing as a feature. There's no voice option and nothing describing an outbound phone call anywhere in their materials. Upwage sells to enterprise, high-volume employers, with public logos including Gopuff, REI, Jackson Hewitt, MasterCorp, and Carenet Health. Their published claims: more than 15x average proven ROI, 102,000 recruiter hours saved, a 32% average turnover drop, and 1.9x faster interviews. In their Carenet Health case study they report completion reaching 77% against a 30% industry benchmark, and turnover falling from 53.2% to 34.6%. Their CEO has cited a 9.01-out-of-10 satisfaction score across 17,000 responses. Pricing isn't published; you go through a demo. Credit where it's due: that's a stronger enterprise ROI story than most vendors in this category can tell, and the fairness and bias-reduction work is real, not decoration. ## What is Classet? Classet automates phone screening for high-volume frontline hiring. Joy, our AI recruiter, reaches out by text and email the moment someone applies. She introduces herself and the company, says she's AI, and asks whether now is a good time to talk. Candidates can reply, ask questions over SMS, and say yes when they're ready. When they do, Joy calls their cell and runs a structured voice interview, 24/7, in 25+ languages. Anyone who misses the moment can call in later. Nobody gets cold-called by a robot out of nowhere, which is the thing candidates actually resent. Joy screens roughly 150 applicants in about 10 minutes. She asks role-specific questions, answers the candidate's questions about the job, and flags knockout criteria during the call. Afterward she writes transcripts, recordings, and structured summaries into your ATS. Joy Sync connects to 100+ ATS platforms including Greenhouse, Workday, Lever, iCIMS, and BambooHR. Our [analysis of 70,000 interviews](/blog/voice-ai-in-recruiting-insights-from-70000-job-interviews) found 12% more job offers, 18% more job starts, and 17% higher retention when AI handled the first touchpoint. On candidate experience, we collected 10,167 post-interview ratings across nearly 2,000 jobs: [88% rated it positively](/land/candidate-experience) and 64% picked Excellent, the top mark on a four-point scale. We publish the negative 12.5% too, because a number you can't inspect isn't evidence. ## Typing Is a Filter, and It Filters the Wrong People Nobody writes down "must be a confident writer" in the job requirements for a CDL driver. But a typed interview enforces it anyway. Think about who actually applies to your frontline roles. The tech whose hands are covered in grease. The nurse coming off a double. The line cook whose English is excellent out loud and effortful on a keyboard. The 58-year-old warehouse lead with thirty years of experience who has never typed a paragraph in his life and isn't about to start for a job application. None of them are bad hires. All of them are worse _typists_ than the office worker applying to the same req, and a typed interview scores that difference as if it meant something. Voice erases it. People who can't write three paragraphs can absolutely talk for five minutes about work they've done for thirty years, and they'll tell you more doing it. You hear the hesitation on a question they're bluffing, the specificity of someone who's actually done the job, the enthusiasm of someone who wants it. Text flattens all of that into whatever the candidate was willing to thumb-type. It shows up in completion. When one home services customer moved off written assessments to Joy, [completion went from about 50% to over 70%](/blog/enterprise-home-services-scaled-skilled-trades). Their talent acquisition manager put it plainly: "The candidates can just explain in their own words." Their applicant pool never changed. What changed was what the interview asked people to do. To be fair to Upwage, their own candidate testimonial makes the opposite case, and it's a real one: "being able to thoroughly write out my responses helps me better represent my skills." For a nervous candidate with a keyboard and time, that's true. Some people do interview better in writing. The question is what share of your applicants are that person. For an HVAC tech in a truck at 7 PM, it isn't most of them. ## The Second Filter: Internet vs. Signal Typing isn't the only requirement a browser interview quietly adds. It also needs an internet connection, held for the length of the interview. Most applicants have data. In frontline hiring, "most" is the problem, because the ones who don't aren't randomly distributed. They're the applicant on a prepaid plan that's out of data until the 1st, or the tech at a rural job site with one bar and no LTE. Skilled trades work happens where the buildings aren't yet, which is precisely where data is worst. Joy's flow needs none of it. SMS rides the cell network, not the internet, and so does the voice call. A candidate with one bar and zero data gets the text, replies yes, takes the call, and finishes the interview. That same candidate cannot load a browser interview at all. The filters stack. A typed browser interview asks a frontline candidate to be a willing typist, to have a live internet connection, and to carry themselves through the whole thing alone with nobody waiting on them. Plenty mean to and don't, because a shift starts or a kid needs picking up or the tab just gets closed. That's why a self-paced interview and a completed interview are different numbers. Timing compounds it: 55% of frontline candidates apply outside weekday daytime hours, and 88% finish their screen within 20 minutes when it fires instantly. As one customer told us: "If candidates have applied to 10 companies, they're only going to talk to the first two that get back to them. We wanted to be that first one." Feature Upwage Classet Interview format Typed. Link attached to the apply button, completed in a browser on any device Spoken. SMS and email on apply, then a voice call to the candidate's cell once they say yes. They can also call in What the candidate does Types answers to structured questions Talks for about five minutes Connectivity required An internet connection held for the length of the interview Cell signal only. SMS and voice ride the cell network, so no data or wi-fi Pacing Self-paced, with pause and resume Live, and 88% finish within 20 minutes of applying Interview method STAR-format structured interviewing, EEOC-trained, per-role tuning Structured role-specific questions, real-time knockout flagging, and Joy answers candidate questions about the job What comes out AI Analyst scores competencies into a consistent scorecard with supporting evidence Transcript, recording, and structured summary in the ATS. No score, no ranking Languages Multilingual, any language 25+ languages, spoken ATS integration Ashby, Greenhouse, iCIMS, Lever, SmartRecruiters, Workday 100+ platforms, deployed in days without engineering Published proof 15x+ ROI, 102K recruiter hours saved, −32% turnover, 77% completion at Carenet, 9.01/10 across 17,000 responses 70,000-interview analysis (12% more offers, 18% more starts, 17% higher retention), 88% positive across 10,167 ratings, 70%+ completion Who decides Recruiting team sets the criteria the Analyst scores against A recruiter makes every advance-or-reject call. Knockouts route to a stage you choose Buying motion Sales-led, pricing not published Standalone plans start self-serve and month-to-month, Joy Sync for larger deployments ## What Each Tool Does With Its Own Output The interview is only half of what you're buying. The other half is what the system hands your recruiter afterward, and the two products diverge sharply here. Upwage's AI Analyst scores competencies against criteria your recruiting team defines and produces a consistent scorecard with supporting quotes. That has real upside: consistency is the thing human screeners are worst at, and a scorecard makes candidates comparable in a way that a pile of recruiter notes never will. Joy doesn't do that. She writes a transcript, a recording, and a structured summary into your ATS, and stops. No score. No ranking. No sorted pile with the good ones floated to the top. A recruiter reads what the candidate actually said and makes the call. Even knockouts, the hard yes-or-no criteria you define yourself, only route someone to a stage you picked. That choice costs us something, because a ranked list is faster to work through than fifty summaries and some buyers want exactly that. We don't build it, because the moment a tool ranks the pile, the ranking becomes the decision. Whatever the policy says about human review, the person at the top gets called and the person at the bottom does not. We'd rather the judgment stay where the accountability already is. Upwage's approach isn't careless, it's just a different answer, and it's worth choosing deliberately rather than discovering after rollout. ## Where Upwage's Numbers Deserve a Closer Read Two things get repeated about Upwage that are worth correcting, and I'd rather correct them in their favor than run a comparison on a sloppy version of their story. The "77% completion rate" that circulates on aggregator sites isn't a platform-wide number; it comes from their Carenet Health case study, one customer. Their own framing is careful about this, the echo isn't. Same with the 9.01 score, which several sites report as an NPS. It isn't one. NPS runs from −100 to +100. It's a 10-point satisfaction score, and Upwage's CEO describes it accurately. Their turnover work is the strongest thing they've published. Cutting a call center from 53% to 35% is a hard, real result, and if retention prediction is your specific problem, they've done more public work on it than we have. One note on the evidence base: Upwage's G2 presence is roughly four reviews. That's not a knock on the product, since a small count mostly means a company sells through a sales team rather than review sites. It does mean third-party proof is thin, so you'll be leaning on their case studies and your own pilot. ## Which One Should You Pick? Pick Upwage if your applicants are comfortable writing, if enterprise ROI and turnover modeling is what you need to defend the purchase internally, and if you want candidates scored and made comparable at scale. That's their design, not an accident, and their scorecard consistency and retention-prediction work are legitimate strengths. I'd rather you buy the right tool than ours. Pick Classet if you want three things: applicants who finish the interview instead of stalling partway, the answers landing in your ATS without anyone rekeying them, and a recruiter holding enough real information to decide who moves forward. That's the whole product. Joy reaches applicants the moment they apply, gets the conversation done in about five minutes, and writes the transcript, recording, and summary straight into the ATS you already run. Your recruiter opens the record and decides with the candidate's own words in front of them, not a number. That works across frontline hiring generally, which is where most of our volume is: retail, healthcare, hospitality, logistics, manufacturing, staffing, and the trades. Anywhere you're hiring at volume and losing people between "applied" and "actually talked to someone," that's the gap Joy closes. It just happens to be _unbeatable_ at the hard end of the range, with applicants in trucks, on ladders, between routes, or on prepaid data that runs out before the month does. Those are the candidates a typed interview quietly loses, and they cost you nothing extra to reach here. Procurement is different too. Classet's standalone plan starts self-serve and month-to-month, so you can run both side by side on live requisitions before anyone signs anything. Upwage, like most enterprise platforms in this category, goes through sales first. ## FAQ What's the main difference between Upwage and Classet? Upwage's AI interview is typed: candidates open a link and write their answers in a browser, at their own pace. Classet's Joy texts the candidate when they apply, asks if now is a good time, and calls their cell once they say yes, then runs a spoken interview of about five minutes. Typing versus talking is the difference, and it drives everything else, from who finishes to what signal you get back. Does Upwage do voice or phone interviews? No. Upwage's interview is typed: candidates open a link and write their answers in a browser. Their own candidate testimonial presents the writing as the benefit. There's no voice option and nothing describing an outbound phone call in their materials. Classet is the opposite by design, with Joy calling the candidate's cell and running a spoken conversation. Why does typing matter for frontline hiring? Because it filters for a skill the job doesn't require. Techs with dirty hands, nurses coming off a double, candidates whose English is stronger spoken than written, and experienced workers who simply don't type are all worse at a written interview than an office applicant, and none of that predicts how well they'll do the job. Voice removes the filter, which is why one home services customer's completion rose from about 50% to over 70% after moving off written assessments. Can candidates interview without wi-fi or mobile data? With Classet, yes. SMS and voice both ride the cell network rather than the internet, so a candidate with signal but no data can get the text, reply, take the call, and finish the screen. A browser interview needs a live connection throughout, so that same candidate can't start one. It's the practical reason phone screening holds up in rural markets and on job sites where coverage is thin. Is Upwage's 77% completion rate a fair comparison to Classet's 70%? Not directly, and we'd rather say so. Upwage's 77% figure comes from a single case study, Carenet Health, not a platform average. Classet's 70%+ figure comes from one home services customer's move off written assessments. Both are single-customer results measured on different funnels. Treat either as a directional signal and run your own pilot on live requisitions. Does the AI decide who gets hired? Not at Classet. Joy runs the interview and writes a transcript, recording, and structured summary into your ATS. She doesn't score candidates or rank them against each other. A recruiter reads the summary and makes every advance-or-reject call, and even knockouts only route a candidate to a stage you choose. Upwage's model differs by design: their AI Analyst scores competencies against criteria your team sets and produces a scorecard. Evaluating either of us, the question worth asking is what the tool does with its own output, and where a human enters the process. ## Key Points - Upwage's interview is typed in a browser. Joy's is spoken on a phone call. Both fire the moment someone applies - Typing filters for writing confidence, which frontline roles don't require and frontline candidates are least likely to have - A browser interview also needs live internet. SMS and voice ride the cell network, so Joy works on signal alone - Upwage's Analyst scores and ranks against your criteria. Joy hands a recruiter a summary and no score, so the judgment stays with a person - Upwage's enterprise ROI and turnover results are real, and their 77% completion figure is one customer's, not a platform average ## Next Steps The fastest way to settle this is to stop reading comparisons and listen to one. [Hear a real Classet phone interview](/demo), or run Joy against your own live requisitions for a month and watch what happens to completion. [![Paul Jones](/_next/image?url=https%3A%2F%2Fassets.basehub.com%2Fe0b5701f%2F6599306507912123f90f150a8bfaaf6c%2Fscreenshot-2026-01-28-at-10.53.16-am.png%3Fwidth%3D100%26height%3D100%26quality%3D100&w=128&q=75) Paul Jones Head of Growth at Classet Paul comes from an operator background running an Alpine-owned company, and brings firsthand experience with the hiring challenges Classet was built to solve. He's driven by a belief that the right technology can make meaningful work more accessible. ](/blog/authors/paul-jones) Follow Classet in Google Add classet.ai as a preferred source and our hiring research is more likely to show up in your Top Stories and AI Overviews. [Add as preferred source →](https://www.google.com/preferences/source?q=classet.ai) ## Explore More ### Use Cases - [RPO / BPO Recruiting](/use-cases/call-centers-bpo) - [Healthcare Recruiting](/use-cases/healthcare) - [Hospitality Recruiting](/use-cases/hospitality) ### Integrations - [Greenhouse](/integrations/greenhouse) - [Bullhorn](/integrations/bullhorn) - [Lever](/integrations/lever)