If you applied to a Fortune 500 job this quarter, odds are the first thing that touched your resume was not a recruiter. It was HiredScore, the AI grader Workday now bakes into its recruiting suite, and it decided within seconds whether you were an A, B, C, or D. General Motors used that grader to cut candidate screening time by 70%, and with over 60% of the Fortune 500 running Workday, this is the default filter most applicants now hit.
What the Workday Recruiter Agent actually does
The Workday Recruiter Agent is an AI layer that ingests every applicant, matches them against the job description, and hands the recruiter a pre-sorted stack labeled A, B, C, or D. It is not one model. It is three agents stitched together: Workday Recruiting (the ATS), HiredScore AI (the grader), and Workday Paradox (the Olivia chatbot that runs conversational pre-screens).
Workday was named a Leader in the May 8, 2026 Gartner Magic Quadrant for Talent Acquisition Recruiting Suites, and the announcement leaned hard on this stack. Adam Godson, the ex-Paradox CEO who now runs Workday's entire TA platform, has been the public face of the pitch. Josh Bersin's July 15, 2026 note "Multi-Agent AI for Talent Acquisition Arrives" put a name on what is happening: the multi-agent TA stack is now the default, not the experiment.
For an applicant, the practical translation is this:
- Your resume gets a letter grade before a human touches it.
- Recruiters are trained by HiredScore's own docs to sort by grade, not by application date.
- HiredScore states there is a "clear correlation" between A/B grades and reaching the interview stage.
- C and D grades functionally do not get seen when a role has 300+ applicants.
Disclosed at Workday's May 2025 Illuminate keynote; the flagship HiredScore case study through 2026.
Why the 70% number at GM matters for your resume
The 70% figure is time saved by recruiters, not applicants cut. But the applicant experience is functionally identical: with 300+ applications per role (Ashby's 2026 Talent Trends benchmark, via Jobscan), a C or D grade is a rejection reached in seconds instead of days.
GM's Head of Candidate Experience Eileen Kovalsky has walked through the deployment publicly with Workday and HiredScore. The mechanic she describes is not "the AI rejects you." It is "the AI sorts you," and the recruiter works top-down until the requisition closes. AdventHealth, the companion HiredScore case study, reported a $68M reduction in external-agency spend over 12 months from the same setup. Neither company is going back.
The honest read: HiredScore's grade is not a literal auto-reject in the software. It is a de facto one, produced by sort order plus math.
The A/B/C/D grade, in plain English
HiredScore assigns each applicant an A, B, C, or D based on how the resume matches the skills, education, and experience in the job description. HiredScore's own help center is explicit that recruiters should sort by grade, and that A/B candidates advance at materially higher rates than alphabetical sorting would produce.
Here is the grade-to-outcome logic, as HiredScore documents it and as recruiter behavior confirms:
| Grade | What it means | Practical outcome at 300+ apps |
|---|---|---|
| A | Strong skills, education, and experience match | Reviewed first, likely interview |
| B | Solid match with some gaps | Reviewed if A pool doesn't fill the slate |
| C | Partial match, notable gaps | Rarely reviewed |
| D | Weak match | Effectively invisible |
| Ungraded | CV missing or in unsupported format | Sorts below D. Silent auto-reject. |
That last row is the one nobody warns you about. HiredScore's docs confirm candidates are not graded at all if the CV is in an unsupported format. A Canva-designed PDF with image layers, a resume exported as a flattened image, or a file the parser cannot read produces zero grade and drops below every C and D. This is mechanical, not judgmental. Fix the file, or you never had a chance.
The real filter isn't keywords, it's inferred skills
HiredScore does not just match the words on your resume to the words in the JD. It infers skills you did not list. Jobscan's July 10, 2026 breakdown of the AI resume screening market names the specific example: HiredScore can infer C++ proficiency from embedded-systems experience, even if you never typed "C++."
This flips a decade of ATS resume advice on its head. Keyword-stuffing a skills bar still helps a little, but it caps out fast. What actually raises your grade is dense, specific project descriptions that give the model adjacencies to infer from. The grader is scoring skill coverage, and coverage rises fastest when you write bullets that reveal the stack around your work, not just the work itself.
Two versions of the same bullet, same person, same job:
- Weak: "Built firmware for automotive ECUs."
- Strong: "Shipped firmware for 3 automotive ECU platforms in a mixed C/C++ codebase on ARM Cortex-M, integrating CAN and LIN buses and running HIL tests through Vector CANoe."
The strong version gives HiredScore ten inferable skills the weak one hides: C, C++, ARM, embedded systems, CAN, LIN, HIL testing, Vector toolchain, automotive protocols, firmware release process. You did not stuff keywords. You described the work honestly, at the resolution the grader reads at.
Rewriting every bullet at that resolution for every posting is the exact work Refolk takes off you: paste the job description, and Refolk pulls the specific adjacent skills the posting implies, then rewrites your history to surface the ones you actually have.
Fortune 500 ATS screening: who is actually behind the screen
Almost nobody who understands the tool. In Refolk's index of professional profiles, only 151 U.S. recruiters and TA specialists list Workday Recruiting or HiredScore as a declared skill, out of 117,790 U.S. recruiter, TA partner, and TA specialist profiles total. That is 0.13% of the human population you are trying to reach.
151 out of 117,790 profiles in Refolk's index. The rest are trusting the A/B sort.
The AI-literacy number is worse. Only 13 U.S. recruiters in Refolk's index list both AI or prompt-engineering skills AND a recruiter title. Against roughly 118,000 recruiters. The human on the other side of your application is not overriding the algorithm. They are working the sorted stack.
Where the tool is spreading, based on the top employer signals for Workday/HiredScore-skilled recruiters in Refolk's index:
| Sector | Named employers |
|---|---|
| Aerospace / defense | Boeing (3 recruiters) |
| Banking | Wells Fargo (2), Citizens Bank |
| Industrials / life sciences | Thermo Fisher, Medtronic, Valvoline |
| Consulting / services | Accenture, JLL, Thomson Reuters, Capita |
| Retail / consumer | Talbots, Alo Yoga, Hungry Jack's |
| AI-native tech | Cohere |
Workday's own May 13, 2026 press release named JLL, Medtronic, Thomson Reuters, Valvoline, Capita, Signify, and Hungry Jack's as recruiting-suite reference customers. Bersin's April 2026 piece put a price tag on the buildout: "almost $3 billion" spent acquiring HiredScore, Evisort, Paradox, and Sana. This is not a pilot. It is the platform.
How to write a resume that earns an A
Write for skill coverage and inferred adjacencies, not keyword matches. Every bullet should reveal the stack, tools, scale, and outcomes around the work so the grader can infer three to five related skills per line.
The mechanical fixes first, because these are the silent kill switches:
- File format. Submit a text-based PDF or .docx. Not an image PDF, not a scanned resume, not a Canva export with flattened graphics. If a parser cannot read it, HiredScore does not grade it, and you sort below D.
- One column, no text boxes. Multi-column resumes and sidebars scramble parsing order. The grader may read your job titles as skills and your skills as employers.
- Standard section headings. "Experience," "Education," "Skills." Cute headers like "Where I've Made Dents" get skipped.
- Dates on every role. Missing dates lower confidence in the experience match.
Then the content fixes:
- Mirror the JD's skill vocabulary once, verbatim, in a Skills section. This locks in literal matches.
- Rewrite every bullet to name the tools, protocols, scale, and outcome. This drives inferred matches, which is where A/B grades are actually won.
- Reorder bullets per posting so the top three lines under each job map directly to the top three requirements in the JD.
- Cut anything older than 10 years unless it is a name-brand employer or a directly relevant skill anchor. Older content dilutes coverage density.
- Add a two-line summary at the top that names your function, years, top three tools, and the domain.
Doing all five for every posting, by hand, is the reason most people apply to 20 roles and rewrite for maybe two of them. Refolk drafts each tailored resume and cover letter from your own history the moment you paste a posting, and scores how well you actually fit before you hit send, so you are not spraying C-grade applications into the Workday funnel.
The auto-reject isn't literal in the software. The sort order and 300 apps per role make it de facto true.
The second gate: Paradox, after HiredScore
Clearing HiredScore is not the finish line. Workday's suite officially unifies three agents in one funnel: Workday Recruiting, HiredScore AI for Recruiting, and Workday Paradox. Paradox (the Olivia chatbot Workday acquired) runs conversational pre-screens after the HiredScore sort.
A resume that earns an A can still be knocked out by Paradox's pre-screen questions: work authorization, availability, compensation range, location, must-have credentials. Treat the application as a two-stage gate:
- Stage 1 (HiredScore): Written for skill coverage and inferred adjacencies. Grade determines whether Stage 2 happens.
- Stage 2 (Paradox): Answered honestly, but with the JD's must-haves in mind. Paradox's knockout questions are binary. One wrong answer ends the pipeline.
The Bersin framing to keep in mind: HiredScore is the grader, Paradox is the interviewer, Workday Recruiting is the file cabinet. You are being read by three different agents with three different scoring functions, and the human recruiter sees the composite. Optimize for all three, or optimize for none.
FAQ
Does HiredScore literally reject my application?
No, not in the software. HiredScore assigns a grade and sorts. What produces the de facto rejection is the combination of that sort with 300-plus applications per role. Recruiters work the A pool first, then B, and most requisitions close before C and D are ever opened. The one exception is silent: if your resume is in an unsupported format, HiredScore does not grade it at all, and you sort below every D.
How do I know if a company I'm applying to uses HiredScore?
Look at the application URL. If it contains "myworkdayjobs.com" or is hosted on a Workday careers subdomain, you are almost certainly in the Workday suite, and HiredScore is likely the grading layer if the employer is a Fortune 500. Workday states over 60% of the Fortune 500 uses its talent management platform, and named reference customers in May 2026 include JLL, Medtronic, Thomson Reuters, Valvoline, Capita, Signify, and Hungry Jack's.
Should I still tailor my resume by hand for every posting?
If you have the hours, yes. Realistically, tailoring 20 applications by hand costs 15 to 20 hours a week, and the win comes from rewriting bullets to name tools, protocols, and scale, not from tweaking the summary. This is where Refolk earns its keep: it rewrites your resume from your own history for each posting, drafts the cover letter, and scores your fit before you submit.
Will a plain-text resume beat a designed one?
For Fortune 500 ATS screening, yes. A clean, single-column, text-based PDF or .docx with standard headings will parse cleanly into HiredScore every time. Designed resumes with sidebars, icons, custom fonts, or image layers risk partial parsing or the ungraded state. Save the designed version for the human interview stage. The grader wants text it can read.