The Hidden Workers ATS Problem in 2026: Qualified Talent Your Filters Reject

What the Hidden Workers research found

Most employers already know their screening software rejects people who could do the job. That's the uncomfortable core of Hidden Workers: Untapped Talent, the September 2021 report from Harvard Business School's Project on Managing the Future of Work and Accenture. The authors, Joseph Fuller and Manjari Raman at HBS with Eva Sage-Gavin and Kristen Hines at Accenture, surveyed more than 8,000 hidden workers and more than 2,250 executives across the US, the UK and Germany.

The headline numbers are blunt. 88% of employers agreed that qualified high-skills candidates get vetted out because they don't match the exact criteria in the job description. For middle-skills roles the figure rose to 94%. The researchers estimated more than 27 million hidden workers in the US alone, with similar proportions in the UK and Germany: people who want a job, or more hours, but whom standard hiring processes keep overlooking.

Who are they? Caregivers coming back after a break. Veterans whose skills don't carry civilian credential names. Immigrants and refugees, people with disabilities or mental health conditions, people without a degree, people who were once incarcerated, and partners who relocated for a spouse's job. The report sorts them into three groups: "missing hours" (part-timers who want full-time work), "missing from work" (long-term unemployed and still looking) and "missing from the workforce" (not looking, but able and willing under the right conditions).

The part that should sting for an HR leader sits in the same survey. More than 90% of the employers surveyed said they use their recruiting management system to do the first filter or rank of candidates (94% for middle-skills roles, 92% for high-skills). The software isn't broken. It does what it was configured to do, which is shrink the pile as cheaply as possible.

Related reading: Project-Based Hiring in 2026: Do Work Samples Beat Interviews? · CSR Employability Programs in 2026: How Employers Fund Programs That Work · Campus Job Expo vs Job Fair in 2026: A Playbook for Career Services

Why the hidden workers ATS problem is bigger in 2026

The hidden workers ATS problem carries higher stakes than it did in 2021, because application volume, legal exposure and regulation have all moved against keyword-only screening. Start with volume. The New York Times reported in June 2025 that LinkedIn was handling an average of 11,000 applications a minute, up 45% in a year, with generative AI doing a lot of the typing. When every applicant can paste the job description into a chatbot and get back a keyword-perfect CV, keyword matching stops telling you much. It rewards whoever tailored hardest, not whoever can do the work.

Then there's the courtroom. In May 2025, Judge Rita Lin of the US District Court for the Northern District of California granted preliminary certification of a nationwide age-discrimination collective in Mobley v. Workday, covering applicants aged 40 and over who applied through Workday's platform since September 2020. Court-authorised notice went out in February 2026, with an opt-in deadline of 7 March 2026. Certification is a procedural step, not a finding that Workday discriminated. Still, the theory being tested (that the vendor behind a screening tool can share liability with the employer using it) has every HR technology buyer re-reading their contracts.

Europe runs on a slower clock. The EU AI Act lists AI systems used to filter job applications and evaluate candidates as high-risk in Annex III, and the Digital Omnibus, agreed in May 2026 and in force since 27 July, moved the start of those obligations from 2 August 2026 to 2 December 2027. Plenty of teams read that as a reprieve. We'd read it as 14 months to find out what your filters are doing before a regulator asks you.

How do keyword filters screen out people who can do the job?

Keyword filters hide qualified people because they test for proxies (titles, degrees, unbroken dates) and treat a missing proxy as a missing skill. The Hidden Workers report calls these "negative" filters: rules that knock an applicant out for something they lack, whatever else is on the page.

The employment gap is the clearest case. In the survey, 48% of employers filtered middle-skills candidates on gaps of more than six months. The rule can't tell a difficult pregnancy from a partner's illness or a military relocation. It sees seven blank months and moves on, and no recruiter ever opens the application.

Job descriptions make it worse. 72% of employers said that when they posted a new middle-skills role they reused the old posting or tweaked it slightly, and only 8% wrote one from scratch. For high-skills roles, 25% started fresh. So requirements pile up year after year, and each one becomes another keyword the ATS treats as mandatory.

Picture a warehouse operations lead who spent eight months caring for a parent and now applies for a shift-manager role at a logistics firm. She has run night shifts for years. Her CV says "team lead" rather than "supervisor", she holds a forklift licence but not the newer safety certificate someone bolted onto the template, and there's the gap. Three knockouts. She never reaches a human, and the hiring manager tells HR there's nobody out there.

The worker side of the survey is grim reading. Long-term unemployed respondents had applied for an average of 44.2 jobs over five years and received an average of 1.2 full-time offers. That isn't a motivation problem.

Keyword screening vs skills-based screening, side by side

Skills-based screening asks candidates to show the work, while keyword screening checks whether the CV uses the right words. Here's how the two compare on the things a talent acquisition lead actually gets measured on.

One honest caveat. Skills-based screening costs more per candidate who reaches the task, so you don't put 3,000 applicants through a two-hour project. Use a short, role-specific task early and keep the heavier assessment for the shortlist.

What changes when screening looks at demonstrated skills?

When screening looks at demonstrated skills, the shortlist changes, but only if the evidence replaces the keyword filter instead of sitting behind it. That's where most skills-based hiring programs quietly fail.

The Burning Glass Institute and Harvard Business School measured this in a February 2024 study of 11,300 roles at large firms. Removing a degree requirement raised the share of non-degree hires by about 3.5 percentage points on average, and the researchers estimated that skills-based hiring had changed the outcome in fewer than 1 in 700 hires the year before. Roughly 45% of the firms made the change in name only. Our view is simple: deleting a line from the posting while keeping the same ATS rules and the same interview script is mostly PR.

The firms that followed through got paid for it. At the study's Skills-Based Hiring Leaders, a group that included Walmart, Apple, General Motors, Cigna and Yelp, non-degree hires had a two-year retention rate 10 percentage points higher than their degree-holding colleagues (58% against 48%). The Hidden Workers survey points the same way: companies that hire hidden workers reported being 36% less likely to face talent and skills shortages. And LinkedIn's 2023 Skills-First report found that talent pools grow nearly tenfold on average when employers look for people with the relevant skills rather than only those who've already held the job.

So what does "demonstrated skills" look like on a Tuesday afternoon? Three formats do most of the work:

• A short real project, scoped to two or three hours, that mirrors something the person would do in their first month
• AI role-play. A simulated customer call, stakeholder pushback or incident, so you see judgement and communication instead of reading a claim about them
• A portfolio or work log the talent already has, reviewed against a rubric rather than skimmed for brand names

This is the pre-hire stage we built into HireOS for employers on Talenlio. You set a project or a role-play for the role, talents complete it, and the output lands in the same pipeline as the application. The tool matters less than the sequence, though: the evidence should arrive before anyone decides who's worth an interview.

Universities, NGOs and CSR teams hold the other half of the fix

Universities, NGOs and CSR teams can shrink the hidden workforce from the supply side by helping people build evidence that employers can screen on. Employers own the filter. You own what the candidate shows up with.

For university placement teams the opening already exists. NACE's Job Outlook 2025 survey, published in November 2024, found that 64.8% of employers use skills-based practices for new entry-level hires. Of those, 90% apply them at interview but only about two-thirds at the screening stage, and 46% of employers still screen on GPA. Plenty of graduates still hit a keyword or grade gate before anyone looks at what they built. Placement offices that help talents package capstone projects, internships and role-play results into a reviewable portfolio give recruiters something better to screen on.

NGOs and workforce programs work with exactly the people the Hidden Workers report describes: long-term unemployed adults, refugees, returners, people leaving prison. A participant who arrives with a finished project and a scored role-play is much harder to filter out on a date gap. That's the logic behind our setup for NGOs and workforce programs, which gives each participant an AI career agent and tracks employment outcomes for funder reports.

CSR and impact leads should take one argument from the report seriously. Fuller's team found that most companies working with hidden workers did it through corporate foundations or CSR efforts, and warned that this framing feeds the myth that hiring them is charity rather than a source of advantage. Fund programs that end in a real hiring pipeline, where participants take the same assessment as everyone else. A separate track with lower bars undermines the people it's meant to help. Demand isn't the bottleneck either: 64% of hidden workers in the survey said they'd enrol in employer-paid online training.

A 90-day plan for HR and placement teams

You can test skills-based screening on one high-volume role in 90 days without replacing your ATS. Here's the order we'd follow.

1. Weeks 1 to 2, audit the knockouts. Export every auto-reject rule and knockout question for one role, then count how many applicants each rule removed last quarter. Plenty of teams never have.
2. In weeks 3 and 4, rewrite the job description from a blank page with the hiring manager. Keep five or six must-haves that predict performance in the first six months and move the rest out of the filter.
3. Swap negative filters for affirmative ones. Switch off the employment-gap and exact-title rules for the pilot role and replace them with a short task that every applicant meeting the legal minimums can take.
4. Weeks 5 to 10, run the task in parallel. Score it with a written rubric, blind to the CV, and compare who your old filter would have passed against who the task passed.
5. Weeks 11 to 13 are for measurement: pass-through rates by segment (returners, non-degree applicants, career changers), time to hire, and how hiring managers rate the shortlist. Book a 90-day retention check for every hire.
6. Ask your ATS vendor, in writing, what its scoring model weights and whether it has been bias-tested. After Mobley, a vague answer tells you something.

University placement teams can run the mirror version. Pick one partner employer, agree on the task together, and send talents in with completed work rather than a CV alone.

The question to put to your hiring managers

The most useful hiring metric in 2026 is how many qualified people your ATS threw away, and almost nobody tracks it. The Hidden Workers authors made the point that no management team would tolerate this error rate in supply chain or customer service. If a supplier-qualification process rejected good vendors as reliably as these filters reject good candidates, procurement would have shut it down years ago.

So bring one question to your next talent review. Of the last 500 applicants your system rejected for the role you can't fill, how many could have done the job? If nobody in the room knows, that's where to start.