A Court Ruled the AI That Screens Your Applicants Can Be Sued for Bias. You Can’t Outsource the Blame Along With the Work.

What the Court Actually Decided

Derek Mobley sued Workday in 2023, alleging that its applicant-screening tools rejected him from more than 100 jobs at companies that use its software, and that the pattern tracked his age, race, and a disability. His claims run under the big three federal statutes: Title VII, the Americans with Disabilities Act, and the Age Discrimination in Employment Act. In July 2024, Judge Rita Lin of the Northern District of California let the case proceed on a theory that had not been tested this way before, writing that Workday’s customers “delegate traditional hiring functions, including rejecting applicants,” to its tools, which makes the vendor an “agent” of the employer and therefore answerable under those laws itself.

In May 2025 the court went a step further and preliminarily certified a nationwide collective action on the age claim, covering applicants 40 and older who were rejected through Workday since September 2020. Court-authorized notice went out to potential members in early 2026, and the case is in discovery as of this writing. It is worth being precise about what has and has not happened: no court has found that Workday, or anyone using it, actually discriminated. What a court has decided is that the question is serious enough to answer in open court, and that the software vendor cannot stand outside the room while it does.

Why This Isn’t a Big-Company Problem

The easy misread is that the ruling moves liability onto vendors, so employers can relax. It does the opposite. An employer has always been answerable for a discriminatory outcome in its hiring; buying a tool never transferred that responsibility, and the court did not lift it now. All the ruling did was put the vendor on the hook beside the employer who chose to deploy it. The delegation the court described, handing the reject-or-advance decision to an algorithm, is not a Workday feature. It is a choice, and it is now a clearly attributable one no matter whose software you run.

And small teams make that choice constantly. AI screening is no longer an enterprise luxury; it is a checkbox in the tools a two-person startup already pays for. In a ResumeBuilder survey of 948 business leaders, roughly 82% of companies using AI in hiring said they apply it to reviewing résumés, and the firm projected AI-assisted résumé screening would reach around 83% of companies by the end of 2025. A founder running an off-the-shelf matcher to thin a flood of applications is making the same delegation Mobley’s case is about, at smaller scale and with far less legal cover than Workday can afford.

The Real Exposure Is the Rejection You Can’t Explain

Here is the part that catches good-faith teams off guard. These laws do not only reach people who set out to discriminate. They reach outcomes. A filter that looks perfectly neutral can still break the law if, in effect, it screens out older applicants or one protected group, and an AI screener is exactly the kind of system that can quietly learn a proxy, a graduation year, an employment gap, the shape of a name, and act on it without anyone writing a rule to do so. Nobody decides to reject over-40 candidates. The model may simply do it, and across hundreds of applications a role, the pattern is real even when no single rejection looks wrong.

What turns that risk into a liability is a single question you may not be able to answer: why was this person rejected? If the honest reply is “the tool scored them low and I could not tell you how,” you have neither a defense if you are challenged nor a way to notice and fix the pattern before it grows. A rejection you can state in plain, job-related terms is defensible and correctable. A rejection nobody can explain is the exposure, whether or not a lawyer ever comes calling.

What a Small Team Should Do

The fix is not to swear off automation and read every résumé by hand. The volume that pushed teams toward AI in the first place has not gone anywhere, and pretending otherwise just trades one failure for another. The fix is to change what you let the machine decide. Let it sort, surface, and organize against criteria you actually chose and can name; keep a human making the reject-or-advance call on the shortlist; and keep a plain-language, job-related reason attached to every no. That record is what makes a decision auditable, and it is, not by coincidence, the same record that lets you give a candidate a real answer instead of a void.

That division of labor, automation on the sorting and the paperwork, a person on the judgment and the reason, is the line we drew when we built Kynto. The point was never to hand the decision to a model; it was to let a small team move at the speed automation gives you while a human stays accountable for every call, with the reason written down rather than buried in a score. In a year when a court has started asking who owns the rejection, being able to answer is not just good manners. It is the cheapest insurance a small team can carry.

Key Takeaways

• A federal court let an AI-hiring discrimination case proceed on an “agent” theory. In Mobley v. Workday, a job seeker who says he applied to 100+ roles is leading a nationwide age-discrimination collective (applicants 40+) that a judge preliminarily certified in May 2025. No discrimination has been proven; the case is in discovery.
• The ruling adds the vendor to the liability; it does not remove the employer. Any team letting AI auto-reject applicants is making the same delegation, and about 82% of companies using AI in hiring apply it to résumé screening, so this is a small-team practice, not an enterprise-only one.
• The exposure is not AI itself; it is the rejection you can’t explain. Anti-discrimination law reaches outcomes, not only intent. Let automation sort and draft, keep a human on the reject-or-advance decision, and attach a job-related reason to every no. A decision you can explain is defensible; one nobody can is the risk.

None of this needs a lawyer on retainer or a fear of the next class action. It needs a small, unglamorous discipline: knowing why you said no, and being able to say it. The teams that get caught out in this era will be the ones that let a black box thin the pile and never asked it to show its work. The ones who come through fine will look almost old-fashioned, a person, a reason, a record, which turns out to be exactly what the law, and the candidate, were asking for all along.