Hiring Algorithms Are Having Their Day in Court

An AI hiring tool is supposed to be defensible precisely because it is consistent. Feed it the same inputs, get the same outputs, no bad mood, no bias, no favouritism. This week that same consistency became the plaintiffs' best evidence. A California court is letting age discrimination claims against Workday's hiring software move forward, on the theory that a tool applying the same flawed logic to millions of applicants isn't a defence against bias. It's the mechanism of it.

The rest of the week told variations on the same story. A principal engineer three years into an AI-forward organisation described his job quietly expanding past what a working week can hold. The labour force participation rate hit a fifty-year low outside the pandemic. Companies that spent the RTO era insisting flexibility was over kept finding exceptions for the people they most wanted to keep. Four separate claims about how AI-era work is supposed to function, four moments this week where the claim met the evidence.

FROM THE EBN NETWORK

Your employees know the truth. Does your EVP? At Fathom we measure the "Credibility Gap" between your promise and their reality.

Close the Gap

This Week’s Paper Cut's

Watch or listen here ↯

We keep tracking to a minimum. The video is hosted on YouTube and the podcast is hosted on PodBean, which may place cookies that measure viewing and personalize content. We only load the player after you consent to these cookies. You can update your cookie settings any time.

Workday's age bias case isn't really about Workday

A federal judge in California has allowed a landmark class action against Workday's hiring software to continue, rejecting the argument that using an algorithm to screen candidates puts a company at arm's length from discrimination law. The case centres on claims that the tool systematically filtered out applicants over 40, and it has been running since 2023 with EEOC backing for the plaintiffs.

What makes this week's development significant isn't the age of the case. It's the precedent risk. Workday's tools sit inside thousands of corporate hiring stacks, which means a ruling against the company doesn't stay contained to Workday. It becomes a template for every organisation that outsourced its screening logic to a vendor and assumed the vendor absorbed the liability along with the task.

Candidate-side fraud is rising in parallel, and for a related reason. As HR teams lean harder on automated screening to manage volume, candidates have found their own automated ways around it, from AI-polished applications to more deliberate misrepresentation. The two trends feed each other: more automation on one side of the hiring table produces more automation on the other, and the humans meant to be doing the judging end up further from the decision each time.

There's a live case study in what happens when a services firm bets against pure automation. Korn Ferry's $1.1 billion acquisition of AMS this week wasn't a bet on an AI hiring platform. It was a bet on a recruitment outsourcing firm with long client contracts and embedded human delivery teams, at a moment when the market narrative says hiring is about to be automated away.

EBN covered this fight in detail when the Mobley case was first certified as a class action, including the detail that more than 11,000 organisations run Workday's hiring tools, which is the scale that makes this week's ruling a precedent risk rather than a single-company problem.

Employer Brand POV
Audit the hiring stack before a plaintiff's lawyer does it for you. If a vendor's tool makes the reject decision, your organisation still owns the outcome, not the vendor, and "the algorithm did it" has already failed as a legal defence once this year. Ask what your tool actually screens for, whether anyone has tested it for age, race, or disability bias, and whether a human sees a rejection before it goes out. Employer brand teams promising fair, human hiring while an unaudited tool runs the first cut are building a credibility problem that a lawsuit will make public on your behalf.

What AI is actually doing to engineers, and to their confidence

The dominant story about AI and software engineering has been productivity: fewer keystrokes, faster shipping, agents writing the boilerplate. This week's reporting complicated that story from two directions at once. Business Insider spoke to engineers who describe AI reshaping not just how they code but how confident they feel doing the job at all, with the tools that were meant to remove drudgery instead unsettling their sense of what their expertise is worth.

Ford supplied the more literal version of the same problem. The manufacturer has started rehiring human engineers after AI-generated work failed to meet quality checks, according to the BBC, a rare public admission that automation moved faster than the organisation's ability to verify what it produced.

Not every executive is framing this gently. Palo Alto Networks CEO Nikesh Arora told Fortune this week that workers are approaching a "Darwinian moment," where adapting to AI tools becomes the difference between staying employed and getting cut. It's a blunter version of a message many organisations are sending more quietly through headcount decisions rather than speeches.

EBN spent time inside exactly that experience earlier this month, in one principal engineer's account of three years watching his own role expand past what a working week can hold, with mentoring time the first casualty and thinking time pushed into holidays.

Employer Brand POV
Give senior engineers back unstructured time, or budget for their replacement. The pattern across this week's reporting is consistent: the people getting the most out of AI tools are also the people running hardest to keep up with the expectations those tools create. That erosion doesn't show up in a survey question about job satisfaction. It shows up eighteen months later on a review site, worded far less carefully than anything your engineers would say to your face today.

The labour market is fraying without anyone calling it a downturn

The US labour force participation rate has fallen to its lowest level in fifty years outside the pandemic, according to CNBC, a statistic that describes people leaving the hunt for work altogether rather than simply struggling to find it. Reuters reported the same week that US job growth missed expectations in June even as the headline unemployment rate fell to 4.2%, a combination that reads as reassuring only if nobody checks the participation figure sitting underneath it.

Reuters also reported a more direct mechanism: companies are cutting jobs specifically as investment shifts toward AI, not as a side effect of a broader slowdown but as a stated reallocation of capital. In the UK, the government's Keep Britain Working programme continued its push to stop people falling out of the workforce entirely, an intervention that only makes sense if the risk of permanent exit is real enough to warrant a national policy response.

None of these are dramatic single events. Each one individually reads as a data point rather than a crisis. Taken together, they describe a labour market where the official unemployment rate is doing less and less work as a signal of how people are actually faring.

Employer Brand POV
Report participation and retention numbers plainly this quarter, not just headcount. A falling unemployment rate paired with people leaving the workforce entirely is not the same story as a healthy labour market, and candidates researching your sector will find the discrepancy whether or not your careers page acknowledges it. Organisations that keep citing the unemployment rate as evidence of a "tight but stable" market will look increasingly out of step with what their own applicant pools are experiencing.

Office mandates are back, but flexibility just relocated

Revolut is making office attendance compulsory for new graduate hires, according to Fortune, with flexibility offered only after a year on the job, a notable shift for a company that built part of its employer brand on remote-first working. It's a pointed reminder that "remote-first" was frequently always a policy for existing staff rather than a permanent commitment to everyone the company would ever hire.

And yet remote work itself hasn't gone anywhere. Axios reported that remote work continues to thrive across large parts of the workforce even as high-profile mandates make headlines, and NPR ran a piece this week asking the more basic question underneath all of it: whether working from home is actually good for people, beyond the convenience argument either side usually leads with.

Nvidia offers a useful counterpoint to the assumption that winning the talent war requires generous perks. Fortune reported that the company built a nearly $5 trillion valuation with a notably frugal internal culture, including making employees pay for their own lunch, while competitors spend heavily on on-site amenities to compete for the same engineers. The lesson isn't that perks don't matter. It's that mission and compensation can outweigh them when the underlying opportunity is strong enough, which is a much harder story to tell honestly than "we offer free lunch."

Where employer brand actually gets tested now may not be the careers page at all. A piece in Human Resources Online this week made the case that authentic employer reputation is happening on Glassdoor and Indeed rather than on company-controlled channels, which is either a threat or an opportunity depending on how honest your reviews currently are.

Employer Brand POV
Match the RTO memo to the actual policy before your employees do it for you on a review site. Revolut's graduate-only mandate is defensible precisely because it's specific and stated plainly, not dressed up as a universal culture decision. Vague or selectively enforced office policies read as favouritism the moment two employees compare notes, and in 2026 they will compare notes somewhere public.

The Cut List

British American Tobacco is cutting 5,500 jobs globally as the company shifts toward smokeless products, per HR.Economic Times.

Microsoft may cut thousands more jobs to control costs, Fortune reported as part of a wider tech-layoffs roundup this week.

TikTok confirmed layoffs in Indonesia as part of continued global restructuring, according to People Matters.

Hollywood's film industry has lost 6,700 jobs as AI reshapes production workflows, People Matters reported.

Newsweek's running list of companies laying off staff in July adds further names to a month that is opening with more cuts than it closed with in June.

Read together, the week's stories share a structure rather than a topic. Each one is a claim about how AI-era work is supposed to function, tested against what actually happened when someone checked. The claims didn't all fail. Some held up. But the gap between what organisations say about fairness, productivity, and flexibility, and what a court, an engineer, or a labour statistic can verify, keeps closing, and it's closing in public.

Takeaways

• A California court is letting age discrimination claims against Workday's hiring AI proceed, on the basis that algorithmic consistency doesn't shield a company from bias liability.
• Candidate fraud is rising alongside automated screening, with both sides of the hiring table now using AI to work around the other.
• Ford has rehired human engineers after AI-generated work failed quality checks, an unusually direct admission that automation outpaced verification.
• US labour force participation has fallen to its lowest level in fifty years outside the pandemic, even as the headline unemployment rate improved.
• Companies are citing AI investment directly as a reason for job cuts, not as an indirect consequence of a slowing economy.
• Revolut is making office attendance compulsory for new graduate hires while retaining flexibility for existing staff, a split policy other RTO mandates rarely state as plainly.
• Nvidia's frugal internal culture is a reminder that compensation and mission can outweigh perks in the talent war, when the underlying opportunity is strong enough.

SOURCES

Trending on EBN