Your HR Team Is Drowning in Admin Work While Workday Faces a Bias Lawsuit. There's a Better Way to Use AI.
Workday is facing a nationwide federal class action lawsuit right now, in 2025, because its AI hiring tools allegedly discriminated against applicants by race, age, and disability. A federal court certified it in May. CNN covered it. And yet companies are still paying Workday tens of thousands of dollars a year to automate their recruiting. That's where we are. The promise of AI in HR has been loud, expensive, and in many cases, actively harmful. But here's the thing: the problem was never automation itself. The problem was the kind of automation companies chose. Black-box scoring models trained on biased historical data aren't the future of HR. A true computer use agent, one that operates like a human actually sitting at a desk, is. And most HR teams have never heard of it.
The Numbers Are Genuinely Embarrassing
Let's put some real figures on the table, because the vague hand-waving about 'HR inefficiency' has to stop. According to research compiled by Deel, HR staff spend up to 57% of their time on purely administrative tasks. Not strategy. Not culture-building. Not even real recruiting. Copying data between systems, chasing down signatures, updating spreadsheets, scheduling interviews, sending the same rejection email for the 400th time. Meanwhile, SHRM data shows the average cost per hire in 2024 hit $4,700, and the average time to fill a role is 44 days. Forty-four days. For a mid-sized company doing 50 hires a year, that's $235,000 in direct hiring costs, plus the hidden cost of 57% of your HR team's brain power going toward tasks a well-configured computer use agent could handle before lunch. A separate study found that workers across industries waste 13 hours per week on low-value manual tasks. For a 10-person HR team, that's 130 hours a week, or roughly three full-time salaries, evaporating into thin air every single year.
The Workday Lawsuit Is a Wake-Up Call, Not an Anomaly
- ●In May 2025, a federal court certified a nationwide class action against Workday, alleging its AI resume screening tools discriminated against applicants based on race, age, and disability under Title VII, the ADEA, and the ADA.
- ●The plaintiff, Derek Mobley, applied to over 100 jobs at Workday-using companies and was rejected by every single one. He's Black, over 40, and has a mental health condition. The algorithm apparently didn't like any of that.
- ●The EEOC settled its first-ever AI hiring discrimination case in 2023. Two years later, the lawsuits are accelerating, not slowing down.
- ●Colorado, California, and Illinois all passed or updated AI employment laws in 2024-2025. Compliance complexity is exploding just as adoption is rising.
- ●The core problem: most 'AI recruiting tools' are scoring models trained on historical hiring data, which reflects every bias your company ever had, baked into math and sold back to you as objectivity.
- ●43% of organizations now use AI for HR tasks according to 2025 surveys. Less than half of them have any formal AI bias audit process in place.
Workday's AI screened out a Black applicant who applied to 100+ jobs. A federal court just certified that as a class action. Your 'neutral' algorithm isn't neutral. It's your worst hiring manager, scaled to infinity.
Why Traditional HR Automation Is Broken by Design
Legacy HR automation, think UiPath bots, old-school ATS integrations, and Workday's scoring models, was built on a fundamentally flawed premise: that you can solve messy human processes by drawing rigid rules around them. RPA tools break the second a UI changes. Scoring models amplify historical bias. ATS keyword filters reject qualified candidates who didn't use the exact right phrase. And all of it requires a small army of IT people and consultants to maintain. Then came the next wave: ChatGPT plugins, Copilot features, and bolt-on 'AI' from every HR vendor on earth. Better? Sure. But still mostly glorified autocomplete. OpenAI's Operator and Anthropic's Computer Use were supposed to be the answer. Both are still in limited research preview stages as of 2025, with real-world enterprise reliability that ranges from 'inconsistent' to 'not ready for production.' Claude Sonnet scores 61.4% on OSWorld, the gold standard benchmark for real computer task completion. That means it fails nearly 4 out of 10 tasks. You want that running your offer letter workflow?
What a Real Computer Use Agent Actually Does for HR
Here's what people get wrong about computer use AI: it's not a chatbot you ask questions. It's not an API integration your dev team spends three months building. A computer use agent actually controls a real desktop or browser, sees the screen, moves the mouse, fills out forms, navigates between applications, and completes multi-step workflows the same way a human would. For HR, that means it can pull a candidate from LinkedIn, cross-reference their resume against your ATS, schedule an interview in Calendly, update the candidate record in Greenhouse, and send a personalized confirmation email, all in one uninterrupted flow, without a single API key or Zapier chain. It can onboard a new hire across six different systems in the time it takes your HR coordinator to find the right tab. It can pull compensation benchmarking data, format it into your internal template, and drop it in the right Slack channel before your 10am comp review. The key difference from old automation: it doesn't need to be pre-programmed for every possible screen state. It reads the interface like a human does and adapts. That's what makes it genuinely useful instead of a brittle nightmare.
Why Coasty Is the Computer Use Agent HR Teams Should Actually Be Using
I'll be straight with you. I work at Coasty. But I'm recommending it here because the benchmark numbers are real and the gap is significant. Coasty scores 82% on OSWorld, the standard benchmark for AI computer use performance. Claude Sonnet, Anthropic's best model for computer use tasks, scores 61.4%. OpenAI's Operator isn't close either. That 20-point gap isn't a marketing number. It's the difference between an agent that actually finishes the job and one that gets stuck on step four and quietly does nothing. For HR teams specifically, Coasty controls real desktops, cloud VMs, and browsers. It doesn't need API access to your HRIS. It doesn't need a custom integration built by your IT team. It works the way a contractor would: you show it the screen, you describe the task, it executes. You can run agent swarms for parallel execution, which means if you need to process 200 applications at once, it doesn't queue them up for three hours. There's a free tier to start, BYOK support if you're particular about your model costs, and you're not locked into a $50,000 enterprise contract before you've proven a single workflow. If you're spending 44 days and $4,700 per hire while your HR team drowns in copy-paste work, the ROI math takes about 30 seconds to do.
Here's my honest take. Most HR technology in 2025 is either too dumb to be useful or too opaque to be trusted. The Workday lawsuit isn't a freak event. It's the inevitable result of companies buying black-box AI, not understanding what it actually does, and deploying it at scale on decisions that affect people's livelihoods. The answer isn't to avoid AI in HR. The answer is to use AI that's transparent, capable, and actually built to handle the messy reality of real software on real computers. That's what computer use AI does. Your HR team should not be spending 57% of their time on admin in 2025. That's not a productivity problem. That's a choice. Stop making it. Start at coasty.ai.