A few years ago, Ken Schumacher was working at a technology company. Part of his job included evaluating job candidates. He hopped on a Zoom call, gave the applicant an engineering test (kind of like a crossword puzzle with codes instead of words), and held “an hour of silence” while the applicant struggled with it.
Except many of the candidates weren’t struggling. The company’s exercises were posted on sites such as Glassdoor. “These smart 23-year-olds, of course, would practice that problem three times and come to me and smash it,” Schumacher told me. “The bigger issue now is that everyone is using AI to write resumes.” They are also using AI-powered chatbots and teleprompters to get to the next round. “It’s become really, really difficult for everyone to tell who’s real and who’s fake,” Schumacher said. (This turned out to be a market opportunity in itself; he now runs a startup that uses AI to detect AI fraud by job applicants.)
This problem can be particularly acute in software engineering. But the same dystopian phenomenon is distorting the labor market as a whole. AI has turned hiring into the Tinder. Workers have applied for hundreds of positions but never heard back. Companies receive thousands of resumes and struggle to respond. More than that, AI has Amazonized jobs. Uniqueness has been removed from applications, similar-looking products have flooded the market, fraud has increased, and personal discretion has been replaced with cruel algorithmic evaluations.
People once hoped that Silicon Valley would not only make the job search process smoother, but also make the process fairer. Unbiased tools will replace alma mater networks. The digital portal accepts applications from anyone, anywhere. Employees will have free access to templates, practice tests, and advice. “Technology generally tends to make job matching more efficient,” said Mitchell Hoffman, a labor economist at the University of California, Santa Barbara. But it looks like AI in particular will destroy it. Kathleen Creel, a philosopher and computer scientist at Northeastern University, said employers and employees are engaged in an “AI vs. AI criminal arms race.”
In just a few years, tools like ChatGPT and Claude have commoditized cover letter and resume writing. A Columbia Business School paper found that the majority of job seekers use generative chatbots to hone their language and summarize their accomplishments. In other words, the average quality of personal documents is improved at the expense of “compression” and “homogenization” of the information being conveyed. In Silicon Valley, this phenomenon is sometimes referred to as “signal collapse.” Resumes used to be full of intentional and inadvertent signals for hiring managers to parse: degrees and certifications, language used, formatting errors, unusual digressions, and too-honest confessions. Now everyone looks better, everyone looks the same, everyone parrots key keywords, and everyone uses punchy action verbs. Recruiters are working hard to “distinguish the underlying expertise” and separate valuable signals from the cacophony of chatbot noise.
Thanks to AI, job seekers spend less time writing applications, so people are submitting hundreds of applications to ZipRecruiter, LinkedIn, and other sites, often for positions that aren’t a good fit, and often never hearing back. Applicants do not receive feedback about their strengths and weaknesses. They don’t have access to information about what they need to do to get hired or the skills they need to get the job they want. Signal decay is occurring in both directions.
To sort through all these applications, companies are once again turning to AI. A recent study by Resume Builder found that 4 out of 5 companies use AI to scan resumes, 2 out of 5 companies use chatbots to communicate with candidates, and 1 out of 5 companies conduct AI interviews. AI selection tools in particular are creating an “algorithmic monoculture” in hiring, a new study of 4 million job applications finds. More and more candidates are being rejected by every company they apply to. Hiring decisions are generally more uniform than they would be if HR managers were not using algorithmic screens. Ten or three years ago, companies “were looking at resumes and picking schools they knew, schools that were well-known, schools that had good programs” in related fields, Creel said. “From an egalitarian perspective, that’s bad in itself. But at least the arbitrary criteria for judging differed from person to person.”
Additionally, the paper said screening algorithms discriminate against black and Asian candidates. “This kind of filtering is happening, but we don’t understand it because these systems are opaque and custom-built for different institutions, and we don’t know how they work,” Sarah Bana, a co-author of the paper and a digital fellow at the Stanford Digital Economy Lab, told me. Nevertheless, the tendency of algorithms to favor the advantaged and disadvantage the disadvantaged seems clear.
Candidates who pass the AI screening may face tests like those conducted by Ken Schumacher. Companies ask applicants to explain how they deal with unruly customers or find engineering solutions to software problems. This step often involves the AI creating the exam, proctoring the exam, and taking the exam. “I see this all day, every day, and sometimes I have to pinch myself. So many scams!” Schumacher said. “That’s completely ridiculous.”
As a result, AI is slowing down HR rather than speeding it up, as companies have to spend more time vetting applications and conducting background checks. Many large companies are re-emphasizing face-to-face interactions in the hiring process. Google makes sure candidates have “at least one in-person interview” to “make sure they have the basics,” CEO Sundar Pichai said. Cisco is ramping up “verification procedures and enhanced background checks that may include an in-person component.” Companies are also extending trial periods or hiring contract workers with the aim of converting them to permanent employees after managers review their work.
Finally, some companies rely on old-fashioned methods such as referrals, alumni networks, local job boards, and headhunters. Schumacher said companies are “regressing to pedigree.” “Because Harvard engineers are the best in the world,” he says. “But in a time when it’s really hard to tell who’s legitimate and who’s full of shit, this is a safer play.”
Even with these techniques, recruitment may simply become more difficult, the overall labor market may become a little more rigid, and business dynamism may diminish a little. Companies are concerned that employees’ tenures will be reduced because managers will have to fire less capable employees, or because unsuitable employees will become dissatisfied and quit. Bana worries that companies are missing out on employees who have the potential to replicate their current workforce, break out of groupthink and expand their ambitions. Hoffman, a labor economist, raised the issue of AI discouraging workers from learning new skills and deepening their expertise.
What should someone looking for a job in this AI hell do? No one I spoke to was sure. But mentioning yourself in your cover letter and submitting it directly doesn’t seem like the worst idea.
