Your hiring software is two robots lying to each other

^{Work takes up most of our waking hours, so getting it right matters. I'm Clif P Lewis, an organisational psychologist, and I write about what the research says makes work work.}

Subscribe | Past essays | Advertise with us | Insane Media

🤝 This edition is kindly brought to you by SurveyMonkey

Retention starts with understanding what people actually want out of their careers. SurveyMonkey's employee engagement program helps you ask the right questions about growth, recognition, and belonging, then use AI-powered analysis to turn the answers into a real employee development strategy. Benchmark against industry data to see how you compare, and build a workplace your best people are excited to grow in.

Start growing your team today

_{We partner with a select group of brands that we use or admire, allowing this newsletter to remain free and independent. Reach out for your campaign here.}

In true Michael Scott fashion, when Ryan arrives promising to “throw out the entire playbook” when it comes to adopting new technology, Michael spends the entire Dunder Mifflin Infinity episode of The Office trying to prove that gift baskets beat websites. Rebuffed by client after client, he blindly follows his GPS into Lake Scranton, insisting that “The machine knows where it is going!”

What strikes me is that he resists the technology all episode and then hands it his judgement in one turn of the wheel. Many organisations I see do both at once, bolting new tech onto processes it was never designed for and chasing efficiency in one place while quietly breaking another. Hiring is where this shows quite prominently.

Employers cannot see a candidate’s ability, so they rely on signals, and the economist Michael Spence (1973) showed that a signal is only believable when it costs a weaker candidate more to send than a stronger one. That cost difference is still the core of signalling theory. A tailored CV and a considered cover letter took time, effort and some writing skill, which is why they offered a signal about a candidate’s abilities. Bangerter et al. (2012) add that the two sides keep outmanoeuvring each other. Employers add a screen, applicants learn to beat it, employers add another.

Generative AI made a polished, tailored application free, and early evidence suggests employers noticed. On one freelancing platform, employers stopped paying a premium for customised applications once large language models arrived (Galdin & Silbert, 2025). On the same platform, the link between how well a cover letter matched the job and getting a callback roughly halved after it added one-click AI cover letters (Cui et al., 2025).

Now add the employer’s AI, scoring every application for fit, while the candidate’s AI writes to that score. Campbell (1979) already clocked this as a problem nearly 50 years ago. He warned that the more a quantitative indicator is used for decisions, the more it attracts corruption and the more it distorts the process it was meant to monitor (p. 85). That inevitable pattern of constant outmaneuvering between applicant and employer now runs at machine speed, and we do the broken thing faster.

The impact of this is unevenly distributed, though. When written signals fail, employers lean harder on work history (Cui et al., 2025) and, I’d argue, on judgements of who seems competent, or is a likely cultural fit - one of the ways gender, race and class inequality survives in organisations (Acker, 2006). Interviewing 21 developers and vendors, Yarrow (2026) found that bias in AI-hiring tools are treated as something to patch after launch, rather than an inherent, systemic problem.

Michael could have stayed dry, since Dwight was in the passenger seat shouting that there was a lake. Most hiring processes rest on assumptions nobody has revisited in decades, starting with the idea that a well-written CV still tells you something about the person who sent it. Technology will only scale an assumption you never questioned, all the way into the lake.

If something here speaks to you, I’d love to hear it.

Stay human,
Clif

P.S. Where do your pre-hire criteria stop matching what your team surely values and rewards? Reply to this email and let me know.

If you’re new here, I’m over the moon you’ve joined us! To help me craft content that’s actually useful (and not just noise in your inbox), I’d love it if you took 1 minute to answer this quick survey below. Your insights help shape everything I write.

💡 Dive into our other newsletters - where psychology meets the founders, creator economy, e-commerce marketing, and AI founders.