
When AI Generates Resumes That Are 5x More Perfect, Why Do Traditional Screening Mechanisms Completely Collapse?
Author: John Hempton, Founder of Bronte Capital
Compiled by: TechFlow
TechFlow Summary: For decades, whether in corporate hiring, consulting interviews, or government agencies handling appeals, modern society has taken for granted “clear and rigorous written expression” as the first moat for measuring capability and filtering out noise. However, the widespread adoption of generative AI is now generating massive volumes of seemingly flawless applications and statements at extremely low marginal costs.
From the perspective of frontline employers and managers, renowned hedge fund founder John Hempton of Bronte Capital deeply analyzes how this mechanism is unraveling. As AI mass-produces flattened resume differences and provides cheat-sheet assistance during interviews, employers face not only the exponential surge in screening costs, but also a systemic failure of public input mechanisms.

If you frequently communicate with younger demographics, you will truly feel how artificial intelligence is impacting the job market, particularly entry-level positions for recent graduates. Many feel anxious about their future career prospects, and these concerns are entirely reasonable.
I believe we can now draw a definitive conclusion: in the vast majority of industries that require intellectual investment, the proliferation of AI is indeed not good news for junior workers.
But what I want to discuss more from the other end of the scale is: how AI is destroying the channels through which employers recruit top talent.
Seeking Mavericks: The Hedge Fund’s Past 5-Minute Quick Screen
At Bronte Capital, the hedge fund I co-founded with friends, we never lean toward conventional, step-by-step traditional backgrounds. We do not value the various glitz and credentials commonplace in the financial markets; in my personal standards, having a CFA on a resume often serves as a slight negative signal.
What we truly seek is raw intellectual capacity and intense exploratory curiosity, qualities that are very difficult to screen for solely by flipping through resumes.
Therefore, we established a transparent set of rules open to all applicants: as long as the submitted resume includes a personalized statement tailored to our fund and the reasoning for applying is logical, we will grant an interview opportunity to every candidate.
When sending invitations, we attach a standard note:
"Our interview coverage is far broader than most institutions, interviewing a large number of people. But the inevitable requirement that comes with this is efficiency; we must quickly filter out unsuitable candidates."
"This means most people will only get 5 minutes of conversation time. When we abruptly end the conversation, it may come across as rigid or cold, for which we sincerely apologize. But given our work pace, this is the only way to ensure rapid interviews."
"Interview times will only be extended beyond 10 minutes if one of the following three situations occurs:"
"First, we genuinely hold a high level of interest in hiring you;"
"Second, we learn new knowledge from your answers; listening to your explanation of what you are doing helps deepen our understanding, and in this scenario, we may still decide to hire you;"
"Third, you act as a very clear 'contrarian indicator': you are vigorously pitching a long position to us, but after hearing your logic, we find it presents an excellent short-selling opportunity, allowing us to make substantial profits from the cognitive blind spots you bring."
"In practice, all three situations have genuinely occurred. Once a conversation is extended, we will frankly tell you at the end of the interview which category you fall into."
"If it's the third situation, we look forward to reconnecting and debriefing with you in a few years, where one of us will undoubtedly be wrong. If we are wrong, we are eager to learn from the mistake and reflect on our initial blind arrogance; if you are wrong, you should likewise follow up and debrief to see what you can learn."
"I know the third approach looks exploitative on the surface, as if using you. But it is not entirely exploitative; if two years later you prove with facts that we were the foolish ones terribly mistaken, we might still sincerely invite you to join."
The Screening Funnel Completely Broken: AI-Derived 5x Resume Surge and Real-Time Cheat Sheets
Over the past few years, across hundreds of brief interviews conducted this way, we hired a small core team, helped another half-dozen candidates secure ideal roles at other firms better suited to their styles, and none harbored any complaints.
Meanwhile, we also unearthed over a dozen exceptional short-selling targets through this process, while conveniently noting which financial institutions ultimately absorbed those fraud peddlers who tried to pitch us bogus schemes.
Of course, our straightforward and rapid rejections did sometimes come across as rigid, with many applicants finding this lightning-fast elimination unacceptable and even voicing discontent to our faces.
Regrettably, this mechanism that worked for years has now completely broken down. In the past, we processed roughly 40 to 60 job interviews annually; spending 5 to 10 minutes per week was a perfectly acceptable time commitment for me, especially when conversations yielded cognitive dividends.
However, the explosion of generative AI directly caused the number of job applications I receive to surge by a full 5x. It is impossible for me to organize 300 interviews a year, otherwise my primary investment work would be completely paralyzed.
Therefore, I am forced to conduct upfront disqualifications, forcibly cutting the vast majority of applications.
I could certainly introduce AI tools to help pick preliminary candidates, but asking an AI system to compare and screen hundreds of AI-bulk-generated cover letters often yields selection results significantly worse than a human picking from genuinely human-written materials.
After completing such a highly error-prone initial screening round, I then call candidates in for a 5-minute phone check-in. Sometimes I hear incredibly impressive answers on the call; faced with complex, tricky questions, the candidate displays nuanced and smart insights.
I once thought I had unearthed extraordinary talent, so I arranged for them to visit the physical office for an in-depth discussion lasting three to four hours. After meeting in person, however, I found them utterly mediocre, failing to meet the initial screening threshold altogether.
We later realized: they were almost certainly using real-time AI tools throughout the phone interview to assist with their responses.
From an employer's perspective, the entire recruitment system has become a completely uncontrolled mess.
A Widespread Predicament: The Funnels of Top Consulting Firms Are Also Breaking
This dilemma is by no means an isolated case. Recently, I discussed the same topic with an executive at a well-known large consulting firm. It is a traditional institution that places extreme importance on elite educational backgrounds and heavily relies on the prestige of top universities like Harvard.
In the past, they received 3,000 to 5,000 applications annually, after which the team would initially narrow them down to around 300 candidates, followed by organizing large workshop-style interviews involving business case analyses, ultimately selecting about 80 hires.
But now they face a massive obstacle: they have absolutely no idea how to reduce 3,000 applications down to 300. With AI-assisted polishing, every applicant's resume and cover letter appears flawless, yet simultaneously terrifyingly homogeneous. Their screening mechanism has suffered the same structural break.
A Deeper Fracture: When the Thresholds of Public Governance Are Completely Flattened
The impact of this extends far beyond corporate hiring. While speaking with friends in Canberra, the political center of Australia, they told me that AI is even disrupting the daily operations of the Ombudsman.
In Australia, the Ombudsman serves as the primary receiving agency for citizens lodging objections against administrative actions or rulings by government departments. This system has operated smoothly for years; after an ombudsman intervenes and investigates, many issues originally bogged down in bureaucratic shuffling are often resolved quickly at a lower cost.
But for a long time, the volume of appeals received by the office has always far exceeded its actual capacity to investigate each case in depth. They previously had a highly effective diversion strategy: establishing a barrier that prioritized requests backed by a 3-to-5-page document featuring clear logic and detailed factual statements.
This logic worked incredibly well in the past: individuals capable of clearly outlining the sequence of events likely reported problems that were genuinely existent and worth investigating; and merely the written threshold of “clarifying the issue” naturally filtered out a huge volume of unreasonable, malicious complaints driven by hysteria or paranoid fixation.
Now this moat is completely gone. Any generative AI today can package illogical accusations into structured, convincing arguments, and under prompt guidance, can even generate seemingly highly persuasive “reasonable hallucinations.” The processing system of the Australian Ombudsman is rapidly spiraling into paralysis.
This is by no means a positive development. A cheap, rapid, and effective public feedback mechanism serves as a crucial cornerstone of public governance and democratic functioning, and this foundation is now being quietly eroded.
The Universal Rule of Systemic Collapse
Looking across these phenomena, we can summarize a highly universal rule:
Any system that opens public input channels and historically relied on “written expression costs” as an anti-spam filter—whether corporate hiring, government appeal channels, or social media content interaction—is bound to trend toward collapse in the face of generative AI.
Faced with such systemic challenges, no one currently possesses a ready-made silver bullet.
I have even jokingly pondered whether we should compel every applicant to submit a handwritten, single-page cover letter. But this is obviously just a last resort born of desperation.
I hope someone can devise a more ingenious solution.















