The Algorithmic Gatekeeper: When AI Rejects You First

Simon Carver

Welcome to the show. The title of this episode is The First AI That Worked Against You — Part II: The Algorithmic Gatekeeper, and the blunt version is this: for a lot of people, the first AI they ever met was not a cheerful chatbot or some clever assistant... it was an algorithm that decided they were not worth a human review. If this conversation lands with you, like it, share it, subscribe to the show. I’m Simon Carver, and I’m here with Lachlan Reed, plus our guest hosts Jacques San Dimas and CJ Murphy.

Lachlan Reed

[warmly] Yeah, and that opener hits a bit too close to home, mate. Because this is the modern job hunt now, isn’t it? You upload the résumé, then re-type the same résumé into seventeen little boxes like a goose in a high-vis vest, and then... nothing. Or you get that chirpy email three minutes later saying they’ve decided to move forward with other candidates, which is a very polite way of saying a machine shut the gate before a person even looked over the fence.

Chris J. Murphy

[reflective] Let’s talk about what’s actually happening. We’ve normalized a hiring experience where silence is part of the process. Not a phone call. Not a note. Often not even confirmation that a real job was available in the first place. And when that happens at scale — hundreds of applications, millions of workers — efficiency stops feeling like progress. It starts feeling like erasure.

Jacques San Dimas

[calm] Yes. And that is the word I would stay with — erasure. In risk work, we often discuss abstractions: systems, throughput, optimization, capacity. But a career is not an abstraction. It is rent. It is medicine. It is a parent trying to remain dignified in front of a child. When a human being disappears into software, the organization may call that efficient. The person living through it experiences something much colder.

Simon Carver

[curious] Jacques, I keep coming back to that little moment right after you hit submit. It’s such a tiny action. A click. But emotionally it’s not tiny at all. There’s hope in it. You imagine maybe somebody sees your name, maybe somebody notices the weird zigzag in your career and thinks, huh, there’s a story there. And then the rejection arrives so fast it almost tells on itself.

Lachlan Reed

Right — the speed is the giveaway. If you get bounced in, like, four minutes, nobody did a deep read. Nobody brewed a coffee, opened your cover letter, had a think, called a manager. Four minutes says software. It says keyword match, field mismatch, ranking threshold, next please. It’s a servo on a factory line, except the product getting sorted is your livelihood.

Chris J. Murphy

And to be fair, there is a real operational argument underneath it. Companies receive enormous application volume. They use applicant tracking systems — ATS platforms — to parse résumés, standardize data, rank applicants, and surface a shortlist. That sounds reasonable until you ask the next question: what happens to everyone below the cutoff, and who is accountable for the cutoff itself?

Simon Carver

[softly] That’s the tension for me. I don’t think most listeners are anti-technology. I’m not. If software helps a recruiter handle 500 applications instead of drowning in them, okay. But if the price of that efficiency is that people never know whether they were rejected by a human being, by a model, by a frozen budget, or by a glitch in a form field... then we’ve saved time by spending dignity.

Jacques San Dimas

[matter-of-fact] And organizations always tell themselves this trade is temporary. Just until things stabilize. Just until budgets return. Just until headcount opens. But systems have a way of becoming culture. What begins as a convenience becomes a habit. Then a habit becomes policy. And before long, no one remembers when hiring stopped being a human conversation and became silent intake.

Chris J. Murphy

[measured] I want to make this personal for a moment, because I don’t think this issue lands unless we admit what it feels like from the inside. After displacement, I sent out more than 125 applications in about three months. One hundred twenty-five. I received three meaningful responses, two interviews, and no offer. And what stuck with me wasn’t rejection by itself. It was watching roles remain online for months, getting delayed rejection emails long after they seemed inactive, and genuinely not knowing whether a human being had ever reviewed my application at all.

Simon Carver

[quietly] One hundred twenty-five is the number that stays with me there. Not twelve. Not twenty. A hundred and twenty-five. That’s not “keep trying.” That’s a system teaching somebody to question their own reality.

Chris J. Murphy

Exactly. You start doing math on your own worth, which is dangerous. You wonder whether your experience is outdated, whether your résumé is wrong, whether your entire professional identity has somehow become invisible. The hardest part wasn’t rejection. It was not knowing whether I was being rejected by a human being... or by a system that had already decided people like me were no longer worth considering.

Lachlan Reed

[skeptical] And “people like me” there — that phrase matters. Because the machine never says what category it thinks you’re in. Too old, too expensive, too nonlinear, not enough exact keyword overlap, wrong title history, whatever. It’s like being knocked back by a locked door that won’t tell you which key it wanted.

Jacques San Dimas

[reflective] There is another layer here, and it is uncomfortable. In large enterprises, public job postings do not always mean active hiring in the way applicants imagine. There are periods of internal hiring freeze, budget containment, delayed backfills, workforce reduction — and yet requisitions remain open. Why? Because appearances matter. Publicly visible roles can signal growth, momentum, confidence. They help build candidate pipelines for later. They reassure investors and the market that the organization is alive and expanding, even when internally the truth is far more constrained.

Lachlan Reed

Wait — “requisitions remain open” is the phrase that gets me. You mean the role can be sitting there on the website, collecting applicants, while inside the company somebody already knows approval is shaky or basically dead?

Jacques San Dimas

[calm] Yes. Not always, of course, but yes, that does happen. The HR system continues intake because intake is easy. Closing the role, communicating clearly, updating governance — that requires friction, accountability, honesty. So the pipeline stays open. From the applicant’s side, time is being spent, hope is being invested, customized materials are being written. Internally, there may be no realistic path to hiring with urgency.

Simon Carver

There’s something almost theatrical about that. The storefront lights are on, the sign says open, the menu’s in the window... and meanwhile the kitchen has quietly shut down.

Jacques San Dimas

[dryly] As a former chef, Simon, that is a painful analogy because it is accurate. In a kitchen, if you take orders while knowing the burners are off, you are not being efficient. You are deceiving people.

Chris J. Murphy

And this is where the technology layer becomes ethically tricky. ATS tools don’t just store applications. Many now assist with ranking, screening, résumé parsing, profile matching, recommendation engines — all the machinery of narrowing the funnel. If those systems are trained on historical hiring behavior, they can reproduce old patterns very efficiently. No villain required. Just yesterday’s preferences encoded into today’s workflow.

Lachlan Reed

[responds quickly] Historical hiring behavior — that’s the bit with teeth. If the company historically favored one kind of candidate, then the software learns the pattern like a bad apprentice. Nobody has to type “exclude this group.” The system can infer proxies and keep the old habit rolling.

Chris J. Murphy

That concern is no longer hypothetical. There’s growing legal scrutiny around AI-driven hiring systems, including a federal case involving Workday and allegations that software-based screening disproportionately filtered out applicants over 40. One of the names tied to that discussion is Derek Mobley, who reportedly applied to more than 100 positions through systems using Workday and alleged repeated rejection tied to algorithmic screening. Courts are now examining whether software vendors themselves may share responsibility for discriminatory outcomes.

Simon Carver

[grave] Derek Mobley and more than 100 positions — that’s the human scale of this. Not an abstract compliance memo. A person doing it over and over, probably hoping each time this one might be different.

Jacques San Dimas

And the truly dangerous part is invisible bias. If the model is complex, if oversight is weak, if auditing is shallow, then even the organization may not fully understand why certain candidates vanish. In risk terms, this is intolerable. A decision system affecting livelihoods should never be a black box to the people deploying it.

Simon Carver

So let me ask the uncomfortable question plainly: if no one can explain a rejection, no one can appeal it, and no one can even say whether a human saw it... in what sense is that fair? I’m not asking whether it’s fast. I’m asking whether it’s fair.

Chris J. Murphy

[steady] I don’t think automation is neutral just because it’s automated. That’s a story we tell to avoid responsibility. We’ve seen this pattern before with technology adoption: the language is efficiency, consistency, scale. But if a system quietly deprioritizes older candidates, or people with nonlinear careers, or applicants whose salary history suggests they won’t be cheap, then the output may look clean while the process is deeply distorted.

Lachlan Reed

And then you get those slippery phrases — “culture fit,” “not aligned,” “better match.” Mate, “culture fit” can be real, sure, but it can also be a cardboard cutout for all sorts of things nobody wants to say out loud. Too senior. Too costly. Too old. Not our vibe. It’s foggy language doing dirty work.

Jacques San Dimas

[firm] Precisely. If an AI system learns from historical behavior inside organizations that already favored younger, cheaper labor, then eventually the machine does not need direct instruction to discriminate. It learns the pattern itself. This is why governance matters. Human review matters. Auditability matters. Otherwise cost pressure becomes model behavior, and model behavior becomes institutional habit.

Lachlan Reed

“Younger, cheaper labor” — that phrase lands like a brick. Because even if nobody writes age into the code, salary compression alone creates an incentive structure. A company under pressure after layoffs may prefer lower compensation. The model sees who got hired before, who made it through the funnel, what profile correlated with offers... and off it goes. Even a kangaroo could trip over that one.

Simon Carver

[chuckles, then softens] There’s the Lachlan shed wisdom. But he’s right. We keep talking about AI as though bias only counts if someone twirls a villain moustache and types a discriminatory rule into the machine. Real life is murkier than that. It’s inherited preference. It’s proxy variables. It’s optimization goals that sound harmless until they touch a person.

Chris J. Murphy

The real question isn’t what AI can do. It’s what decisions we should never allow it to make alone. Hiring is one of those decisions. Technology can assist human judgment — organize information, reduce administrative burden, maybe even flag overlooked talent. But it should not erase judgment, and it absolutely should not erase accountability.

Jacques San Dimas

[reflective] The workforce was never meant to become a silent data-processing pipeline where human beings disappear into scoring systems. At some point we have to ask: are these systems helping humans work... or helping organizations remove humans faster? That is not anti-technology. It is a moral question about design, intent, and care.

Lachlan Reed

So what does accountability look like in practical terms? For me, a few basics. Tell applicants when AI screening is being used. Don’t keep public postings live if a role is effectively frozen. Give people a path — even a narrow one — to human review. And audit the thing properly, not with a glossy vendor slide deck and a thumbs-up.

Chris J. Murphy

[warmly] Yes. Transparency, oversight, dignity. Those are not nice-to-haves. They are the conditions that make technology legitimate in human systems. Otherwise the future of work becomes a place where people are measured constantly and understood rarely.

Simon Carver

[softly] And maybe that’s the question I want to leave hanging in the room. When someone applies for work, are they entering a relationship with an organization... or are they being fed into a sorting mechanism? Because those are not the same thing, and pretending they are has already cost people more than we admit.

Jacques San Dimas

[warmly] Thank you for spending this time with us.

Lachlan Reed

If this episode gave you something to chew on, share it with somebody who’s job hunting right now — or somebody building these systems.

Chris J. Murphy

And if you liked the conversation, subscribe to the show. We appreciate you being here.

Simon Carver

Take care of yourselves, and of each other. We’ll see you next time.