The Machine Remembers You: Workforce Dirty Secret No. 1
How AI hiring systems track and score you across every application, what your data shadow carries, and three moves to route around the machine.
In the 1960s, when you applied for a loan, a job, or an insurance policy, a company most people had never heard of yet often knew more about you than your neighbors did, providing the data on which others decided your fate. The Retail Credit Company of Atlanta kept files on millions of Americans, and those files held far more than payment history. They recorded your reputation, your habits, your household, your associations, all gathered by field investigators and sold to the employers, lenders, and insurers deciding your future. As an individual, you could not see the file, so you could not correct it. Nonetheless, it followed you like a judgy mother-in-law detailing your worth. Eventually, public outrage and a run of congressional hearings finally dragged the system into daylight, and in 1970 the Fair Credit Reporting Act gave Americans the right to see, dispute, and correct what was written about them. And the Retail Credit Company of Atlanta, now operating under new scrutiny, quietly rebranded as Equifax.
Here we are today. After we dragged one secret dossier into daylight, we built another, faster, and aimed it at the hiring line, among many other things.
Apply for a job today and you are applying to a system as much as to a company. Ninety-nine percent of Fortune 500 employers run an applicant tracking system, and a handful of vendors own most of the market. Workday alone screens hiring for roughly four in ten of the largest companies in the country, and a few more names cover most of the rest. Send out fifty applications and you are not knocking on fifty doors. You are passing through the same three or four machines, each one reading you the same way.
By some estimates three-quarters of résumés are scored so low that they never reach a human. The score is a composite of:
1. how closely your words match the posting,
2. how your work history parses,
3. where your employment gaps fall, and
4. how closely you resemble the people the employer has already hired.
That last factor is where the bias lives. Age, race, and disability ride inside it. Nobody types them in but the training data encoded them, and the model repeats them.
I keep hearing this dismissed as theory, or paranoia. The proof is in the courts. In Mobley v. Workday, certified as a nationwide collective in 2025, applicants over forty allege the AI screened them out across employer after employer; Workday’s own filings acknowledge more than a billion applications ran through the system in the relevant years.
In Kistler v. Eightfold AI, filed in 2026, applicants allege the company scraped data on more than a billion workers, scored each of them from zero to five, and discarded the low-ranked before a human ever looked. Eightfold’s clients include Microsoft, PayPal, Morgan Stanley, and Starbucks. A number, zero to five, attached to a billion people, deciding who is worth a glance. The Retail Credit file has a great-grandchild, and it computes.
The hardest part is what the machine does across the whole market at once. A recently published Stanford study, the first to examine hiring algorithms at scale, followed 3.4 million people through four million applications screened by a single vendor’s AI. The researchers found what they call an algorithmic monoculture: when one system judges you, its verdict travels. People who applied to several jobs screened by the same vendor were rejected far more often than chance would predict, and one in ten who sent four applications were turned away from all four. So wherever that vendor’s AI was in use, you were not getting a fresh read. The same logic ran, and reached the same verdict. The same tool recommended Black and Asian applicants at lower rates across many roles. When the researchers looked at hiring data from before AI screening, the shut-out-everywhere pattern was non-existent.
What we have now is consistency. Consistent bias, built into programs built by the few for the many, with few to no guardrails and a race to deploy that scales the error. Human fallibility, baked into the original model, felt across the whole market at once. The machine delivers the same biased outcome at every employer that licenses it.
Can we reset the score?
People ask how to reset the score, whether a fresh email or a new browser wipes the record. Not entirely, because the thing that follows you is not a login. Inside a single employer’s system, your profile is tied to your account, so reapplying surfaces your history there. Across the market, your data shadow travels: the scraped work history, the parsed résumé, the public profile the models feed on. You cannot log out of that. The machine remembers you. The law may eventually catch up. One argument running through these cases holds that an AI hiring score is a consumer report and belongs under the same Fair Credit Reporting Act that pried open Retail Credit half a century ago. For now, the machine keeps its files, and it does not take corrections.
There is no direct path to resetting the score, and chasing one misses the point. Resetting the score treats a symptom. The reality is a system built to remove the human from the hiring process, and the productive response is to put the human back. Three moves:
1. Apply on purpose, not just online. On purpose means something more specific than a list of companies you want to work for. It starts with understanding why you want to be there: what in their priorities and public work genuinely intersects with yours, and how you communicate that alignment. That intersection is what makes your outreach coherent and your name worth remembering before you ever send an application. Choose the company and the person, speak in the language of what they and you already champion, and let that alignment lead the conversation. A hiring manager who can see why you belong in their work reads your name differently than a stranger's. A name, with a voice, travels a different route: requested, referred, pulled from the pile by someone who went looking. Rapport opens a door the machine does not guard, and alignment is what builds it.
2. Stand in the right rooms. A name with a voice still needs somewhere to land. Classes, cohorts, conferences, working groups: these are the addresses where your targets learn, debate, and solve problems alongside each other, and where the opinions that generate referrals get formed. Proximity is a channel the algorithm does not control, and most senior roles still move through people who can vouch for you in one.
Getting in is not always a matter of registration. Some rooms reward attendance; others require you to present, apply for a competitive slot, volunteer for a committee, or build a credential that earns the seat. The entry path is part of the strategy. Once you are in, the goal is one real relationship, someone who knows your work well enough to say your name first when the right role surfaces.
When the front door is locked by the algorithm, it is time to use the side door. The GWHQ Side Door is a curated community board for finding the gatherings, cohorts, and working groups worth entering. Start there.
3. Reframe your Mindset. All of this is relationship work, and as anyone continuously on the dating app grind will tell you, instant gratification in relationships is fleeting to non-existent. Your career search will not materialize with a swipe or a submit button. Set your expectations accordingly, and find joy in the build: a growing presence, a compounding network, and conversations that open rooms you did not know existed.
Five years ago, I told my daughter not to be rude to our kitchen Alexa. I told her that just as she learns by absorbing what I do and say me, the machine was learning her, cataloguing every pattern she gave it. The kitchen was only where it started. The same logic now sits at the hiring gate, and it remembers a version of you that you did not write and cannot read. The antidote is to build, live, and promote the version you can: a presence so clear and so human that, by the time the algorithm files you, the people who matter have already met you. That is the work, and it is yours to design.
To start, I’m sharing my Build Your Personal Brand workbook: the prompts that turn this into a plan, including what you stand for, the companies and industries you’re aiming at, and the rooms worth walking into. The machine keeps its file. This week, start keeping yours.
The aspiration machine was built on the same foundation as everything else in this series — the labor of people whose contribution was captured, rebranded, and sold back to them as something to want. Understanding that history is not academic. It is the most practical thing a working professional can do. Career Communiqué exists to make that history legible. Founding subscribers make that work possible.