The hiring process fails at an extraordinary rate — not by accident, but by design. I spent an evening building automation around my job search — scraping, scoring, outreach. Hit a rate limit that didn’t clear for hours, so I went digging. The platform’s public API is effectively dead. The terms of service prohibit job seeker automation explicitly.
But the rules themselves said something worth paying attention to. The platform prohibits automation for candidates. For employers, it is the entire product.
Not a policy detail. That is the business model.
The paying customer always gets the optimized experience
Job boards have one actual customer: the employer. LinkedIn’s Talent Solutions — recruiter tools, job posting, candidate search — accounts for roughly 60% of its $17 billion annual revenue. Indeed markets itself to employers with the line “Indeed reaches 96.7% of online US job seekers” — the candidate is the reach metric being sold, not the person being served.
Job seekers use these platforms for free. And in any two-sided marketplace, when one side pays and one side doesn’t, the paying side gets the optimized product.
The result is a platform where employers get systematic access — APIs, ATS integrations, automated screening, AI-powered sourcing. Candidates get a search box and a submit button. You cannot be discovered without a paid recruiter searching for you. You can only apply, manually, one posting at a time, into a process you have no visibility into.
Fuel. Not client.
I could have finished the automation anyway. Stayed in the gray zone, waited out a ban that might never come. I chose not to — but that was a decision about how I want to operate, not a technical limit the platform imposed on me. The platform never built anything for candidates that would make that choice difficult.
AI arrived — on the employer hiring process side first
When AI entered the hiring process, it entered through the employer door. ATS systems that parse and rank resumes before a human sees them. Automated rejection emails. Screening tools that filter on keyword density. The candidate-facing AI products came later, and they came with a different structure: you pay for them yourself to compete harder in a system already tilted against you.
The effect was predictable. AI made writing CVs easier and applications faster to submit. LinkedIn saw a 45% increase in applications submitted through the platform in a single year — 11,000 per minute. 64% of recruiters reported a flood of nearly identical, AI-optimized resumes that actually increased their screening workload.
Employers responded with more walls. Stricter ATS filters. More screening stages. More human gates.
This is the arms race: candidates use AI to get past filters; employers build better filters. Each cycle produces more volume, less signal, and longer time-to-hire. The platform collects fees at every point of contact regardless of outcome.
The volume is a hiring process design signal, not a market signal
The platforms built low-friction applying deliberately. More applications mean more data, more engagement metrics, more justification for employer spend. What they did not tell candidates is that taking that logic one step further — automating the process entirely — is a terms of service violation. The ease they built into the product becomes a rules violation the moment a candidate tries to systematize it.
The side effect of all that deliberate friction removal is that a single posting now absorbs hundreds or sometimes thousands of applications, and most people read that as a sign of fierce competition in the market.
This is not a market signal. The platform designed it to produce exactly this. High volume is a feature of the business model, not a failure of it.
What that volume actually costs employers
74% of employers — nearly three in four — admit to having hired the wrong person for a position. That number comes from a CareerBuilder survey of over 2,000 hiring managers, independently confirmed across multiple studies. 41% said a single bad hire cost them over $25,000. Harvard Business Review puts 80% of all employee turnover down to poor hiring decisions — not performance issues, not market conditions, just a broken hiring process. The U.S. Department of Labor estimates the direct cost at a minimum of 30% of that employee’s first-year salary, before counting the productivity drag.
Managers spend an average of 17% of their working week supervising, correcting, and documenting the mistakes of a wrong hire. Nearly one full day every week, consumed by a decision that shouldn’t have been made.
The problem runs in both directions. 66% of workers say they’ve accepted a job only to realize it was a bad fit — half of them quit within six months.
The mismatch is mutual. The platform collects its fee either way. And every failed hire eventually cycles back through the same system that produced it.
The human guard is not the solution
The logic behind adding humans to the screening process is that judgment catches what the algorithm misses. In practice, the human guard is a non-technical recruiter matching keywords on a CV against a job description they may not fully understand, under pressure to move volume.
Developers know this pattern well. Recruiters pitching PHP roles to engineers who haven’t touched PHP in a decade. Screening calls that test for stack acronyms instead of problem-solving ability. Candidates with nine years of relevant experience rejected because someone didn’t tick a checkbox correctly.
The pattern repeats constantly: a candidate clears the HR screen, then gets rejected by the technical lead who actually understands the role. Or gets passed by HR and the engineering team spends the next two weeks rebuilding the brief because whoever screened the CV didn’t understand what the job required. Each stage in the process adds a person operating on incomplete information, making a judgment call they may not be qualified to make.
The platform language is not hypocrisy — it’s a category error
Job boards say they exist to help people find work. Creating opportunity. Connecting talent to companies. The language is everywhere, and it sounds genuine.
It is not dishonest. It is just aimed at the wrong person.
When these platforms say “people,” they mean the employer. That is the entity with a problem worth paying to solve — a gap in the team, a deadline, a cost. The job seeker is not the problem being solved. What the job seeker actually is: millions of profiles, searchable, filterable, rankable — the raw material that makes the employer’s problem solvable. The job seeker’s experience of the platform matters little to the business, because the job seeker is not paying for anything.
The mission language and the product decisions are consistent. They just describe a product built for someone other than you.
What comes next is not another job board
The current state is the ugliest version of a transition. The old model — post a job, collect CVs, screen manually — is still alive. AI is arriving into it without replacing the underlying logic. The result is more volume, more noise, more cost, less accuracy, and a 74% failure rate that every finance department eventually has to explain.
Business counts money. That number is too large to ignore indefinitely.
A hiring process built around the actual problem
The model that replaces this one doesn’t look like a better job board. It looks like the person who actually owns the hiring decision — the team lead, the founder, the department head — uploading a description of a business problem — not a job title, a problem — and an AI system that knows the difference between what the job posting says and what the role actually needs, then finds the people who fit that. No keyword matching, no non-technical screening gate, no intermediary layer running on guesswork and volume incentives.
The platforms that profit from the broken cycle will call this a threat to quality. What they mean is a threat to their fee structure.
The HR layer that currently acts as a pipeline between decision-maker and candidate will shrink — not because AI is taking jobs, but because the people signing off on headcount will eventually look at what bad delegation is actually costing them. Delegation was always the problem. The platform just made it easy not to notice for a long time.
Right now, the old model still has enough inertia to keep running. Job boards are still collecting fees, the ATS is still filtering on keywords, and recruiters are still sending PHP offers to engineers who hate PHP.
Nobody actually likes this hiring process — not candidates, not employers, not even the hiring managers signing off on decisions they’ll regret three months later.
The only constituency that benefits from keeping it exactly as it is — is the one collecting fees every time it fails.
