The hiring process most companies use today was designed for a world that no longer exists. The resume format, the ATS, the keyword filter, the shortlist — all of it was built in the 1990s for paper applications, local newspaper ads, and recruiters who might see 20 candidates for a role over three weeks.

None of those conditions hold anymore. And yet the stack that was built for them is still the stack most companies are running.

That's not a maintenance issue. It's something bigger. The whole system broke silently over the last decade, and AI is the first thing to make the break visible.

How We Got Here?

The first Applicant Tracking System worth the name was Taleo, which launched in 1999. Its core innovation was simple: parse incoming resumes, extract text, match that text against keywords from the job description, rank candidates by keyword density. This was genuinely useful in 2001, when a recruiter might get 30 resumes for a posting and needed a fast way to flag the top ten.

That original design has barely changed. Every major ATS since — and there are more than two dozen of them now — essentially runs the same pattern. Parse the resume. Extract keywords. Rank. Surface the top candidates. The interfaces got prettier. The logic stayed the same.

This worked for twenty years because the volume was manageable and the candidates were writing resumes by hand.

Neither is true anymore.

The Scale Problem

The average corporate job posting now receives 250+ applications. Senior roles often clear 500. Engineering roles at well-known companies regularly exceed 1,000. Remote roles can push past 2,000.

There's nothing wrong with a lot of applications. The problem is what it does to a system designed for 30.

At 30 applications, a keyword match is a useful filter. It cuts your review from 30 to maybe 8, and you read those 8 carefully. At 250, the keyword match cuts you from 250 to maybe 60. You can't read 60 carefully. You scan them, miss things, and fall back on proxies that don't actually predict performance — company names, school names, job title pattern-matching.

The signal that made the filter work — a recruiter reading the top of the pile carefully — doesn't survive the volume.

This has been true for about a decade. Companies coped by hiring more recruiters, sharpening their boolean strings, adding more filters. None of it fixed the underlying problem: the screening layer was built for scarcity and is now operating under abundance.

The GenAI Accelerant

Then, in late 2022, ChatGPT happened. And from that moment, the other side of the system — the one producing the resumes — also started running on AI.

Today, 50%+ of job seekers admit to using AI to write their resumes. The real number is almost certainly higher, because people don't love admitting it. Tailored resumes that used to take 30 minutes of thoughtful writing now take 90 seconds of prompt engineering. AI reads the job description, mirrors its language, and produces a resume optimized to rank well in the exact keyword-matching system that's about to read it.

You can see where this goes.

The screening side reads resumes with AI. The candidate side writes resumes with AI. Both are optimizing for each other. Neither is optimizing for whether the candidate is actually good at the job. The human signal — the thing the whole system was trying to surface — is getting washed out in an AI-vs-AI arms race that nobody is winning.

There's a term for what this does to a resume: "ATS bloating" — the inflation of a resume with keywords and AI-generated content designed to game the ATS layer. It isn't a moral failure by candidates. It's a rational response to a system that rewards it. If the filter reads keywords, candidates will produce keywords. If the filter reads more keywords, candidates will produce more keywords. That's the whole story of the last three years in resumes.

The 2028 Problem

If you think this is bad now, the forecasts are stark. Industry research suggests that by 2028, 1 in 4 applications will be fraudulent — not just inflated, but materially misrepresenting experience, credentials, or identity. Companies are already reporting candidates who pass phone screens and then fail basic technical tasks, because the person on the call wasn't the person applying for the job.

The ATS was never designed to handle this. It was designed to filter on keywords. Keywords are exactly what the bad actors are best at producing.

So What Actually Changed?

Here's the honest answer, and it's more radical than "AI is in hiring now."

The system everyone is running was quietly broken for a decade and is now obviously broken. Keyword matching doesn't work at 250 applications. It especially doesn't work when half of those applications were written by a tool specifically designed to beat keyword matching. The ATS as a filter layer has effectively stopped doing the job it was designed for.

What's emerging isn't a patch. It isn't a better filter or a smarter keyword engine. It's a different paradigm entirely — one where the screening layer actually reads the resume the way a good recruiter would, looking for the story, the progression, the plausibility. Not the words.

Calling this "AI in hiring" undersells what's happening. It's more accurate to say the thirty-year-old hiring stack is being retired, and the one replacing it works differently enough that most of the old assumptions need to be rebuilt from scratch.

The next generation of hiring tools isn't patching the old system. It's replacing it with something that reasons, not just matches.

That's the change. Everything else in this curriculum is downstream of it.

Next: Lesson 03 — From Shortlisting to Orchestrating