Key takeaways
- The cost of producing an application fell by an order of magnitude; everything else follows from that.
- More applications per posting means less attention per application and stricter, more literal first filters.
- Documents now carry content but no longer carry evidence of effort.
- Employers respond by weighting what is expensive to fake: samples, referrals, structured assessment.
- Logistics changed enormously; the substance of why anyone gets hired did not change at all.
Applying became close to free
The direct change is that producing an application now costs minutes rather than hours. Finding roles, tailoring documents and submitting are all substantially automated.
When the cost of an action falls by an order of magnitude, the volume rises accordingly. That is the change everything else follows from.
It is worth being precise about what became cheap, though, because it was not all of it. Producing and submitting became cheap; being suitable, being credible and being remembered did not, and the whole set of second-order effects comes from that gap widening.
Employers receive more and read less
More applications per posting means less attention per application. Screening becomes more automated in response, and the initial filter becomes stricter and more literal.
For candidates that means clear, checkable evidence against stated requirements matters more, and a well-written narrative matters less at the first gate than it used to.
Some employers have gone further and simply stopped advertising certain roles. If a posting reliably produces eight hundred applications of which most are automated, filling the role through a network becomes the cheaper option — which shrinks the open market precisely for the people who depend on it.
- Higher volume per posting
- More automated first-stage screening
- More weight on explicit requirement evidence
- More scrutiny of applications that look generated
- More roles filled before they are meaningfully advertised
| Stage | Then | Now |
|---|---|---|
| Finding roles | A few boards, manually | Many sources, continuously |
| Producing an application | One to two hours | Minutes |
| Applications per posting | Dozens to low hundreds | Often many hundreds |
| First-stage screen | A person, partially | Software, almost entirely |
| What the document proves | Content and effort | Content only |
| What decides the shortlist | The application | Evidence outside it |
The documents carry less information
A CV and cover letter used to indicate effort and care as well as content. Now they indicate content, and even that is harder to trust when the text was generated.
Employers respond by weighting things that are harder to produce: work samples, referrals, structured assessments, and conversations. The document becomes an entry ticket rather than the evidence.
This is the same pattern that plays out whenever a signal becomes cheap to produce. The signal is not improved, it is replaced, and the replacement is always something more expensive — which is why hiring processes have got longer rather than shorter despite the screening being faster.
Where candidates are genuinely better off
Discovery improved substantially. It is far easier to find relevant roles across many sources than it was, and far easier to keep track of a search.
Screening automation also means more applications actually get read, which is a real improvement over a human reading the first fifty of four hundred.
The quieter gain is in who can compete at all. Writing a strong application in a second language, or in a professional register nobody ever taught you, used to be a barrier that had nothing to do with capability — and a substantial part of that barrier has simply gone.
The effects that are not yet settled
Several things are clearly in motion without having landed. Whether employers converge on work samples as the standard first filter, whether the open application recovers or keeps shrinking, and whether screening automation is regulated in the way other automated decisions have been.
Verified credentials are another open question. If there were a cheap, trustworthy way to confirm that someone holds a qualification or did the job they claim, much of the current problem would dissolve — and several attempts at this exist without any of them having won.
It is worth holding these lightly. The confident predictions made when the first screening systems arrived mostly did not happen, and the same restraint is warranted now about anything described as inevitable.
What has not changed at all
Employers still hire people they believe can do the work and want to work with. Referrals still outperform cold applications by a wide margin. Being genuinely well suited to a role still matters more than how the application was produced.
Agents changed the logistics of a job search considerably and its substance very little. Treating them as a way to do the logistics faster is the use that holds up.
Which suggests a fairly unglamorous strategy: let the tools handle everything mechanical, and put the recovered time into the things that were always decisive — being genuinely good at the work, being visible to people who hire, and being able to explain both.
Frequently asked questions
What is the main change AI agents have made?
Applying costs minutes rather than hours. Everything else — higher volume, stricter automated screening, less attention per application — follows from that.
How has screening changed for candidates?
More applications get read, but the first gate is more literal. Clear checkable evidence against stated requirements matters more than a well-written narrative.
Are candidates better off overall?
In discovery and tracking, clearly yes. In the odds on any individual application, less so, since employers adjusted in response to the volume.
What has not changed?
Employers still hire people they believe can do the work, referrals still outperform cold applications, and being genuinely suited still matters more than how the application was produced.
Why are hiring processes getting longer if screening is faster?
Because a cheap signal is replaced rather than improved, and the replacement is always more expensive — work samples, structured assessment, more conversations.
Who benefited most from the change?
Candidates for whom writing was a barrier unrelated to capability — a second language, or a professional register nobody taught them. That barrier has largely gone.
Check this against your own resume
Scan your CV against a real job description, or build a parse-safe one from scratch. Your first scan costs nothing.