Getting the go-ahead to hire is supposed to be good news. Budget approved, help on the way, someone to take work off the team.

Then the applications start arriving. 200 by Friday. 2,000 by the end of the month. And the person who has to read them is doing it at 9pm, because their day was full of everything else.

Having too many job applicants is supposed to be a good problem. Nobody who’s actually had one describes it that way.

And the numbers back that up. According to Employment Hero’s Work That Works report, three in four UK business leaders say recruitment is a challenge, and 45% are spending more on talent acquisition than in previous years. Hiring managers still report a talent shortage, while sitting on more CVs than they can physically read.

The complexity is easy to measure, and it’s moving fast. In a YouGov survey of 1,016 UK business leaders commissioned by Employment Hero in April and May 2026, employers rated the difficulty of hiring at 6.2 out of 10, against 4.7 three years ago. That’s a 32% increase in three years.

Simon Ryan, Managing Director at Chapter 2 summed it up well when speaking on a recent Employment Hero hiring panel. Simon coined “the applicant avalanche”, and traced it to AI-generated CVs and one-click applications arriving on the HR and recruitment scene at precisely the same moment.

The cost of being buried in the avalanche? The hours it takes to work out who’s actually worth speaking to. As one UK employer told us at a recent HR event, they don’t have the capacity to endlessly sit sifting through CVs.

And here’s the biggest part that should worry any employer reading this. The talent shortage and the application pile are the same problem. 

Employment Hero research with One Poll, surveying 1,000 UK working adults in November 2025, found that 61% of workers say the job searching or hiring process itself discourages them from looking for a new role. Among 18 to 34-year-olds that rises to 69%, and among women to 67%.

On paper, there’s no shortage of people, but there’s a shortage of people still willing to take part.

More applications have meant less information

The mechanics are simple enough. One-click apply removed the friction from applying. Generative AI removed the effort from writing a CV and a covering letter. And both were, individually, good for candidates.

But together they produced something neither intended… Applications now arrive fluent, keyword-matched, tailored to the advert, and almost entirely indistinguishable from one another.

Volume went up but real hiring signals went down.

That distinction matters more than the raw numbers, because it explains the contradiction. Employers aren’t short of applicants. They’re short of useful, distinguishing information about applicants. A hundred well-written CVs that reveal nothing leave you worse off than twenty rough ones that reveal something, and they take five times as long to sift.

It works both ways, and candidates are feeling the burn too. Employment Hero’s One Poll research also found the following:

• Workers judge only 38% of the roles surfaced in job platform searches to be relevant to them, a figure that barely shifts across age groups
• A quarter suspect they have applied for a ghost job, rising to 37% among 18 to 24-year-olds

Research shows that candidates aren’t applying indiscriminately because they’re careless. They’re doing it because matching is typically broken at source, and volume is the rational response to a system that surfaces the wrong roles.

What employers are actually doing about it

Three responses are common. Two of them make the problem worse.

Response one: Read everything. Someone commits to reviewing every application by hand. This is the honourable option but with an avalanche to dig through it’s completely unsustainable.

Daniella Angel, Talent Acquisition Manager for the Northern Hemisphere at Employment Hero, did exactly this on a role that brought in approx 4,000 (3,994!) applications a few months ago. 

Daniella recalls reading every one, in the evenings, across six to ten weeks, while recruiting for five other roles and leading a team across three time zones. Forty-six applicants reached a screening call. And only one reached the finalist stage.  

She spent between 57 and 67 hours on it but could only review CVs in the evenings because her days were full of calls.

This was a next-level commitment from Daniella and not a regular or sustainable expectation. For many, this same scenario looks like reading the first 80 and abandoning the rest—the worst of both worlds, because the 80 are ordered by arrival time rather than by quality.

There’s also an issue that doesn’t appear on any timesheet. 

“The thing people don’t see is the queue,” Daniella says. “When you’re working through a pile systematically, the people at the bottom might be sitting there for days or weeks waiting on you. That’s a constant anxiety ticking over in the background.”

That queue has a measurable effect at the other end. In the same One Poll research, the single most frustrating part of job hunting was applying and hearing nothing back, shared by 54% of workers. 

Eight in ten have applied for a job and never heard anything at all. One in four have been left waiting on five or more applications. Once people are in a process, long delays between interviews frustrate 44%, and 42% report being ghosted after interviewing.

None of that is anyone deciding to treat candidates badly. It’s what happens when the sift outgrows the hours available to do it.

And it doesn’t hold up as fairness either. Screening consistency drops sharply after roughly 20 to 30 CVs in a row, from decision fatigue alone. Daniella is blunt about what that means. 

“If consistency drops off a cliff after the first 30 CVs, then fair is not what’s happening to applicant number 3,000.”

Response two: raise the proxy bar. 

Filter to a shortlist of universities, or require a recognisable employer on the CV. It’s fast, and it feels rigorous. But only one of those is true. It excludes people who built genuine capability through apprenticeships, side projects, startups and self-teaching, and it doesn’t reliably predict performance. 

A big-name employer on a CV tells you someone got through that company’s hiring process, not that they will succeed in yours. CIPD’s own figures point the other way. The large majority of UK employers now say they prioritise skills over formal qualifications. Filtering by pedigree in 2026 is filtering by an attribute you’ve already said you don’t rank.

Response three: let the role drift. 

The role stays open. The sifting keeps sliding down the list. Weeks pass and nobody actually decides to slow the hire down, it just slips behind whatever else is on fire that day.

This is the one that shows up as a “talent shortage” in the surveys while 400 applications sit there unread. It’s also the most expensive. The median UK time to hire is 40 days and the median cost per hire is £6,125, and neither of those numbers improves while the pile sits there. You’re carrying the cost of an empty seat and calling it bad luck.

There’s evidence this is already happening at scale, though cost is doing much of the driving. The same YouGov research found full-time employment costs have risen by almost 10% in a year, and in response 39% of UK businesses have reduced full-time recruitment, while 25% are turning to freelancers and contractors instead. When hiring is both more expensive and more work, the roles that get opened are the ones somebody has the capacity to fill.

So the obvious question is: Why haven’t more employers already handed the sift to AI? The tooling exists. And many HR platforms now offer some version of it.

The answer is that plenty of people have tried it and didn’t like what they found. But the technology is changing fast and so are the benefits. 

The things that actually help

Notice all three responses above accept the same premise: that screening is a fixed cost you either pay, dodge, or approximate. 

But it isn’t. In fact, screening is the part of hiring that’s changed most in the last two years, yet most UK teams are still doing it exactly as they did in 2021.

Here’s what changes when you stop treating it as fixed and introduce AI. Every applicant answers the same structured questions, against the same criteria, and you see the result before you’ve read a single CV. That’s the same evidence you were going to gather anyway, just arriving at the start instead of the end.

Three things follow from that. Every applicant is assessed against the same standard, rather than against whoever happened to read them and whatever mood they were in. The ones who look wrong on paper but right in person can surface. And the reading time stops scaling with the application count, which is the link that’s been breaking hiring since 2022.

And here’s the bit Daniella didn’t see coming. Automating the front of the process moved her first human contact earlier, not later.

“The AI interview replaces the faceless CV as the first real contact. Candidates get their faces and their answers in front of me sooner, in their own time.”

This is what an AI Recruitment Agent powered by Hero AI does. It scores and ranks every applicant against criteria you set, runs a structured first-round screen, and hands you a shortlist with the reasoning attached. You set the rubric, you edit the questions, you make every call meaning the judgment stays with you.

You also don’t have to replace your HR system to do it. Plenty of UK teams have an ATS they renewed recently, or an HR platform they’re happy with, and no appetite for a migration to fix screening. There’s a standalone option that connects to the job boards you already post on.

Daniella went from spending 10 to 13 hours a week on CV review and screening calls to between 2 and 3. For a role like her Canadian scenario above, she worked out it’s the difference between a six-to-ten-week slog and something that could wrap in one to two.

And Employment Hero customers are seeing the same thing:

• Thermosash Group reported  99% faster screening and 90% less recruitment admin.
• While Alchemy Saunas went from two to three hours per applicant to just seconds.

What UK employers are actually asking about AI screening

Before getting to the fix, the objections deserve straight answers, because they come up in almost every conversation we have with UK employers weighing this up. 

These are the six we hear most:

1. “Is it as good as somebody actually screening?”

The answer is that it’s good at a different thing. It isn’t making a hiring decision, and it shouldn’t be. It’s ordering a pile so a person can spend their judgment where judgment matters. 

The comparison people imagine is AI against a fresh, focused recruiter on their first CV of the day. The real comparison is AI against that same recruiter on their 400th CV of the week, at 9pm. Daniella explains: 

“It’s the difference between being shown the aces up front and flicking through the whole deck.”

Like with any technology the learning curve has been steep.  

“In the early days, the AI screening tools I encountered were at the start of their evolution and the scoring was usually inaccurate… Back then, I felt confident I could ignore what the AI was suggesting, because my own judgment was better,” Daniella describes.  

What’s changed is twofold. The technology has moved on a long way from those early tools. And, just as importantly, what she asks it to do has changed.

“I don’t expect it to decide who gets seen. I expect it to point me in the right direction early and I make the human judgement from there.”

The practical difference shows up in who she ends up meeting. Manual screening meant a fixed number of calls a week, squeezed around three time zones, so candidates got cut quickly to fit the diary. Now candidates record in their own time and nobody is chasing a slot.

The upshot is counterintuitive. She looks at more people, not fewer, because she can hear them instead of reading them. The applicant who writes a poor CV but interviews well finally gets a look.

2. “Does the AI filter the candidates, or is that done manually by the recruiter?”

This is the one that matters most, and it’s the question to press any vendor on. 

Employers don’t necessarily want AI sifting people in and out for them. That’s the right instinct. Scoring should narrow and order. A person should decide. 

In Daniella’s process, strong matches are shortlisted and invited to interview, but lower-scoring candidates aren’t discarded. She still reviews their CVs, just faster, because a mismatch is quick to confirm.

3. “What actually happens in an AI interview?”

This is where most of the resistance and fear lives among Talent Professionals , and it’s usually based on something people have read rather than something they’ve seen.

It’s closer to a video interview than to anything science-fictional. The candidate gets a link, records their answers to a set of structured questions in their own time, and the follow-up questions adapt to what they say. No diary wrangling, no 8am call squeezed around their current job, no travelling to an office for 45 minutes.

“Think of it as replacing the telephone screening, not the interview itself,” says Daniella. 

That first call was never the moment anyone got hired. It was the moment someone confirmed the basics and formed a quick impression. The difference is that every candidate now gets the same questions, at a time that suits them, and their answers get watched by a person rather than summarised from memory by someone taking notes on their 10th call of the day.

4. “What candidate feedback have you had on AI interviews?”

Asked constantly, and fairly. The answer isn’t universal. 

“It might not land everywhere,” Daniella says. 

“We’re hiring in Romania at the moment and the software engineers there are lukewarm on it. They’d rather have a human screen first. Sales candidates are the complete opposite; they love it!” 

Any employer adopting this should be reading the room per market and per role rather than switching it on everywhere. Daniella’s own red line: “I’d stop using it tomorrow if candidates hated it, thankfully that’s not the case.”

5. “What stops candidates using AI to answer the AI?”

This comes up constantly, and it’s worth turning around: does it actually matter?

If someone uses AI to tidy up their CV or prepare for an interview, they’ve done what any sensible person does with a tool that’s sitting there. Plenty of candidates have good reasons beyond convenience. English as a second language. Dyslexia. A brain that works better with a bit of structure. Screening those people out isn’t rigour, it’s just a different kind of bias.

What you actually need to know is whether the person in front of you can do the job. A recorded interview answers that better than a CV ever did, because you’re watching someone think, not reading something they polished for a fortnight.

So the question worth asking a vendor isn’t “how do you detect AI.” It’s “will I be able to see the person.” Structured questions you wrote. A recording you can watch yourself. A real conversation before anyone gets an offer.

6. “How customisable is it? Can I set my own criteria?”

If the answer’s no, it’s a black box, and you should walk away.

This was Daniella’s condition too, before she’d agree to use anything. 

As a hiring and talent professional you should expect to write or guide the interview questions and set the scoring criteria. 

Daniella explains:

“The job description and scorecard candidates get matched against are mine. I created them and I still own those important guardrails.”

The EU AI Act is starting to come up more and more in conversations too, and it’s worth a check on your own side. Most businesses have a GDPR privacy notice. Far fewer have anything written down about how they use AI in hiring.

If that sounds like you, write the policy first. It’s a much easier job before the tooling goes in than after. It doesn’t need to be long, either: who gets assessed by AI, at what point they’re told and how you’d explain a decision if someone asked.

Where this leaves the avalanche

The avalanche isn’t going to stop. One-click apply isn’t going away and AI-written CVs are now the default rather than the exception. Application volume per role is the new baseline rather than a spike to wait out.

What can change is whether volume translates into cost. Right now, for most UK employers, it does. Every extra applicant costs you hours, so every extra applicant is another reason not to advertise. That’s the link worth breaking. Once the 200th application costs you the same as the second, a big pile stops being a problem and goes back to being what it was meant to be, which is choice.

Fifteen years of doing it the hard way gave Daniella one sentence. “The grind wasn’t the job. It was the workaround.”

AI screening: Try it on one role

If any of this is familiar, the useful next step is small: Pick one role, the one with the worst pile, and run it differently. 

“Working without the AI Recruitment Agent or AI screening is a bit like shopping under stress,” Daniella describes.

“You want to scan the whole shop before you decide what to buy, so you know your shortlist genuinely is the best of the pile. With 4,000 applications, that luxury disappears.”

Employment Hero’s AI Recruitment Agent scores and ranks every applicant against criteria you control, runs structured first-round video interviews in the candidate’s own time, and hands you a shortlist with the reasoning attached. You review it. You make every call.