2024 was a year of significant change in recruiting, driven primarily by technology.

Two forces defined it. The talent shortage hit a record high, with three quarters of employers worldwide unable to fill the roles they had open. And artificial intelligence arrived in recruiting for real, though at roughly a quarter of the scale the widely circulated statistics claimed.

That second point is why this report exists in the form it does. Most 2024 recruitment statistics you will find online were copied from other statistics pages, not from the surveys underneath them, and the copying introduced errors that are still being repeated today. What follows is the 2024 data on hiring, workforce trends, and technology adoption in recruitment, traced back to the primary sources wherever a primary source exists. It closes with a section on what actually happened next, because these numbers now have a two-year track record to be judged against.

How to Read These Numbers

The single most useful thing to know about recruitment statistics is that most of them are laundered. A vendor blog cites a statistics aggregator, the aggregator cites another aggregator, and somewhere at the bottom of the chain there is either a real survey with a real sample, or there is nothing at all. By the time the number reaches a conference slide, the two are indistinguishable.

The clearest example from 2024 is the claim that 87% of companies use AI in recruitment. It appears on dozens of sites and is usually traced to Demandsage, a statistics aggregator, which is not the same thing as a survey of employers. Over the same period, SHRM asked more than 2,000 working HR professionals whether their organisation used AI for HR at all, and 26% said yes in 2024. Both figures cannot describe the same world. One of them has a published sample size and field dates. The other does not.

Three habits protect you from most of this. Prefer sources that publish their sample size and field dates (SHRM, ManpowerGroup, Gallup, CompTIA and national statistics offices all do). Watch the denominator, because "14% of workers are fully remote" and "26% of remote-capable workers are fully remote" are both true and describe different populations. And treat any figure without a traceable origin as marketing, however often you have seen it. Several numbers that appeared in the first version of this report, including a widely repeated $152,000 cost per tech hire, did not survive that test and have been removed rather than repeated.

The State of Hiring 2024

The defining feature of hiring in 2024 was scarcity that had stopped improving. Employers were not struggling because the economy was overheating, they were struggling because the supply of qualified people had structurally thinned, and the data showed that pattern holding across almost every developed market simultaneously.

The headline number came from ManpowerGroup, which surveys a very large employer panel every year and publishes both the sample and the method. Its results describe a market that had been tight for long enough to become normal.

• 75% of employers worldwide reported difficulty filling roles in 2024, a record in the survey's history. The figure comes from ManpowerGroup's Global Talent Shortage report, based on 40,077 employers across 41 countries.
• The shortage was worst in Japan (85%) and lowest in a handful of large emerging markets. The United States sat at 70%, slightly below the global average (ManpowerGroup).
• Healthcare, IT and logistics were the sectors reporting the most acute difficulty, and employers responded mainly with flexibility (65%) rather than pay, with only 30% raising wages (ManpowerGroup).
• Prime-age labour force participation (ages 25 to 54) averaged 83.6% in 2024 and peaked near 83.9% over the summer, the strongest readings in more than two decades - FRED.
• Overall US labour force participation was 62.6%, and BLS projections have it falling to 61.1% by 2034, roughly 4.3 million fewer workers than if it held steady - Indeed Hiring Lab.

Those last two lines are the ones worth sitting with, because they explain why the shortage is not a cyclical inconvenience. Prime-age workers were participating at close to a generational high, which means the obvious slack had already been absorbed. The pressure on the overall rate comes from demographics rather than from discouraged workers: the population aged 55 and over is growing while participating at roughly 37%, and the 16 to 24 cohort is both shrinking and participating less. There is no policy lever that reverses that quickly, which is why the shortage persisted through both a hiring boom and a hiring slowdown.

The long-range framing most often quoted here is Korn Ferry's Global Talent Crunch, which modelled a shortfall of 85.2 million workers and $8.5 trillion in unrealised revenue by 2030 across 20 major economies - Korn Ferry. It is worth citing accurately and worth caveating honestly: that study was published in 2018, before the pandemic reshaped participation and before generative AI changed the demand side of skilled knowledge work. It is a scenario built on pre-2018 assumptions, not a forecast anyone has re-run. Use it for the direction, not the decimal.

Technology Hiring in 2024

Tech hiring in 2024 looked weak in the headlines and tight in the data, and both were true at once. Layoff announcements dominated coverage, but they were concentrated in a handful of large employers and in non-technical functions, while the underlying unemployment rate for people who actually do technical work stayed roughly half the national rate all year.

CompTIA tracks this monthly against the official household survey, which makes it one of the few tech-labour figures with a real statistical basis rather than a vendor sample. Its numbers describe a market where finding a job was still easier than filling one.

• The unemployment rate for tech occupations was 2.5% in November 2024, against a national rate of 4.2% - CompTIA.
• Employers had more than 475,000 active tech job postings open in November 2024, adding roughly 184,000 new postings that month alone (CompTIA).
• Employers ran nearly 331,000 active postings for AI roles and AI skills across 2024, a 71% year-over-year increase and the fastest-growing skill category in the market (CompTIA).
• AI skills carried a wage premium of up to 25% in 2024, measured across job advertisements rather than self-reported salaries - PwC's Global AI Jobs Barometer.

The practical reading for recruiters is that 2024 was not a buyer's market for technical talent, whatever the layoff news implied. A 2.5% unemployment rate means essentially everyone competent is already employed, so every hire is a poach, and poaching requires outbound sourcing rather than posting and waiting. The AI premium compounds this: a candidate with genuine AI skills in 2024 could expect a meaningfully better offer somewhere else, which shortened the window between first contact and lost candidate.

On cost, be careful what you benchmark against. The most credible published figure comes from SHRM's benchmarking programme, which put average cost per hire at $5,475 for non-executive roles and $35,879 for executive roles in its 2025 report, based on 2,371 responding members - SHRM. Agency-led technical searches run far above that because the fee is a percentage of salary, but the six-figure per-hire numbers that circulated in 2024 conflate cost per hire with total first-year cost of employment, and no primary source supports them.

Hiring Trends

Beyond the raw supply numbers, three shifts defined how companies actually hired in 2024: flexibility settled into a stable equilibrium rather than continuing to move, skills-based hiring gained real ground in technical roles, and pay transparency became a compliance question rather than a philosophical one.

Remote work is where the denominator problem bites hardest, so it is worth stating precisely. Gallup measures remote-capable employees, meaning people whose jobs could be done from home, which is the population where the decision is actually live. Among that group the split has been remarkably stable: roughly 26% work exclusively remotely, 52% work hybrid, and 22% are fully on-site - Gallup. Measured across the whole workforce, including jobs that cannot be done remotely, every one of those percentages drops sharply. Statistics that quote a small remote share and a large one are usually both right and simply counting different people.

• Roughly six in ten remote-capable employees prefer hybrid, about a third prefer fully remote, and fewer than one in ten prefer full-time on-site (Gallup).
• Six in ten fully remote employees say they are extremely likely to job-hunt if remote flexibility is withdrawn, which makes an RTO mandate a retention event as much as a policy change (Gallup).
• Employers save an estimated $11,000 per half-time remote worker per year, mostly through reduced real estate, utilities and turnover - Global Workplace Analytics. Note the denominator again: that is per half-time telecommuter, not per employee.

The productivity question, which is the one executives actually argue about, does not have a clean answer and it is dishonest to pretend otherwise. The published research disagrees with itself depending on whether it studies fully remote or hybrid arrangements, whether the work is individual or collaborative, and whether the comparison group was selected or randomised. Studies of hybrid arrangements generally find neutral to positive effects, while studies of fully remote individual output are more mixed. Any single percentage quoted with confidence in either direction is quoting one study and ignoring the rest.

Skills-based hiring made its real progress in technical roles, where the signal is verifiable. Employers increasingly evaluated coding ability directly rather than filtering on degree, which widened the pool to self-taught engineers and bootcamp graduates. The honest caveat is that adoption was far patchier than the "death of the degree" commentary suggested: dropping a degree requirement from a job advert is easy and cheap, and many employers did exactly that while continuing to screen on it informally. Pay transparency moved faster, largely because it was legislated rather than chosen, with disclosure requirements in a growing set of US states and the EU pay transparency directive already on the clock.

AI and Technology Adoption in Recruitment

AI adoption in recruiting in 2024 was real, fast-growing, and much smaller than advertised. The gap between those three facts is the most important thing in this report.

The best primary measurement comes from SHRM's Talent Trends survey, which asks HR professionals directly and publishes its method. It found that 26% of organisations used AI for HR tasks in 2024, rising to 43% in 2025 (n=2,040, fielded 3 to 12 February 2025) - SHRM. That is a 17-point jump in a single year, which is a genuinely fast adoption curve. It is also nowhere near 87%.

Where AI was used, recruiting was the first place it landed, and the specific tasks tell you what the technology was actually good at in this period.

• 51% of AI-adopting organisations used it to support recruiting, making talent acquisition the leading HR use case (SHRM).
• 66% used it to write job descriptions and 31% to customise job postings, which are text-generation tasks with a human approving the output (SHRM).
• 44% used it to screen resumes and 32% to automate candidate searches, the two applications with real decision consequences (SHRM).
• Reported benefits skewed heavily to speed: 89% cited time savings, 36% cited lower hiring costs, and only 24% said it improved their ability to identify top candidates (SHRM).

That last cluster is the finding to remember, because it contradicts how the category is sold. Nearly nine in ten adopters got time back. Fewer than one in four got better candidates. In 2024, AI in recruiting was overwhelmingly an efficiency technology rather than a quality technology, and the honest way to justify a purchase was on recruiter hours saved rather than on hire quality, which most organisations were not measuring anyway (SHRM's benchmarking found only 20% of organisations track quality of hire at all).

The tooling itself divided along the same line. Automation of scheduling, resume parsing and pipeline movement delivered reliably because those tasks are mechanical and their failure modes are visible. Candidate relationship management gained ground as teams shifted from reactive applicant tracking toward maintaining talent pools over time. Sourcing tools, including autonomous sourcing platforms such as HeroHunt.ai, pushed further by finding and contacting candidates rather than only organising the ones who applied. Meanwhile AI-scored video interviewing, which assessed candidates on non-verbal cues, drew the most vendor enthusiasm and the least evidence, and it is the corner of the market that has since attracted regulatory attention rather than validation.

Resistance was real on both sides of the table. Recruiters worried that algorithmic screening discards non-standard but capable candidates, which is a well-founded concern given that a screening model reproduces whatever its training data rewarded. Candidates were warier still, and the industry's response has been explainability: making it possible to see why a system surfaced or rejected someone. That work matters more than it sounds, because it is the difference between a tool a recruiter can defend in a hiring meeting and one they cannot.

How the 2024 Numbers Held Up

Two years of subsequent data have now tested the 2024 picture, and the scorecard is mixed in an instructive way: the labour-supply story held almost exactly, and the AI story moved faster than any of the credible 2024 numbers implied but never approached the incredible ones.

The talent shortage eased slightly and structurally persisted, which is what the demographic argument above predicted. ManpowerGroup's global figure went from 75% in 2024 to 74% in 2025 to 72% in 2026, with the US at 69%, based on 39,063 employers across 41 countries - ManpowerGroup. A three-point move over two years, from a record high, is not a resolution. It is a plateau.

What changed is which skills are scarce. In 2026, for the first time in the survey's history, AI model and application development (20%) and AI literacy (19%) became the hardest skills to find globally, displacing engineering and IT (ManpowerGroup). The 2024 signal, 331,000 AI postings and a 71% growth rate, pointed directly at this. The wage data followed the same line: PwC's measured AI skills premium went from up to 25% in 2024 to 56% in 2025 - PwC.

AI adoption in HR kept climbing but on the primary curve, not the mythical one. SHRM's tracking went from 26% of organisations in 2024, to 43% in 2025, to 39% currently using AI in HR in its 2026 report, with 27% using it for recruiting specifically (n=1,908, fielded 5 to 23 December 2025) - SHRM. The apparent dip between 2025 and 2026 is a question definition difference rather than a retreat, and the useful takeaway is the shape: fewer than half of organisations, two years after the technology became unavoidable in the discourse. Adoption was also deeply uneven by size, with 60% of the largest organisations using AI in HR against roughly a third of small and midsize ones. Anyone who told you in 2024 that 87% of companies were already doing this was, two years later, still wrong by a factor of more than two.

Remote work is the cleanest case of a trend that simply stopped moving. Gallup's remote-capable split has flattened at roughly 26% fully remote, 52% hybrid and 22% on-site, and it has stayed there through several years of loudly announced return-to-office mandates (Gallup). Hybrid was not a transition state. It was the destination.

What to Take From 2024

2024 is worth understanding as the year two clocks became visible at once. A demographic clock made skilled hiring structurally hard in a way that no hiring slowdown reverses, and a technology clock started changing which skills were scarce faster than the education system could respond. The 2026 data confirms both: the shortage flattened at a high level rather than clearing, and the scarcest skill in the world is now the one that barely registered as a category three years earlier.

For recruiters, the operational conclusions from the 2024 data have aged well. A 2.5% unemployment rate in your target population means outbound sourcing is not a growth tactic but the baseline, because the people you want are not applying anywhere. AI is worth adopting for the hours it returns, which is what adopters overwhelmingly reported getting, and worth being sceptical about for hire quality, which most of them did not get and most organisations still cannot measure. And flexibility is a retention instrument with a measurable cost of withdrawal, which is a more useful way to think about it than as a perk.

The broader lesson is about the numbers themselves. The 2024 statistics that turned out to be right, the ones on shortage, participation, tech unemployment and AI demand, all came from organisations that published their sample size and their method. The ones that turned out to be wrong came from pages that published neither. That test cost nothing to apply in 2024 and it would have sorted the entire list correctly.

Written by Yuma Heymans (@yumahey), founder of HeroHunt.ai. He has been building AI recruiting technology since 2021, which is a useful vantage point from which to notice how far apart the industry's adoption statistics and its actual adoption have been.

Statistics re-verified against primary sources in July 2026. Figures without a traceable primary source have been removed rather than repeated. Survey figures are restated with their sample sizes and field dates so you can check them yourself.