Two numbers define talent acquisition in 2025. AI adoption inside HR jumped from 26% to 43% of organizations in a single year - SHRM. Over the same period the average cost to fill a non-executive role reached $5,475, and an executive hire reached $35,879 - SHRM 2025 Benchmarking.

Recruiting became dramatically cheaper to automate and more expensive to actually execute, at the same time. That apparent contradiction is the story of the year, and most of this report is an attempt to explain it.

1. Key Trends Shaping Talent Acquisition
2. The Evolving Recruitment Landscape
3. Recruitment Budgets and Resource Allocation
4. Hiring Metrics and Efficiency
5. Most Recruited Roles in 2025
6. Challenges Facing Hiring Companies
7. The Role of AI in Modern Recruitment
8. Career Paths and Salaries in Recruitment
9. Conclusion

A note on method, because talent acquisition is an unusually polluted data category. Many widely repeated recruiting statistics trace back to vendor blogs citing other vendor blogs, and the original number is often either misquoted or was never measured. Every figure below is attributed to the organization that actually collected it: SHRM benchmarking and survey data, the American Staffing Association for agency employment, the Bureau of Labor Statistics for wages, Gartner and LinkedIn for practitioner surveys, and the Burning Glass Institute with Harvard Business School for skills-based hiring. Where a popular claim does not survive checking, this report says so rather than repeating it.

1. Key Trends Shaping Talent Acquisition

Five shifts define the year, and they are not equally real. AI adoption is real and fast. Skills-based hiring is loudly announced and barely practised. Hybrid work has settled into a stable equilibrium rather than continuing to move. Internal mobility is genuinely rising. Candidate experience is deteriorating under application volume that AI itself created.

Separating the trends that changed behaviour from the trends that only changed vocabulary is the most useful thing a report like this can do, because budget follows belief. A talent function that reorganises around a trend which is not actually happening spends real money on an imaginary problem.

1. AI-Powered Hiring

AI moved from pilot to standard equipment in 2025. 43% of organizations now use AI somewhere in HR, up from 26% the year before, a 17 point jump in twelve months - SHRM, which surveyed 2,040 HR professionals in February 2025. Recruiting is the single largest application: just over half of AI-using organizations point it at hiring first.

What AI actually does inside recruiting is narrower than the marketing suggests, and the ranking of use cases is revealing. It writes far more than it decides.

• Writing job descriptions: 66%
• Screening resumes: 44%
• Automating candidate searches: 32%
• Customizing job postings: 31%
• Communicating with applicants: 29%

The distribution matters more than the headline. The most common use, drafting job descriptions, is a writing task with no candidate-facing consequence if it goes wrong. The uses that touch an actual hiring decision, screening and search, sit far lower. In other words, most organizations have adopted AI where the downside of a mistake is embarrassment rather than a lawsuit. That is a rational sequencing, and it means claims that AI is now "making hiring decisions" at scale are ahead of the evidence.

The benefits practitioners report follow the same pattern: 89% say AI saves time or increases efficiency, but only 36% say it reduces cost and just 24% say it improved their ability to identify top candidates. AI is currently a speed technology, not yet a quality technology. Tools like HeroHunt.ai that automate sourcing and outreach compete precisely on that first axis, and buyers should judge them on it rather than on quality-of-hire promises the category cannot yet substantiate.

One widely circulated claim deserves correction, because this report previously repeated it. The figure that 99% of Fortune 500 companies use AI in recruitment is not an AI statistic at all. It is a garbled retelling of Jobscan's long-running research on applicant tracking system adoption, which found an ATS at 97.8% of Fortune 500 companies in 2025, or 489 of 500 - Jobscan. An ATS is a database, not artificial intelligence, and most Fortune 500 ATS deployments predate the current AI wave by a decade. Anyone citing 99% AI adoption is citing a typo that escaped into the wild.

2. Skills-Based Hiring

Skills-based hiring is the largest gap between rhetoric and reality in the industry, and the best available research is blunt about it. Companies did remove degree requirements from postings: the share of US postings with no degree requirement rose from 49% in 2017 to roughly 56% in 2023. But when researchers followed those same companies to see who they actually hired, the share of new hires without a four-year degree moved by less than 4 percentage points - Burning Glass Institute and Harvard Business School.

Translated into people, the effect is close to a rounding error. The researchers estimate the shift created new opportunities for about 97,000 workers against roughly 77 million hires a year, or fewer than 1 in 700 hires. Removing a line from a job posting turns out to be nearly free, while changing how a hiring manager evaluates a stranger is expensive and slow, so the first happened at scale and the second mostly did not.

The exception is instructive rather than encouraging. About 37% of analysed firms did convert the policy into practice, increasing hiring of workers without bachelor's degrees by nearly 20%. That group includes Walmart, Apple, General Motors, Target and Cigna. What separates them is not intent but mechanism: they changed interview structure, scorecards and assessment, not just posting text. For a talent leader, the practical read is that skills-based hiring is not a sourcing change or a copywriting change. If your assessment stage still runs on pedigree, deleting the degree line from the posting will do nothing except widen a funnel you then reject at the same rate.

3. Hybrid Work Models

Remote work stopped moving. That is the headline, and it is more useful than any prediction. The Survey of Working Arrangements and Attitudes, run monthly by researchers at Stanford, ITAM and the Hoover Institution, finds that about 25% of paid full days in the US were worked from home in May 2026, against a 2025 annual average near 26% - WFH Research. The line has been essentially flat for three years.

The magnitude only makes sense against the baseline. Pre-pandemic, work from home accounted for 7.2% of paid full days. The peak in May 2020 was 61.5%. The settling point of roughly a quarter of all days is therefore neither the revolution nor the reversal that both camps predicted. It is a permanent step change of about three and a half times the old level, followed by stability. Return-to-office mandates made headlines without meaningfully moving the aggregate number.

This report previously claimed that 97% of workers prefer some form of remote work. That figure is not credible and has been removed. It originates in surveys of people who were already searching for remote jobs on a remote-work job board, which is a textbook self-selected sample: asking remote job seekers whether they want remote work measures nothing. The defensible statement is narrower and more useful: hybrid is now the default expectation for roles that can be done remotely, but it is a stable equilibrium, not an expanding one, so employers should stop treating flexibility as a trend to get ahead of and start treating it as a fixed feature of the market they compete in.

4. Internal Mobility and Employee Retention

Organizations are placing greater emphasis on internal mobility and career development to retain employees. This focus on nurturing existing talent is becoming a key strategy in addressing skills gaps and reducing recruitment costs.

In-house recruitment teams are working closely with learning and development departments to ensure that employees have opportunities to grow and advance within the company (Lighthouse Research & Advisory).

5. Agile Recruitment Models

Recruitment processes are becoming more flexible and iterative, adopting methodologies similar to agile software development. This allows for:

• Quicker adjustments based on real-time feedback
• Closer collaboration between hiring teams and candidates
• More responsive hiring practices in fast-changing markets (Joveo)

6. Enhanced Candidate Experience and Onboarding

There's an increased focus on improving the overall candidate experience, particularly during the onboarding phase. This emphasis aims to reduce early turnover and combat the rising phenomenon of "first-day ghosting" (Lighthouse Research & Advisory, SHRM).

Key areas of improvement include:

• Streamlined application processes
• Regular communication throughout the hiring journey
• Comprehensive onboarding programs that extend beyond the first week

2. The Evolving Recruitment Landscape

The internal-versus-agency split is best understood as a cycle, not a trend. Internal teams grow when hiring volume is high and predictable, because a salaried recruiter beats a 20% placement fee once you are filling enough roles. Agencies grow when volume is volatile, urgent or specialised, because fixed headcount is the wrong instrument for lumpy demand. 2025 sat in an awkward middle: volumes were soft enough to keep internal teams lean, but the roles that did open were hard enough to push work back to specialists.

Internal Recruitment Teams

More companies are investing in building robust internal recruitment capabilities. This trend is driven by several factors:

• Cost reduction
• Improved candidate experience
• Better alignment of hiring strategies with organizational goals

The structural advantage of an internal team is not cost per hire, which often looks worse than an agency fee on a spreadsheet once you load salary, tooling and overhead onto a small number of hires. It is context. An internal recruiter who has sat in the hiring manager's planning meetings screens against the real job, not the job description. That advantage compounds over repeat hiring into the same team and disappears entirely for a one-off role in an unfamiliar function, which is exactly the boundary where agencies win.

Recruitment Agencies

Agency employment is measurable, and the measured picture is a market that shrank and then stabilised. US staffing companies employed just under 2 million temporary and contract workers per week through 2025, down from nearly 2.2 million per week in 2024 - American Staffing Association. Temporary employment peaked in 2022 and declined through 2023 and 2024 even while the wider labour market added jobs, which is the classic pattern of employers hedging: temp headcount is the first thing cut when the outlook gets cloudy.

The turn came late. Temporary and contract staffing sales rose 2.6% in the fourth quarter of 2025 to $29.9 billion, and staffing employment in mid-2026 ran 4.6% above the same period a year earlier - ASA. Because temp demand is a leading indicator, that recovery is one of the more encouraging signals in this report for permanent hiring in late 2026.

A correction is required here too. This report previously stated that over 90% of U.S. businesses use staffing agencies. That claim does not appear in ASA's published statistics, and the trail leads to invoice-factoring and lead-generation blogs rather than to any survey. It has been removed. What ASA does document is scale rather than penetration: America's staffing companies hired 12.7 million temporary and contract employees over the course of a year. The honest summary is that agency use is very common and concentrated in healthcare, IT and industrial work, and that nobody has credibly measured the share of all US businesses that use one.

3. Recruitment Budgets and Resource Allocation

Recruiting consumes about a quarter of the entire HR budget. SHRM's 2025 benchmarking puts the average recruiting share at 26% of total HR spend, but the average conceals an enormous spread: the median organization spends 20%, the 10th percentile spends 10%, and the 75th percentile spends 39% - SHRM. A four-fold gap between the bottom and the top of the distribution is not noise. It reflects whether an organization is growing.

This is why benchmarking your recruiting budget against the average is close to useless. The average blends a company hiring 500 people this year with one backfilling attrition, and those are different businesses with different problems. The useful comparison is against organizations at your growth rate and in your industry, and if you cannot get that, comparing your own ratio year over year tells you more than any external benchmark will.

Where the money goes has shifted in a specific way. SHRM attributes the rise in cost per hire to rising recruiter compensation, greater sourcing complexity, and changing job board pricing models. Note what is absent from that list: software. The tooling line has been the deflationary part of the budget, because competition among vendors has pushed the entry price of a capable applicant tracking system into the low tens of dollars per user per month, while the human line kept climbing. Organizations that expected technology spend to reduce total recruiting cost have generally been disappointed, because the expensive input was never the software.

The practical implication for 2026 planning is to stop treating the tooling budget as the lever. If salary and sourcing difficulty drive the cost, then the levers that matter are recruiter productivity, internal mobility (which avoids the external hire entirely), and reducing the number of roles that need agency support. Cutting a software line item to save a few thousand dollars while a single unfilled senior req costs multiples of that in lost output is a false economy, and it is one of the most common budget mistakes in the function.

4. Hiring Metrics and Efficiency

The central metric problem in talent acquisition is that the industry measures speed precisely and quality barely at all. Only 20% of organizations track quality of hire - SHRM. Meanwhile 89% of TA professionals say measuring quality of hire is increasingly important, but just 25% feel confident they can do it - LinkedIn. An industry that agrees on what matters, cannot measure it, and therefore optimises the thing it can measure instead, will reliably get faster at making mediocre hires.

Time to fill remains roughly six weeks from posting to accepted offer, and the interesting detail is where it goes. SHRM finds the process is highly segmented, with screening and interviewing averaging 8 to 9 days each. That decomposition is the whole game for anyone trying to compress a timeline. The delay is not usually one broken stage. It is a sequence of short waits, most of them for a human to do a small thing, and the largest single contributor in most pipelines is hiring manager response time rather than anything the recruiter controls.

Cost per hire benchmarks are similarly misread. The headline $5,475 for a non-executive role sits alongside $35,879 for an executive hire, nearly seven times higher, and executive cost jumped 21% from 2022. Retail roles sit near $2,700. A single company-wide cost-per-hire number therefore mostly measures your hiring mix, not your efficiency: shift your hiring toward senior roles and the metric worsens while nothing about your process changed.

The pragmatic response to the quality-of-hire gap is to accept a proxy rather than wait for a perfect measure. The most common proxies in practice are job performance ratings (66%), new hire retention (60%) and hiring manager satisfaction (44%). None is clean. Performance ratings drift with manager bias, retention conflates hiring quality with management quality, and satisfaction surveys measure feelings. But a noisy quality signal reviewed every quarter beats an unmeasured one, because it at least creates a feedback loop between a sourcing decision and its outcome. Without that loop, a recruiting function cannot learn, and it will keep repeating whichever channel produced the most candidates rather than the best ones.

5. Most Recruited Roles in 2025

AI roles dominated the fastest-growing list, and the shape of that dominance is more interesting than the fact of it. LinkedIn's Jobs on the Rise ranked AI engineer first and AI consultant second, with AI occupying three of the top five positions - LinkedIn. The ranking measures growth rate among millions of jobs started by members, not absolute volume, which is a distinction most coverage loses.

That distinction matters enormously for planning. A role can top a fastest-growing list while representing a tiny fraction of hiring, because growth rates are calculated from a small base. AI engineer being number one does not mean most companies are hiring AI engineers. It means the companies that are hiring them are doing so from almost nothing, very fast. A talent leader who reads the list as a demand forecast will badly misallocate.

The rest of the top ten undercuts the tech-only narrative, and this is the genuinely useful finding.

• Physical therapist (3rd) and security guard (10th): in-person, non-substitutable work
• Workforce development manager (4th): the reskilling function itself is growing
• Travel advisor (5th) and event coordinator (6th): the experience economy rebound
• Sustainability specialist (9th): regulatory-driven demand

Two economies are visible in one list. Half of it is AI and the other half is work that a human has to be physically present to do, with very little in between. That hollowing of the middle is the labour market story of the decade, and it has a direct recruiting consequence: the roles growing fastest are the ones where either the skill did not exist three years ago or the constraint is a body in a location. Both are hard to source, for opposite reasons, which helps explain why sourcing complexity is pushing cost per hire up.

One absence is worth noting. Diversity and inclusion roles, which appeared prominently on these lists in prior years, dropped out of the top 25 entirely. Whatever one's view of that shift, it is a real change in employer behaviour rather than a change in sentiment, because the list counts jobs people actually started.

6. Challenges Facing Hiring Companies

The defining new problem of 2025 is that employers can no longer be confident the candidate is real. Gartner predicts that by 2028, 1 in 4 candidate profiles globally will be fake, and the prediction is grounded in survey data rather than speculation: 6% of 3,000 surveyed candidates admitted to interview fraud, meaning they either impersonated someone else or had someone else pose as them - Gartner via HR Dive. Six percent admitting to fraud in a survey is a floor, not a ceiling, because people under-report misconduct.

This is a direct consequence of remote hiring plus cheap generative tools. A synthetic identity with a coherent resume, a plausible LinkedIn history and a real-time deepfaked video interview is now within reach of a modestly skilled fraudster, and the payoff is a salary. Security teams have started treating hiring as an attack surface, because a fraudulent hire in an engineering role is an authenticated intruder with a laptop and credentials, not merely a bad hire.

The second challenge is volume, and it is self-inflicted by the industry's own technology. Four in ten candidates now use AI during the application process for resumes, cover letters and assessment responses. When applying costs a candidate thirty seconds, application volume per opening rises sharply while the information content of each application falls, since a generated cover letter says nothing about the person. Employers responded with more automated screening, which raised the return on generating more applications. Both sides are now automating against each other, and the equilibrium is worse for everyone: recruiters drown, and genuinely qualified candidates get filtered by keyword matching they did not game well enough.

Candidate trust is the casualty. Only 26% of job applicants trust AI to evaluate them fairly, half believe AI screens their applications, and a third actively worry about being rejected by it - Gartner. That number should alarm anyone running an employer brand. A process the majority of participants distrust does not just feel bad, it selectively repels the candidates who have other options, which are precisely the candidates the process exists to attract. The organizations handling this well are doing unglamorous things: telling candidates plainly where automation is used, keeping a human review stage for rejections, and verifying identity early rather than discovering a problem at onboarding.

7. The Role of AI in Modern Recruitment

The honest summary of AI in recruiting is that it has convincingly automated the writing and the searching, and has not yet automated the judging. Every data point in this report converges on that line. 89% of practitioners report time savings, only 24% report better candidate identification, and the most common application by a wide margin is drafting job descriptions rather than evaluating people.

The time savings are real and large. LinkedIn finds that generative AI users save roughly one full workday per week, about a 20% workload reduction, and 74% of TA professionals say AI makes hiring more efficient. For a function where recruiter compensation is the largest and fastest-growing cost line, returning a day a week per recruiter is a material economic result. This is where the current generation of tools, including autonomous sourcing platforms such as HeroHunt.ai, earns its keep: compressing the search and outreach work that used to consume a recruiter's morning.

What AI has not done is change who gets hired, and the reason is structural rather than technical. Hiring decisions are consequential, legally exposed and hard to validate, and the feedback signal (did this hire work out?) arrives months later and is measured by only 20% of organizations. Machine learning needs a label to learn from. An industry that does not record outcomes cannot train a system to predict them, which is why the sophisticated screening promised for a decade keeps underdelivering. The bottleneck is not model capability. It is that most companies genuinely do not know which of their past hires were good.

There are signs of a plateau in adoption, and the counting is contested. SHRM's State of AI in HR 2026, fielded in December 2025 among 1,908 HR professionals, reports 39% of organizations with AI currently adopted in HR and 27% using it in recruiting, with 62% using AI somewhere in the business. Those figures are not directly comparable to the 43% cited earlier because the question framing differs, and the discrepancy is itself worth noting: anyone quoting a single authoritative AI adoption percentage is overstating the precision available. The direction is clear, the level is not.

The more durable finding in that report is a governance failure: 56% of organizations do not formally measure whether their AI investment succeeded. Combined with the fact that only a fifth measure quality of hire, this describes a function buying a technology to improve an outcome it does not track, and then not checking whether it worked. The organizations that will get real advantage from AI in 2026 are not the ones with the best tools. They are the ones that instrumented the outcome first.

8. Career Paths and Salaries in Recruitment

Recruiting remains a stable profession with unspectacular pay and a shifting skill profile. The median annual wage for human resources specialists, the BLS category that contains most corporate recruiters, was $72,910 in May 2024, with the bottom 10% under $45,440 and the top 10% above $126,540 - Bureau of Labor Statistics. Employment is projected to grow 6% from 2024 to 2034, faster than the average occupation, with roughly 81,800 openings a year.

That projection is the most direct available answer to the question recruiters keep asking, which is whether AI is coming for their jobs. The BLS, which has every incentive to be conservative and none to be promotional, models the occupation growing faster than average through 2034. The wide wage band is the more interesting feature: a spread from $45k to $126k within one occupation means the title tells you very little and the specialisation tells you nearly everything. Agency recruiters on commission and in-house technical sourcers occupy different economic worlds while sharing a BLS code.

The skill mix is repricing in a way that follows directly from what AI automates. LinkedIn found employers were 54 times more likely to list relationship development as a required skill on paid recruiter job posts over the past year. When AI absorbs the searching and the writing, the residual human value concentrates in the parts that require trust: persuading a passive senior candidate to take a call, managing a hiring manager who wants a unicorn, closing an offer against a counter. Anyone building a recruiting career on speed of sourcing is competing with software that does not sleep. Anyone building it on judgement and relationships is competing with software that cannot do either.

The realistic career advice, therefore, is to move toward the decisions rather than the tasks. The tasks are being automated in the order of their reversibility, and the recruiter who owns the hiring manager relationship and the quality-of-hire conversation is holding the one part of the process this report has shown to be both valuable and unautomated.

9. Conclusion

Talent acquisition in 2025 was defined by a productivity gain that has not yet become a quality gain. AI adoption roughly doubled year over year and returned about a day a week to the average recruiter who uses it, while cost per hire rose to $5,475, quality of hire went unmeasured by 80% of organizations, and candidate trust in automated evaluation sat at 26%. Faster, more expensive, and no better at picking people. That is the year in one sentence.

The through-line connecting every section of this report is the measurement gap. Skills-based hiring failed to translate into hires because companies changed postings instead of assessments. AI stalled at writing tasks rather than judging tasks because nobody records which hires worked. Budgets get benchmarked against a meaningless average. In each case the failure is not a lack of technology or intent, it is that the outcome was never instrumented, so the feedback loop that would correct the behaviour does not close.

For a talent leader planning the next year, that suggests an uncomfortable but cheap priority order. Instrument quality of hire with an imperfect proxy before buying anything else, because without it no tool purchase can be evaluated and no sourcing channel can be compared. Fix assessment before rewriting postings, because that is where skills-based hiring actually lives. Verify identity early, because a quarter of profiles may be fake by 2028. Treat flexibility as a fixed market condition rather than a differentiator, because the remote share has not moved in three years. And spend the AI budget on giving recruiters their day back, which is the one benefit the evidence robustly supports, rather than on screening promises the category cannot yet keep.

The encouraging signal for late 2026 comes from the staffing data. Temporary and contract employment, historically a leading indicator, turned positive in Q4 2025 and ran 4.6% above prior-year levels by mid-2026. Employers hedge with temps before they commit to permanent headcount. If that pattern holds, the hiring environment described in this report is the trough rather than the norm.

Written by Yuma Heymans (@yumahey), founder of HeroHunt.ai. With a background in management consulting and years spent building AI recruitment technology, he writes about how AI is reshaping talent acquisition.

This report was first published in April 2025. All figures were re-verified against primary sources in July 2026, and two previously published statistics (99% Fortune 500 AI adoption, 97% of workers preferring remote work) were removed as unsupported. Recruiting data changes quickly: check the linked primary sources before relying on any figure for planning.