Recruiting & HR: where the industry stands with AI

Where recruiting and HR stand with AI

Recruiting and HR in 2026 use AI for almost every routine hiring task, yet hiring itself has not become faster. Software screens CVs, writes job ads and books interviews, and much of it now runs as an agent, a program that carries out a chain of tasks on its own between human checks. Candidates use the same tools to write their applications. So far the result is more volume on both sides and few measured gains. What the law increasingly demands is a person who makes the final call.

Use is close to universal among the US employers surveyed. A survey of more than 1,000 US HR and recruiting leaders by Paylocity, a payroll software company, found 91% use AI somewhere in recruiting and 67% use it to screen CVs (HR Dive, August 2026).

Smaller firms and Latin America are further behind. In Argentina, a survey run by the research firm Trendsity for Deel, a payroll company, found 39% of small and mid-size companies use AI in at least one HR process, falling to 20% among the smallest (BAE Negocios, September 2026).

What is still promise is speed. In a survey of about 40,000 employers worldwide published in September 2026, ManpowerGroup, the global staffing group, found 28% fill roles faster than in 2025 and 30% slower, with the rest unchanged.

An earlier ManpowerGroup study with the research firm Everest Group, published in June 2026, asked 80 senior leaders in the US and UK: almost all use AI in recruiting, and fewer than 5% report results that changed how they hire.

What changed recently

California put dates on the human check. On September 30, 2026, Governor Gavin Newsom signed three workplace AI laws (governor's office, Associated Press). From July 1, 2027, an employer in the state may not discipline or fire anyone on an automated system's word alone: the reviewer must gather their own evidence, and the worker gets a written notice naming the data used and a person who can explain it. From January 1, 2027, tools that infer a worker's emotions from face, voice or tone are banned, and a large layoff caused mainly by automation must say so in the legal notice. Era Haus covered the package in Recruiting & HR: California puts a person behind every AI firing. Hiring decisions sit outside the firing law.

Europe delayed its strict rules for hiring tools and kept the disclosure rule. The EU's AI Act treats hiring software as "high-risk", which brings bias testing, record-keeping and human oversight. A law published on July 24, 2026 pushed those duties to December 2, 2027 (Mayer Brown, July 2026). The duty to tell a candidate they are dealing with an AI, such as a chatbot or an AI-run first interview, applied from August 2, 2026. The AI Act's ban on reading emotions at work has applied since February 2025.

Agents arrived inside products recruiters already pay for. LinkedIn announced the second version of its Hiring Assistant in late September 2026; it learns how each recruiter has hired before and reaches existing users in November at no extra cost (HR Brew, Recruiting News Network). In mid-September Adecco, the Swiss staffing group, began rolling out agent software from Salesforce, the US customer-software company, to about 27,000 staff in more than 40 countries. If you recruit on LinkedIn, you get an agent in November without buying anything; if you run an agency, expect clients to ask which steps an agent handles.

Candidates are automating too. On September 21, Gupy, the recruiting system many large Brazilian employers use, released a connector that lets job seekers search and apply through general AI assistants such as Claude and Gemini (IT Forum). In the UK, 67% of hiring managers told the staffing firm Robert Half that AI-written applications have made hiring slower (September 2026).

What works in practice

Keep a person who actually decides. Europe and Brazil already require it: Europe's data-protection law, GDPR, restricts rejections made purely by software, and Brazil's equivalent, the LGPD, gives a right to human review of automated decisions. US states are moving the same way, Colorado from January 2027 and California, for firing, from July 2027. Clicking "approve" on a score does not count. Write down who decided, what the tool recommended and why you agreed or overrode it.

Automate tasks, keep the roles. Careerminds, a US outplacement firm, surveyed 600 US HR leaders who had run layoffs in early 2026: about two in three of those who cut jobs citing AI were rehiring, and only 8.4% said they would make the same cut again. The firms reporting gains count them as recruiter capacity.

Test the work instead of the CV. When a model writes the CV and another model reads it, the CV stops telling candidates apart. Era Haus's recommendation is a 20-to-30-minute work sample, AI tools allowed and the candidate asked to explain their choices. It answers a gap employers report: 78% of HR leaders in a US and UK survey by Talogy, a company that sells hiring tests, said they struggle to assess AI skills (HR Dive, August 2026).

Tell candidates, and question the vendor. One plain sentence in the chat window or interview invitation covers Europe's disclosure duty. Before buying a screening tool, ask in writing how it tests for bias and whether it reuses a candidate's score across other employers. Switch off any feature that scores emotion or "engagement".

What named companies are doing

Adecco is the clearest European case with numbers. At its second-quarter results on August 6, 2026, it said agents are now in use across half of its Adecco-brand business, and reported recruiter productivity up 25% to 35% and fill rates up 10%. Those are the company's own figures; it frames them as more work per recruiter and has announced no cuts tied to them.

IBM built an assistant that answers about 94% of its staff's HR questions, and said it would triple US entry-level hiring because cutting junior roles now would leave it short of experienced staff later (Fortune, February 2026).

Workday, whose HR system many large employers use to screen applicants, is fighting a US federal lawsuit, Mobley v. Workday, which claims its screening discriminated by age, race and disability. The age claims have run as a nationwide group action since May 2025; in June 2026 a judge refused to dismiss most of the remaining claims, and the plaintiffs have since been collecting the names of employers whose applicants were screened (Duane Morris, HR Dive). Customers of a screening vendor end up in the case record.

In Argentina, the job board Bumeran ran an online fair in late August 2026 with more than 9,000 vacancies, where its AI assistant built each candidate's profile from the uploaded CV (Fortuna). Bumeran's own survey found 46% of Argentine HR specialists use AI in selection, mostly to filter profiles by keyword (Infobae, August 2026).

The tools, by size of employer

Corporate. Large employers get agents inside the HR suite they already run. Workday customers use a chatbot from Paradox, a company Workday bought in October 2025, to schedule interviews and answer applicants' questions. Recruiters on LinkedIn's paid Recruiter product use its Hiring Assistant to find and message candidates. In Brazil, large employers screen applicants in Gupy. US and UK staffing agencies mostly run on Bullhorn, whose AI "digital workers", launched in May 2026, find prospects, check candidate details and transcribe calls.

Small business. The stack is regional. In Europe, Personio, a Munich HR system with about 16,000 customers, now ranks incoming applications before a recruiter reads them, using technology from aurio, a startup it bought in April 2026. In the US, mid-size hiring teams use Greenhouse, an applicant tracking system (where job ads and applications are kept), which bought a voice-interview startup in May 2026 so employers can run AI phone screens. In Argentina, small firms post on Bumeran and its sister sites. Many employers also draft job ads with a general chat assistant; it was the third most common use among the US HR leaders Paylocity surveyed.

Freelance and solo. We could not verify which tools independent recruiters and one-person HR consultants use; this section stays short until we can.

What to watch over the next year

First, January 1, 2027 in the US. Several rules switch on together: California's emotion-monitoring ban and layoff-notice change, California's rules on automated decisions for businesses covered by its privacy law (notice before use and a right to opt out), and Colorado's rewritten AI law, which asks for notice, an explanation after a rejection and human review. They share one design, so one process can meet all of them. No US federal law has paused any of them.

Second, the courts. As Mobley v. Workday proceeds, employers that used the tool will face questions about their own hiring records. A separate US case filed in January 2026 against Eightfold, an AI recruiting company, argues that AI match scores are credit-style reports, which would bring consent and dispute duties.

Third, Europe's definition. The European Commission's draft guidance of May 19, 2026 says a tool counts as high-risk when its output "materially influences" a hiring decision, even if a person signs off. If the final text keeps that line, most screening tools a small European employer uses will carry the full duties from December 2027. In Brazil, an AI bill that treats recruiting software as high-risk is waiting in the lower house after the October 4 elections.