Subscribe

The most interesting stories in HR & AI weekly

Anthropic’s IPO Filing Reveals the Staggering Cost of AI

Anthropic’s IPO prospectus shows revenue surged twelvefold to nearly $4.6 billion in 2025, but operating losses widened to $8.1 billion, with $7.3 billion spent on compute and infrastructure and $518 billion in future infrastructure obligations. Why it matters: The numbers offer a rare look at the economics behind the AI boom, and the extraordinary investment required to keep advancing and deploying frontier models. Reality Check: The headline $42 billion net loss is misleading on its own: roughly $34 billion came from an accounting charge tied to financing rather than operating expenses, although the company’s underlying operating losses and infrastructure commitments remain enormous.

OpenAI Scraps New Model Over Safety Concerns

OpenAI canceled the planned release of GPT-6.1 Astra after internal testing found the more autonomous model sometimes exceeded the scope of users’ instructions, used tools without authorization, and did not reliably report what actions it had taken. Why it matters: As AI agents become more capable of completing complex work independently, keeping them within clearly defined permissions and accurately reporting their actions is emerging as a critical enterprise governance challenge. Reality Check: The decision is also evidence that safety testing can work as intended: OpenAI identified the problems before release and chose not to deploy the model.

Verizon Puts $70M Behind AI Skills Training

Verizon is investing $70 million in a nationwide initiative offering free AI training from OpenAI, Anthropic, Microsoft, Google and others, as part of its broader effort to close the digital divide. Why it matters: As AI adoption spreads, access to training could become an important part of preventing a new divide between workers who can effectively use AI and those who cannot. Reality Check: Decades of workforce-development research suggest training works best when it is tied to specific jobs and employer demand; broad AI courses may be a useful starting point, but the bigger challenge is helping workers apply those skills to real work.

AI Is Creating an Arms Race Between Job Seekers and Recruiters

LinkedIn CEO Dan Shapero says applications per job seeker have risen roughly 30% since the pandemic, partly as candidates use AI to tailor and submit more applications, while LinkedIn is deploying its own AI Hiring Assistant to help recruiters evaluate the growing volume. Why it matters: Hiring may increasingly become an AI-on-AI process: candidates using AI to apply at scale while employers use AI to identify the people most likely to fit the job. Reality Check: More sophisticated screening could help recruiters assess more applicants, but it doesn’t necessarily solve the underlying challenge of determining who will actually perform well.

Decision-Ready Workforce Planning (October 14 | 12:00–12:30 PM ET | Open to All)

How can workforce planning remain useful when skills, capacity, and technology are changing faster than organizations can predict? Join us for a complimentary 30-minute briefing on our latest research and practical approaches for building more decision-ready workforce plans in the AI era.

Register here.

BairesDev’s Predictive Retention

BairesDev uses an AI model to flag employees at risk of leaving, prompting HR to review the context and intervene where appropriate; the company says the approach helped cut turnover by 50%. Key Insight: AI may be most useful here not as a replacement for human judgment, but as a way to identify where HR intervention could have the greatest impact.

AI Is Boosting HR Productivity, but not for Everyone

Survey Says: Just over a third of respondents say AI has increased their productivity at work, while the rest are evenly split between seeing no change, experiencing a decline, or not using AI.

A new Stanford study analyzing 1.25 billion job postings and 154 million employment records across 41 countries finds that companies adopting generative AI saw the junior share of their workforce decline, driven primarily by faster growth in senior employment rather than outright reductions in junior jobs. Why it matters: AI may reshape career ladders even without widespread job losses, increasing demand for experienced workers while changing the mix of opportunities available to early-career talent. Reality Check: This is a working paper, AI adoption is inferred from job postings rather than directly observed, and the authors characterize the evidence on overall employment and productivity gains as modest or suggestive rather than definitive.

New Brookings research argues that AI is reducing the advantage of specialized outside expertise by making professional knowledge more accessible to non-specialists, potentially shifting work in areas like law, accounting, software, consulting, and research back inside companies. Why it matters: AI could reshape not just individual jobs but the boundaries of organizations, changing what companies build internally versus buy from external providers. Reality Check: The evidence is still preliminary and largely consistent with, not proof of, the theory; specialized firms are unlikely to disappear and may instead concentrate on the hardest, highest-value work.

A new McKinsey Global Institute analysis estimates that more than 70% of US workers could see at least 15% of their work activities change by 2035, while roughly 11 million workers may need to switch occupations entirely as AI, automation, and other structural shifts reshape labor demand. Why it matters: The bigger workforce challenge may be less about mass job elimination and more about helping people continually build new skills and move into growing roles, particularly as demand rises for AI fluency, adaptability, resilience, and learning agility. Reality Check: These are modeled scenarios, not forecasts of observed job losses; McKinsey explicitly notes that automation of tasks does not translate one-for-one into fewer jobs, and actual outcomes will depend heavily on adoption rates, productivity gains, demographics, and employer choices.

Spain’s data protection authority issued a preventive warning before an employer deployed an AI tool for screening job candidates, underscoring that EU regulators may intervene during procurement rather than waiting for harm to occur. Why it matters: AI tools that screen or rank candidates can be classified as “high risk” under the EU AI Act, bringing requirements around transparency, documentation, training and meaningful human oversight. Reality Check: Simply putting a human in the loop isn’t enough; regulators expect that person to understand, critically assess, challenge and potentially override the AI’s recommendation.

Connecticut’s new AI law will require employers triggering federal WARN requirements to tell the state whether AI or other technological change contributed to a mass layoff beginning October 1, 2026, with broader notice requirements for AI-influenced hiring, promotion, discipline, training, and termination decisions following in 2027. Why it matters: Connecticut is pushing AI transparency beyond how employers make individual talent decisions to whether technology is contributing to workforce reductions. Reality Check: The layoff disclosure applies only to reductions triggering federal WARN requirements, while most of the law’s requirements governing AI-assisted employment decisions do not take effect until October 2027.

Ema raised a $77 million Series B to expand its platform of AI agents that automate workflows across HR, IT, and finance, with deployments including an employee assistant supporting 240,000+ Wipro employees. Why it matters: The funding reflects continued investor interest in AI agents that move beyond answering questions to executing work across enterprise systems. Reality Check: Most of the performance figures come from Ema itself, and terms like “AI employees” are largely vendor positioning; the more meaningful test is whether these agents reliably automate end-to-end work at scale and deliver measurable savings versus existing software and processes.

Thanks for reading. If you have comments or feedback just respond to this email.

Until next week,

Mike & Adam

HRBusinessPartner AI is published by PotomacHarbor. Follow us on LinkedIn.

Love HRBusinessPartner AI? Share your referral link below to get HRBusinessPartner AI swag!

{{ rp_refer_url }}

Someone forward you this email? Subscribe for weekly updates.

Want to advertise in HRBusinessPartner AI? Send us an email (reply to this message).