Subscribe

The most interesting stories in HR & AI weekly

Google’s Own AI Team Bypasses Its HR Screening Tools

Google DeepMind’s AGI Safety and Alignment team created a separate application form to ensure candidates reach a human reviewer, warning internally that Google’s hiring system has a “non-trivial probability” of screening out CVs incorrectly or delaying them, though Google disputes that characterization. Why it matters: If one of the world’s leading AI organizations is creating a human-review workaround for automated screening, HR leaders should be cautious about relying on AI as the sole gatekeeper for talent. Reality Check: Google says its systems do not incorrectly filter candidates and described the form as a way to route resumes directly to the team rather than a workaround for a faulty system.

Meta Bets on Open AI to Challenge Rivals

Meta launched its open-weight Muse Glimmer model and plans to open the weights of its more powerful Muse Spark 1.2, as Mark Zuckerberg argues that advanced AI should be broadly accessible rather than controlled by a handful of companies. Why it matters: More capable open-weight models could let companies customize and run AI internally, potentially lowering costs and accelerating enterprise use cases including HR. Reality Check: Meta’s openness is also a competitive strategy, giving it a way to challenge proprietary leaders like OpenAI and Anthropic even if its models don’t lead on performance.

Anthropic Will Watermark Claude-Generated Text

Anthropic will embed watermarks in text and files generated by Claude and its other AI products to comply with new EU transparency requirements, with text watermarks designed to persist even when content is copied and pasted. Why it matters: As AI-generated content becomes commonplace in recruiting, employee communications, and other HR workflows, organizations may gain new ways to identify and govern when AI is being used. Reality Check: Watermarking won’t provide a foolproof AI detector; the durability of Anthropic’s watermark after users substantially edit generated text remains unclear.

JLL worked with recruiters to design its “Hire Buddy” around their biggest pain points, cutting scheduling from three days to 15 minutes and reportedly reducing time-to-fill for some roles from 45 days to 10. Key Insight: JLL shows how to overcome a common barrier to AI adoption: start with employees’ real workflow pain points and involve them in designing the solution, rather than introducing AI in search of a problem.

As AI increasingly handles routine knowledge work, new SHRM survey data indicates critical thinking (questioning outputs, evaluating evidence, and exercising judgment) is becoming a core workforce skill. Why it matters: HR leaders may need to rethink hiring and development around employees’ ability to work critically with AI, not simply use it. Reality Check: Critical thinking has always mattered, but AI raises the stakes because polished, confident outputs can still be incomplete or wrong.

Salesforce’s latest Agentic Enterprise Index finds companies nearly tripled their average number of active AI agents from 5 to 13 between February 2025 and April 2026, while agents expanded from an average of 2 to 6 business skills. Why it matters: The shift from AI experimentation to scaled “digital workforces” could force HR leaders to rethink roles, workflows, skills, and how human and agent work is organized. Reality Check: The findings come from organizations already running Salesforce agents in production, making them useful evidence of what leading adopters are doing but not necessarily representative of the broader enterprise market.

Rippling launched an AI Spend Console that helps customers connect employee AI spending with productivity, extending its HR and IT platform into AI workforce measurement. Why it matters: AI usage and ROI measurement could become a natural extension of the HR tech stack as companies seek to understand where AI investments are actually improving workforce productivity. Reality Check: Connecting AI spend to employee outputs can provide useful signals, but measuring individual AI ROI risks oversimplifying productivity and incentivizing the wrong behaviors.

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).