Agents in production, updated Copilot and security gaps widened.
Here's what HR should do.
Welcome to FullStack HR, and an extra welcome to the 84 people who have signed up since last edition.
If you haven’t yet subscribed, join the 9700+ smart, curious and like-minded future of work people by subscribing here:
Happy Friday!
This was my first week back from summer vacation, and the autumn season kicked off with a two-day AI workshop with a client in Chicago. Chicago was hot and humid, but the workshop was 👌🏻
It’s a joy to be able to work with clients full-time now and to collaborate with those who are pushing themselves and their organisations forward.
But enough of this, and on to the update!
PS. If you'd like me to speak at your conference or host a workshop similar to the one in Chicago (e.g., moving beyond just using ChatGPT), just ping me!
OpenAI launches GPT-5
On August 7, OpenAI released GPT-5 to all ChatGPT users. The model is faster, more accurate, and stronger at reasoning and coding than previous versions. It features a massive 256,000-token context window, reduced hallucinations, and improved ability to process long documents and complex tasks. GPT-5 comes in multiple variants, including smaller, more efficient models (mini and nano) for different performance needs. There has been an outcry about the launch, but it seems like most of the dust has settled now.
Why it matters for HR: With the ability to process entire policy manuals, multi-month project histories, or complex employee datasets in one go, GPT-5 changes the scope of what can be automated. This affects admin, analytics, training material creation, and even employee self-service. HR leaders should, as always, ensure employees have AI literacy training so they can collaborate effectively with these more capable AI systems.
Read more: Wired
GPT-5 feels like a colleague, not a chatbot
A developer’s Medium post describes GPT-5 as working like a team member rather than a traditional chatbot. It can decide when to think slowly or respond instantly, switch between problem-solving modes, scan large codebases, summarise entire reports, and write production-ready code. It also now runs through a single endpoint, meaning users no longer have to pick between “fast” and “deep” models.
Why it matters for HR: When AI becomes a “colleague” that can handle mid-complexity work, the human-AI dynamic shifts. Teams will need onboarding processes for AI tools, workflows that define when to hand tasks to AI, and updated performance metrics that consider hybrid human-AI output.
Read more: Medium — Level Up Coding
Employees want to work with AI agents, not be managed by them
Workday’s new global study found that 75% of employees are comfortable working alongside AI agents, but only 30% are comfortable being managed by one. The research highlights a gap in trust when AI takes on decision-making authority, especially in areas like performance reviews and pay.
Why it matters for HR: This is a clear warning sign about employee adoption. AI can handle many operational tasks, but leadership decisions must remain human-led to preserve trust and morale. HR must define governance boundaries, train managers on responsible AI delegation, and maintain transparency about how AI is used in decision-making.
Read more: Workday newsroom
Microsoft Copilot gets Smart Mode and GPT-5
Microsoft updated Copilot with a “Smart Mode” that automatically selects the most capable model for each task, and GPT-5 is now live across its ecosystem. This means Word, Excel, Teams, and Outlook users can access GPT-5’s improved reasoning without manual configuration
Why it matters for HR: Employees will experience changes in how Copilot behaves without actively switching models. HR and L&D teams must update training materials, create example prompts relevant to internal processes, and refresh AI usage guidelines to align with the new capabilities and risks of GPT-5.
Read more: Microsoft Copilot — release notes
US government launches “USAi” for federal workers
The US General Services Administration (GSA) has rolled out a secure AI platform allowing federal agencies to use models from OpenAI, Anthropic, Google, and Meta while protecting sensitive data. The goal is to enable safe experimentation and adoption without exposing internal information to public AI systems.
Why it matters for HR: This sets a blueprint for AI adoption in large, complex organisations. HR leaders in both public and private sectors can study this approach to balance innovation with security. It also signals that AI skills will become essential for many public sector roles, requiring targeted upskilling programs.
Read more: POLITICO
Security gaps as AI agents scale
Okta’s latest survey of 260 executives found that while 91% of companies are already using AI agents, only about 10% have mature governance for non-human digital identities. Many organisations lack clear ownership, access controls, and audit trails for these agents.
Why it matters for HR: AI agents are effectively “digital employees” and should be managed as such. HR must work closely with IT to integrate AI agents into joiner-mover-leaver processes, ensure role-based access control, and clarify accountability for the actions these agents take.
Read more: Okta — AI at Work 2025
AI reduces entry-level jobs for recent graduates
Forbes reports that companies are cutting back on junior roles as AI takes over simpler tasks. A LinkedIn survey showed over 60% of managers plan to use AI to replace routine work, and firms are prioritising candidates with a few years’ experience.
Why it matters for HR: This trend threatens the talent pipeline and future leadership bench. HR must create structured internships, apprenticeships, and entry-level roles that incorporate AI from the start. They should also emphasise human skills like critical thinking, relationship-building, and creativity that AI cannot replicate.
Read more: Forbes
AI linked to growing job cuts in the US
Data from Challenger, Gray & Christmas shows 62,000 job cuts in July in the US, with over 10,000 directly attributed to AI adoption. The tech sector is the hardest hit, but other industries are also accelerating automation-driven restructuring.
Why it matters for HR: Leaders must anticipate where AI adoption will trigger redundancies and invest in reskilling and redeployment strategies before cuts happen. This includes mapping vulnerable roles and creating career transition pathways for affected employees.
Read more: Computing
Leading with empathy in AI-driven change
On the other hand, LACE Partners argues that AI is rarely the sole reason for layoffs, high operating costs, and weak demand also play a role. They propose five people-first strategies for AI transitions: communicate early, lead with empathy, provide retraining, support those who remain, and build alumni networks.
Why it matters for HR: Change handled poorly damages employer brand and retention. HR teams should embed empathy into restructuring plans and use AI adoption as an opportunity to enhance, not erode, employee capability and culture.
Read more: LACE Partners
Gemini in Google Drive can now handle 1M-token PDFs
Google has expanded Gemini’s context window in Drive to around one million tokens, enabling question-answering and summarisation of extremely large or scanned PDF documents (or even books!)
Why it matters for HR: This upgrade allows faster contract reviews, policy analysis, and compliance audits. HR teams can process vendor agreements, benefits documents, and investigation files more efficiently, but they must also ensure redaction and access controls are in place to prevent exposure of sensitive information.
Read more: TechRadar Pro