Highlights
Top Insights
1. Agentic AI works at speeds and scales humans can’t match, which means traditional management tools built for people often don’t fit. Leaders will need new ways to track, audit, and guide these systems in real time.
2. Unlike humans, AI can’t feel guilt, be punished, or be sued. This means people, not the AI, must always remain responsible for what AI systems do, from design through deployment.
3. Things managers take for granted with people (judgment, ethical boundaries, escalation paths) need to be made explicit and coded into AI governance, since AI won’t “fill in the blanks.”
Source: Agentic AI at Scale: Redefining Management for a Superhuman Workforce (Sloan Management Review)
Top News
1. Microsoft has rolled out Copilot Chat across its Office.
2. Google announced the Agent Payments Protocol (AP2), an open standard for agent-led transactions.
3. Notion 3.0 introduces powerful AI Agents that can autonomously create and manage workflows across Notion.
4. Researchers developed Delphi-2M, a transformer-based AI that predicts disease progression across 1,000+ conditions,
5. Scientists used AI to design and synthesize the world’s first viral genomes, creating bacteriophages that can kill E. coli.
Additional Insights
1. The AI-Native Office Suite – Can AI Do Work For You? (A16Z)
The AI-native office suite is rapidly evolving into a world where AI acts as a true teammate, handling tasks from email drafting to research and financial modeling. The market is splitting between generalist tools (broad, flexible assistants like Manus and Comet) and specialist tools (deep, workflow-focused solutions like Gamma for presentations or Fyxer for email). Benchmarks reveal a clear tradeoff: vertical tools excel in polish, reliability, and external-facing use cases, while generalists provide breadth and faster adaptability across multiple tasks. For example, Gamma shines in presentation design, Manus in deep research and spreadsheets, and Mem or Notion in meeting notes. Meanwhile, Comet and Claude deliver speed in research and task execution. Overall, the competition is intensifying as horizontals race to be the “all-in-one” entry point, verticals push into adjacent categories, and convergence looms, with both sides striving to balance speed, accuracy, depth, and usability.
2. AI Is Moving Faster Than Your Workforce Strategy. Are You Ready? (BCG)
AI is reshaping work far faster than most workforce strategies can keep up: tasks are shifting from manual execution to AI-led orchestration, roles are broadening and blending, and teams are flattening into agile, AI-augmented pods. Tech workers, the first to experience these shifts, illustrate a broader trend in which AI takes on routine tasks while humans focus on strategy, oversight, and problem framing. Success now requires AI fluency, systems thinking, and judgment, while traditional pipelines, job ladders, and management structures are being redefined. Organizations must act intentionally—whether scaling tools, cautiously piloting, streamlining for speed, or fully reinventing—to align talent strategies with rising AI maturity. What seems advanced today will be baseline by 2030, and the leaders who proactively redesign roles, retrain teams, and restructure organizations around AI-human collaboration will secure lasting competitive advantage.
3. Buy, boost, or build? Choose your path to generative AI (Ideas Made to Matter)
Organizations adopting generative AI must decide whether to buy, boost, or build solutions based on their strategic objectives, resources, and capabilities. Buying offers speed and simplicity with vendor-managed tools but limits differentiation and creates dependency risks. Boosting enhances vendor models with proprietary data for greater accuracy and relevance, though it requires strong data governance and comes with higher operational costs. Building custom AI provides full control and competitive differentiation, but it is costly, complex, and demands advanced technical expertise. Success depends on aligning the chosen path with business priorities, establishing transparent governance and decision-making processes, and cultivating long-term vendor partnerships that encourage collaboration, adaptability, and innovation.
Innovation Radar
1. AI Model Releases and Advancements
Grok 4 Fast matches Grok 4’s performance with 40% fewer tokens, state-of-the-art web/X search, a 2M token context window, and a unified reasoning architecture, delivering frontier-level intelligence at dramatically lower cost (xAI).
Tencent launched Hunyuan 3D 3.0, a free AI model that triples 3D modeling accuracy with advanced geometry, texture, and character generation features (Pandaily).
H Company has released Holo1.5, a family of open-weight vision models optimized for computer-use agents, delivering state-of-the-art accuracy in GUI element localization and UI-VQA for reliable, high-resolution interface interaction (MarkTechPost).
Google Research has released VaultGemma, a 1B-parameter open language model trained from scratch with differential privacy, marking the most capable DP-trained LLM to date and advancing private, safe AI development (Google).
Reve has launched a new all-in-one AI-powered image creation and editing platform that makes advanced visual design simple, precise, and accessible to everyone (Reve).
Luma AI has launched Ray3, a generative video model that produces studio-quality HDR videos with advanced reasoning, visual control, and professional editing support, now available for free via its Dream Machine app (The Decoder).
2. AI Tools and Features
OpenAI has released GPT-5-Codex, a faster, more reliable version of GPT-5 optimized for agentic coding, real-time collaboration, and code review, now integrated across Codex CLI, IDE extensions, the cloud, and GitHub (OpenAI). OpenAI is preparing to add an “Orders” section in ChatGPT for native checkout and order tracking, alongside updates to voice mode and upcoming parental controls, signaling a shift toward making ChatGPT a multifunctional assistant beyond text generation (Testing Catalog).
Kling AI has launched a photorealistic avatar tool that turns any image and audio into multilingual, expressive, studio-quality animated avatars for as little as $1 per minute (Financial Content).
Replit’s Agent 3 is its most autonomous AI yet, capable of testing and fixing apps in-browser, running up to 200 minutes unsupervised, and even creating other agents and automations to streamline workflows (Replit).
Microsoft has rolled out Copilot Chat across Word, Excel, PowerPoint, Outlook, and OneNote, giving users an AI assistant directly inside their Microsoft 365 apps to help draft, analyze, create, and collaborate (Microsoft).
Google announced the Agent Payments Protocol (AP2), an open, payment-agnostic standard developed with over 60 partners to enable secure, authenticated, and accountable AI agent–led transactions across traditional and emerging payment systems (Google). Google is rolling out its biggest-ever Chrome upgrade by integrating Gemini AI to simplify browsing, handle tasks, enhance search, improve safety, and streamline productivity across desktop and mobile (Google).
Amazon has upgraded Seller Assistant with agentic AI, turning it into a proactive partner that helps sellers manage, optimize, and grow their businesses around the clock (Amazon).
Notion 3.0 introduces powerful AI Agents that can autonomously create, manage, and personalize workflows across Notion, with Custom Agents for teams coming soon (Notion).
Alibaba’s Tongyi Lab has released Tongyi DeepResearch Agent, a powerful open-source web research AI that rivals or outperforms much larger proprietary models in benchmarks and real-world use cases, marking a major leap for Chinese AI agents on the global stage (VentureBeat).
Researchers developed Delphi-2M, a transformer-based AI trained on large-scale health records, that predicts disease progression across 1,000+ conditions, generates synthetic health trajectories, and offers insights into comorbidities, showing promise for personalized risk prediction and precision medicine (Nature).
Scientists used AI to design and synthesize the world’s first viral genomes, creating bacteriophages that can infect and kill antibiotic-resistant E. coli, marking a step toward AI-generated life (Nature).
4. Other
Researchers have developed a novel ultrasound helmet that can non-invasively and precisely modulate deep-brain circuits, offering new possibilities for studying brain function and treating conditions like Parkinson’s disease and depression without surgery (Science Alert).
At Meta Connect 2025, the company unveiled new Ray-Ban and Oakley AI smart glasses, VR features like Hyperscape and Horizon TV, and upgrades to its Horizon Engine, highlighting advancements in wearables, entertainment, and metaverse technology (The Verge).
