Highlights
Top Insights
- Firms may consider integrating AI tools to simulate negotiation outcomes, anticipate counterpart reactions, and accelerate decision-making in high-stakes, time-sensitive environments.
- Persona-based AI models may be used to explore diverse leadership perspectives, helping executives stress-test decisions through the eyes of competitors, regulators, or geopolitical actors.
- Companies might invest in AI systems grounded in both language modeling and game theory to enhance the rigor and foresight.
Source: AI and the art of negotiations (The Economist)
Top News
1. OpenAI has released its most advanced models yet—o3 and o4-mini, with the former scoring 136 in an IQ test, setting a new record.
2. Google DeepMind has added a “reasoning dial” to its Gemini Flash 2.5 to address the growing problem of overthinking.
3. Kling 2.0, the latest AI video generator, is making waves with its stunningly realistic and creative animations.
4. Microsoft has introduced the “computer use” feature in Copilot Studio, enabling AI agents to automate tasks on websites and desktop applications.
5. Some Spanish engineers have developed FlexiVol, the world’s first physically interactive 3D hologram system.
Additional Insights
1. Busting the myths of gamification (IDEO)
Gamification is often misunderstood as simply adding points, badges, or competition to drive engagement, but without thoughtful design, it can lead to fatigue, manipulation, and missed opportunities for meaningful impact. True game design goes deeper by leveraging intrinsic motivators like challenge, progress, and agency to foster sustained and empowering experiences. Projects from IDEO demonstrate how well-crafted game mechanics—like playful feedback tools or systems that reward effort over perfection—can effectively support goals in mental health, education, and workplace transformation. Misapplied gamification, on the other hand, risks turning experiences into shallow systems that feel more like obligations than choices. Ultimately, effective gamification isn’t about surface-level fun—it’s about designing purposeful systems that align with users’ values and lead to long-term engagement and growth.
2. Making AI work (Board of Innovation)
Many companies pursue AI without a clear strategy, often chasing hype-driven ideas that fail to deliver real ROI. To unlock true value, leaders must treat AI as a strategic capability, starting with an outside-in analysis of market trends and competitor activity while building executive fluency to avoid hype traps. Exploring AI opportunities through both business-led (optimizing current operations) and AI-led (enabling new offerings) lenses helps identify high-impact use cases. A cross-functional scan grounds ideas in operational reality, followed by rigorous prioritization based on value and feasibility. Finally, success depends on defining the enablers—data, talent, tech, and governance—needed to make top opportunities actionable and sustainable.
3. Navigating tariffs with a geopolitical nerve center (McKinsey)
Amid a surge in global tariffs and trade controls, businesses face escalating uncertainty and complex supply chain disruptions, particularly in sectors like automotive where parts sourcing is highly globalized. To manage these pressures, companies are increasingly adopting a “geopolitical nerve center”—a centralized team structure that tracks developments, guides tactical responses, and supports strategic planning across immediate, mid-term, and long-term horizons. This nerve center coordinates cross-functional teams focused on nine core initiatives such as tariff operations, supplier diversification, commercial optimization, and business portfolio shifts. Supported by analytics like tariff scenario modeling and trade flow analysis, the nerve center enables companies to swiftly adapt to policy changes and preserve competitiveness. Ultimately, this approach helps organizations shift from reactive crisis management to proactive, resilient strategy in a volatile trade environment.
Innovation Radar
1. AI Model Releases and Advancements
1.1 Video Models
ByteDance introduced Seaweed, a powerful 7B-parameter foundational video generation model that uses multimodal data to create high-resolution, realistic, and controllable videos—featuring lifelike characters, dynamic environments, synchronized audio, and consistent storytelling—for a wide range of creative and interactive applications.
Alibaba has released its latest open-source video generation model, Wan2.1-FLF2V-14B, which enables users to create realistic, high-quality videos using start and end frames, reflecting the company’s continued leadership in accessible, cutting-edge AI innovation (Alizila).
Google has launched Veo 2 for Gemini Advanced and Whisk Animate, enabling Google One AI Premium users to generate high-resolution, eight-second AI videos from text or image prompts with cinematic quality and easy sharing features (Google).
Kling 2.0, the latest AI video generator from Kuaishou, is making waves with its stunningly realistic and creative animations, positioning itself as a strong rival to OpenAI’s Sora (Tom’s Guide).
1.2 Others
OpenAI has released its most advanced models yet—o3 and o4-mini—which combine top-tier reasoning, full tool use (like web browsing, image analysis, code execution), and improved efficiency, enabling faster, smarter, and more personalized responses across complex tasks (OpenAI). The o3 model achieved a score of 136 on the Mensa Norway IQ test, outperforming Gemini 2.5 Pro and setting a new record.
Google DeepMind has added a “reasoning dial” to its Gemini Flash 2.5 to address the growing problem of overthinking, which, while potentially improving complex task performance, can also waste resources and money when applied unnecessarily (MIT Technology Review).
GPT-4.1 is a powerful new family of models—featuring standard, mini, and nano versions—that significantly improves on coding, instruction following, and long-context understanding (up to 1 million tokens), while offering lower latency, better real-world utility, and more affordable pricing for developers building intelligent, agentic applications (OpenAI).
Cohere’s Embed 4 is a powerful multimodal, multilingual embedding model that enables secure, accurate search across complex business data to support AI agents and enterprise applications (Cohere).
2. AI Tools and Features
Anthropic has introduced new capabilities to Claude—Research and Google Workspace integration—that enhance its ability to deliver fast, in-depth answers and personalized support by combining web and work context, streamlining tasks across roles from marketers to parents (Anthropic).
Google Maps is introducing new AI-powered tools to help cities and businesses manage traffic, monitor infrastructure like utility poles, and analyze location trends—aiming to improve road conditions, reduce congestion, and support smarter urban planning without always needing on-site inspections (CNET).
Airtable has launched Airtable Assistant, a conversational AI tool that lets users build and manage apps, analyze data, and automate tasks using natural language—putting powerful AI capabilities directly in the hands of everyday builders (Airtable).
Microsoft has introduced a new “computer use” feature in Copilot Studio, enabling AI agents to automate tasks on websites and desktop applications by interacting with user interfaces like a human would, even without APIs (Microsoft).
Grok has introduced Grok Studio, a new canvas-like tool that lets users collaboratively create and edit documents, code, and simple apps, now with Google Drive integration for working with various file types (TechCrunch).
OpenAI has added a new “Library” section to ChatGPT, allowing users to easily view and manage their AI-generated images on both mobile and web (The Verge).
3. AI for Research and Social Impact
The integration of diverse omics data into multimodal foundation models promises to revolutionize molecular biology by enabling AI-driven insights into cellular states, gene regulation, and biomarker discovery, ultimately advancing our understanding of life sciences (Nature).
Google’s DolphinGemma, an AI model developed with the Wild Dolphin Project and Georgia Tech, is helping researchers analyze and predict dolphin vocalizations, potentially enabling two-way communication and deepening our understanding of dolphin language (Google).
ProGen3 is a suite of large-scale AI models for protein generation that enables single-shot design of diverse, high-functioning antibodies and compact gene editors, demonstrating real-world benefits of scaling laws in biological design through both computational and experimental validation (ProFluent).
4. Other
A team of Spanish engineers has developed FlexiVol, the world’s first physically interactive 3D hologram system using elastic diffusers, allowing users to intuitively manipulate holographic objects with familiar gestures like swiping and pinching—bringing us closer to sci-fi-style hologram interaction (New Atlas).
Leave a Reply