#26 AI Integration in Investment Management
Plus: Directors' 2025 Priorities and Class Recommendation, AI Trends, Expert-Curated AI, and More
Welcome back, dear readers!
After taking some time to immerse myself in advanced AI coursework and wrap up several projects, I'm excited to return to our bi-weekly conversations. Thank you for your patience.
In this issue, I cover:
Notable Developments:
Microsoft and Google Bet Big on AI Agents for 2025
NVIDIA's Cosmos Platform Signals Robotics' Big Moment in 2025
AI in Healthcare: New Rules for Medical Devices and Insurance Practices
AI Integration in Investment Management
Directors’ Corner: Recommending Stanford's New AI Governance Program
What’s Next in AI: 3 Excellent Interviews
AI Tools Spotlight: Learning from Expert-Curated AI
Enjoy. All past issues can be viewed here.
1. Microsoft and Google Bet Big on AI Agents for 2025
Microsoft and Google are intensifying their push to integrate advanced AI agents into their productivity tools, aiming to redefine workplace efficiency in 2025. Microsoft 365 Copilot introduces customizable agents for automating tasks, analyzing data, and enhancing collaboration across its suite, including Word and Excel. Meanwhile, Google's Gemini-powered Agentspace offers multimodal AI capabilities, enabling employees to query enterprise data, automate workflows, and create specialized agents. This trend could reshape the vertical AI landscape, particularly for business intelligence and productivity tools. Their established enterprise relationships and proven track records give Microsoft and Google some advantage, especially when AI agents’ semi-autonomous nature poses new challenges to security and compliance.
2. NVIDIA's Cosmos Platform Signals Robotics' Big Moment in 2025
NVIDIA is setting the stage for robotics' breakthrough year with Cosmos, its new physical AI platform unveiled at CES 2025. The system makes robot training dramatically cheaper and faster by letting companies test robots in virtual environments instead of requiring extensive real-world trials. Rather than spending months teaching robots basic physical tasks through trial and error, businesses can now rapidly simulate thousands of scenarios in a digital space. Cosmos allows manufacturers and logistics companies to validate robot performance virtually before deployment, potentially accelerating automation adoption in 2025.
3. AI in Healthcare: New Rules for Medical Devices and Insurance Practices
U.S. regulators are taking decisive steps to govern AI in healthcare, with the FDA issuing its first comprehensive guidance for AI-enabled medical devices while states tighten oversight of insurance practices. The FDA framework emphasizes transparency and bias mitigation throughout the product lifecycle, setting guardrails for medical device innovation. Meanwhile, states like California and New Jersey are implementing stricter controls on insurers' use of AI in healthcare delivery, including bans on automated claim denials and accelerated coverage decision requirements. These parallel initiatives signal how AI regulation is evolving across both medical technology development and healthcare administration.
AI Integration in Investment Management
Investment firms have moved past the question of whether to adopt AI - Mercer's new survey of 150 asset managers shows over a third began implementing AI more than three years ago, with some pioneers starting over 15 years ago. Today, 91% of managers are either using or planning to use AI in their investment processes.
Recent academic research suggests some promising capabilities. University of Chicago researchers found large language models can outperform analysts' median estimates, particularly when following senior analyst methodologies. This makes sense - in periods of limited uncertainty, AI's lower bias and systematic approach can beat human projections.
“Does investment research make sense in the age of AI?”
Asked Juan Luis Perez, former global head of research at Morgan Stanley and former group head of research, data and analytics at UBS
However, can AI meet the late Byron Wien's definition of valuable research - making non-consensus recommendations that prove correct? In an recent Financial Times article, author Juan Luis Perez argues AI cannot anticipate major market disruptions, identify breakthrough stocks like Nvidia, or detect subtle management evasions during earnings calls. Markets are non-stationary - opinions and conditions constantly shift, requiring human intuition and flexibility.
Mercer's findings reflect these real-world complexities. While AI adoption is widespread, only 14% of managers consider it a "default and key part" of their process. Most use it to augment rather than drive decision-making. Data quality remains the biggest challenge (68%), followed by integration issues (54%) and ethical considerations (51%).
Looking ahead, managers expect AI to increase both market efficiency (60%) and market concentration (44%). They're investing primarily in machine learning, natural language processing, and generative AI capabilities, with particular focus on applications in equities, hedge funds, and digital assets.
The debate continues over AI's ability to generate true investment insight. But Mercer's data shows the practical reality: AI is becoming a standard tool in investment management, valued more for augmenting human capabilities than replacing human judgment.
A Timely Resource: Stanford's New AI Governance Program
A recent report from EY Center for Board Matters, "American Boards Priorities 2025," surveyed over 500 directors across seven countries in the Americas. Their top four priorities emerged as:
Support reshaping the portfolio to reinvest for the future
Navigate the upside and downside of technology
Assess the company's resiliency posture
Enable a talent advantage in an era of workforce instability
Notably, "Innovation and evolving technologies" ranked highest among areas where directors want both more time and information. As I've discussed in previous newsletters, this heightened focus comes at a critical time. We're seeing AI evolve beyond simple chatbots and copilots into more autonomous features embedded across business tools – creating new urgency and complexity for board governance.
In this context, Stanford's new online program "From Insight to Action: Advancing Enterprise AI Governance" launched this January, is very timely for directors.
Having viewed several modules, I'm finding it directly addresses many of the governance challenges highlighted in the EY report. This free course delivers exactly what board directors need right now - practical, current guidance on AI oversight that aligns with the increasing complexity of enterprise AI adoption.
What stands out is how the content maps to real board priorities around technology governance. I particularly appreciate how the program balances high-level strategic insights with actionable governance practices. The content is current and practical. The faculty and industry experts don't just present theoretical frameworks; they share practical experiences and lessons learned from actual boardroom discussions about AI.
Stanford has made this valuable program freely available through June 2025, with a flexible self-paced format perfect for busy directors. Optional live Q&A sessions provide opportunities to engage directly with experts on specific governance challenges.
For board members looking to strengthen their AI oversight capabilities, I highly recommend checking it out here. This is exactly the kind of focused, relevant training that helps directors ask better questions and provide more effective governance of AI initiatives.
What’s Next in AI: 3 Excellent Interviews
Three recent episodes of BG2Pod, the podcast hosted by Brad Gerstner and Bill Gurley, bring us deep insights from three technology leaders shaping the AI landscape. In separate conversations, Jensen Huang (NVIDIA CEO, Episode #17), Satya Nadella (Microsoft CEO, Episode #22), and semiconductor expert Dylan Patel (SemiAnalysis Founder, Episode #23) shared their perspectives on AI's trajectory and implications. Their highly insightful and analytical discussions moved beyond typical tech predictions to explore structural changes in computing, strategic decisions in partnerships, and the economics of AI scaling.
I highly recommend setting aside time for all three episodes. Together, they provide a comprehensive view of AI's future from complementary angles: Huang's deep dive into the full computing stack, Nadella's strategic insights on partnerships and scalability, and Patel's sharp analysis of the semiconductor landscape. While each episode runs over two hours, the depth and quality of discussion make them well worth your time – especially if you're trying to understand both the technical foundation and business implications of AI's evolution.
Here are the key themes that emerged from these conversations:
AI represents a profound platform shift, demanding reinvention, not just optimization. Jensen Huang emphasizes that AI requires a completely new computing stack, from hardware and software to algorithms and development processes. This is not merely about making existing systems faster; it's about fundamentally rethinking how we build and interact with technology.
Nvidia's "moat" is deeper than just chips; it's a full-stack advantage built on years of investment. Huang argues that Nvidia's strength lies in their deep understanding of the entire AI ecosystem. This includes not just GPUs, but also CPUs, networking, and especially the crucial software libraries and algorithms that power AI applications. This full-stack approach makes it difficult for competitors to simply replicate their success with a better chip.
Microsoft's strategic bet on OpenAI is a prime example of recognizing a "permission to play" in a new era. Satya Nadella highlights the importance of understanding a company's "structural position" when entering a new market. Microsoft recognized that OpenAI had a unique capability in large language models (LLMs), and rather than trying to replicate those capabilities internally, they chose to partner and leverage OpenAI's expertise. This strategic decision has given Microsoft early access to cutting-edge AI technology and a strong position in the emerging market for AI applications.
The future of AI is about "agents" and "digital employees," not just one-shot answers. Both Huang and Nadella envision a future where AI permeates every aspect of our lives, both personal and professional. This means moving beyond simple question-answering systems to sophisticated AI agents that can interact with us in complex ways, understand context, learn from experience, and engage in extended problem-solving.
Inference, the often-overlooked aspect of AI, is about to see surging demand. As AI models become more sophisticated and capable of "inference-time reasoning," the computational demands of running these models will grow significantly. This shift will have profound implications for the AI hardware landscape, potentially creating opportunities for competitors to Nvidia's dominance in training.
The push for ever-larger models is meeting economic constraints. While there's still momentum for larger models, Nadella suggests a growing awareness of economic constraints and the potential for diminishing returns. This indicates a shift toward efficiency, cost-effectiveness, and the ability to generate revenue from AI applications.
AI development is still in its early stages. Success for both businesses and investors depends on understanding the underlying dynamics of this rapidly evolving landscape. This requires looking beyond immediate trends to identify the core technologies, strategic partnerships, and emerging business models.
Learning from Expert-Curated AI
Here's something fascinating I've been exploring: AI tools that give us interactive access to distilled wisdom from industry experts. Let me share two examples.
1. Imagine being interviewed by an insightful podcast host
I recently came across Griffin Cook's experience with LennyBot, an AI trained on Lenny Rachitsky's extensive podcast interviews and newsletter content about product strategy and growth. Griffin, a Head of Engineering, used it to simulate an interview about his professional journey. What struck me was how this tool lets you engage with Lenny's curated insights through natural voice conversation, making years of product leadership knowledge more accessible and interactive.
2. Imagine having a seasoned executive coach guide your reflection
In another creative application, executive coach Katia Verresen has created Mana, a custom GPT trained with her proven year-end reflection framework. It's an interactive guide that walks you through her carefully developed approach to meaningful year-end reflection and future planning – making her well-sought-after insights and process available to anyone, anytime.
What excites me about these free tools is how they transform static content into dynamic conversations. They're not replacements for real mentorship, but rather new ways to explore and internalize expert insights through dialogue. You can experiment freely, without worry of judgment or time pressure. Do you have other similar tools you’d like to share?
If you’re finding this newsletter valuable, share it with a friend, and consider subscribing if you haven’t already. I greatly appreciate it.
Sincerely,
Joyce 👋
Great insight Joyce! Thank you!