How To Manage AI Use Case Portfolios

Hello and welcome to the latest edition of Banking On AI Strategy. In this edition we’ll review some recent industry developments in AI, and look at how leaders and innovators can manage portfolios of AI use cases in their organisation.

AI Industry Updates:

Technology: New OpenAI Models Go Live

Energy: Does AI Need Nuclear?

Technology: Microsoft Updates Focus On Agentification

Technology: New OpenAI Models Go Live

OpenAI has released two new AI models, o3 and o4-mini, which represent a major advance in reasoning and tool use within ChatGPT. These models can autonomously decide when and how to use ChatGPT’s full suite of tools including web browsing, file analysis, code execution, image generation, and visual reasoning to generate accurate, multimodal, and high-quality responses. o3 is OpenAI’s most capable reasoning model to date, setting new performance benchmarks across coding, science, and complex problem-solving, while o4-mini is a lighter, cost-efficient model optimized for high-throughput use cases with strong results in math, data science, and visual tasks. Both models leverage reinforcement learning not just to use tools, but to reason strategically about tool usage, bringing ChatGPT closer to functioning as a proactive digital agent.

The release of OpenAI’s o3 and o4-mini underscores a growing shift toward agentic AI, systems that can reason, plan, and autonomously act across multiple tools and modalities. This trend is accelerating, with similar efforts across the ecosystem (e.g. Anthropic’s Claude 3 with tool use, Google’s Gemini, Perplexity’s AI agents). For the financial services sector, this wave portends a near-future where client advisors, analysts, and operational teams can offload entire workflows—like portfolio diagnostics, regulatory reporting, or risk scenario modelling, to intelligent agents. These agents won’t just assist; they’ll execute, adapt, and deliver results across structured and unstructured data. Over the next year, expect experimentation with agentic copilots across trading desks, audit teams, and customer service channels, accompanied by renewed focus on governance, oversight tooling, and human-in-the-loop design to mitigate risk and preserve compliance.

Energy: Does AI Need Nuclear?

Xcel CEO Bob Frenzel is advocating for a renewed push toward building large-scale nuclear power plants in the U.S., arguing that the surging energy demands driven by AI data centers, industrial electrification, and new manufacturing require reliable, carbon-free, and dispatchable energy sources. While Xcel is not currently planning a new large reactor in its own territory, Frenzel sees growing political and industry openness to nuclear, particularly in states like Wisconsin and North Dakota. The challenge, he notes, will be overcoming the high costs and risks associated with large nuclear projects through coordinated partnerships between utilities, major energy users (like data centers), and government entities. His comments come as existing plants are being reconsidered for reopening, and as small modular reactors continue to attract interest, but may not be sufficient to meet the scale of future demand.

For the financial services and banking industry, this signals a long-term capital reallocation opportunity as AI-driven infrastructure growth begins reshaping utility-scale energy financing. Banks, asset managers, and private credit providers will increasingly need to assess nuclear viability as a mainstream investment again, not just through ESG or infrastructure lenses, but as a critical enabler of digital economic expansion. Large data center operators like hyperscalers and AI firms may also become co-investors or financing partners in energy assets, pushing banks to create new financing models and risk-sharing mechanisms. If AI continues to drive multi-gigawatt demand growth, institutions that can underwrite and structure deals around nuclear resurgence could gain a strategic foothold in the next wave of energy-linked financial innovation.

Technology: Microsoft Updates Focus On Agentification

The 2025 Microsoft 365 Copilot Wave 2 Spring Release introduces a significant new feature: the Agent Store, a centralized hub where users can find, pin, and interact with AI-powered agents directly within their workflow. This store includes first-party agents like Researcher and Analyst, built on OpenAI’s deep reasoning models, as well as partner-created and custom enterprise agents. These agents can assist with complex tasks—from conducting multi-step research to deriving insights from raw data—making on-demand expertise accessible in real time. Combined with enhancements like Copilot Notebooks, enterprise-wide AI search, and personal memory features, the update marks a clear shift toward seamless human-agent collaboration embedded across the Microsoft 365 ecosystem.

For the financial services sector, this marks the beginning of the agentic AI wave, where reasoning agents become integral to decision-making, compliance, and productivity at scale. Banks, asset managers, and insurers can harness these agents for everything from real-time portfolio analysis and fraud detection to policy drafting and regulatory research. The Agent Store model also sets a precedent for a modular AI workforce—where firms deploy, govern, and measure the ROI of internal and partner-built agents just as they do software or staff. This shift empowers financial institutions to scale specialist knowledge across thousands of users while maintaining control over data security, risk posture, and strategic alignment—essential in a regulated, knowledge-intensive industry. 

To continue reading this article, and to receive all future editions direct to your inbox, subscribe for premium or free access on Substack today at:

https://BankingOnAIStrategy.substack.com/