LLMs versus agents explained

In this episode of Excess Returns, we sit down with Matt Russell of Business Breakdowns to explore how AI is actually being used in investing today. We go beyond the hype and break down practical use cases for AI in portfolio management, stock research, due diligence, monitoring, and idea generation. From deep research models and agentic AI to prompt engineering and workflow design, this conversation walks through how professional investors can use AI tools to increase productivity, improve decision-making, and reduce blind spots without losing their edge. If you are an asset manager, analyst, allocator, or DIY investor wondering how AI will impact investing and stock picking, this episode offers a clear, practical roadmap.

Main topics covered:

• The evolution from early large language models to deep research and agentic AI for investors

• LLMs vs agent-based AI and why the distinction matters for investment research

• How AI fits into an investor’s workflow, from due diligence to portfolio monitoring

• Using AI to monitor KPIs, earnings calls, and cross-industry signals in real time

• How AI can help kill bad ideas faster and surface deal breakers early

• Prompt engineering for investors, including mindset framing, audience targeting, and output design

• Building mental models into AI systems to reflect your investment philosophy

• AI tech stacks for investors, including writing tools, deep research models, and browser-based AI

• Iteration, experimentation, and standardized testing of prompts across model upgrades

• The impact of AI on alpha generation, active management, and generalist vs specialist investors

• Organizational adoption strategies for investment firms considering AI

• Customization, agentic workflows, and what AI in investing could look like five years from now

Timestamps:

00:00 How AI tools increase investor productivity
01:16 Why early ChatGPT was a head fake for investors
03:07 The inflection point with deep research and agentic AI
05:00 LLMs vs agents explained in plain English
07:01 Where AI fits inside an investment workflow
09:28 Replacing manual earnings transcript work
11:40 Real-time monitoring and AI alerts
19:24 Using AI to kill bad investment ideas faster
22:01 Trust but verify, hallucinations and safeguards
25:29 Matt’s AI tech stack for investing
30:00 Prompt engineering breakthroughs
33:00 Standardized experimentation across new AI models
36:07 Building idea generation prompts step by step
40:15 Using AI as an editor and critical reviewer
43:50 Does AI compress investor skill differences
46:10 How funds should adopt AI internally
50:40 Fear of falling behind in asset management
53:05 Generalists vs specialists in an AI world
55:18 AI and the pursuit of alpha
57:00 Customization, agents and the future of investing
01:01:10 Coding agents and building tools with AI