• Prioritize the vertical AI theme, especially procurement automation, but favor companies that demonstrate reliable end-to-end execution, handle exceptions, and deliver measurable customer savings—not just impressive demos.
• Treat Leo as a private-company example of this opportunity, not an available investment based on the information provided; monitor production adoption, customer value, and how much human review its agents still require.
• The discussion offers no actionable stock call, price target, or timeframe for CRM, SAP, or ORCL; evaluate incumbents’ distribution advantages against focused AI firms’ ability to automate workflows across systems.

Enterprise vertical AI and procurement automation

• The central investment opportunity discussed is vertical AI: agents that take responsibility for an entire business task rather than simply retrieving information or adding a chatbot to existing software.
• Procurement is presented as a promising use case because much of the work happens outside systems of record—in emails, spreadsheets, contracts, engineering files, and conversations—and often requires coordination across procurement, finance, legal, and engineering.
• The potential value comes from handling exceptions and making decisions, not just automating routine steps. The speakers cited missed supplier-delay emails as an example of a small oversight that could delay a project and cause hundreds of millions of dollars in damage.
• The discussion also emphasizes potential operating benefits: negotiating purchases that companies previously lacked capacity to negotiate, reducing friction, and improving project timing. One speaker said that, in their framing, a 1% procurement savings could have a P&L impact comparable to generating 10% more sales.
• Actionable takeaway: When evaluating this theme, look for products that can reliably complete multi-step work across systems and handle exceptions—not merely demonstrate an impressive chatbot or prototype.
  • The speakers identified meaningful hurdles: poor-quality or fragmented company context, complex integrations and custom workflows, and the fact that performance that is not sufficiently accurate may still leave humans doing all the work.
  • Trust is also a constraint. The discussion describes gradual adoption, with human review used to build confidence before agents take on more autonomous or higher-stakes tasks.
• No public-market recommendations, price targets, or investment timelines were given.

Leo (private company; no ticker)

• Leo builds AI agents for enterprise procurement, including sourcing, invoice exceptions, supplier communications, and negotiations.
• The company’s approach is to coordinate agents across the workflow, with the level of autonomy varying according to the task’s risk and complexity.
• The CEO said Leo can handle some purchases autonomously, while complex, multi-million-dollar negotiations retain expert humans in the loop. Human feedback also helps train and evaluate the agents and build customer trust.
• Leo’s longer-term opportunity, as described in the episode, is to support work between buyers and suppliers—not just automate tasks inside one company. The speakers suggested that agents on both sides could reduce friction on shared tasks, even when the parties have conflicting interests over price.
• Actionable takeaway: Leo is a case study in the vertical-AI thesis, but the transcript provides no financial results, valuation, or evidence of investment availability. For investors assessing similar companies, relevant indicators would include measurable customer value, increasing automation in production, and the ability to handle exceptions without creating additional review work.
  • The discussion specifically flags trust, accuracy, implementation complexity, and the need for human oversight in sensitive negotiations as challenges.

Salesforce (CRM)

• Salesforce is discussed as an incumbent with meaningful customer trust and distribution. Its Agentforce product was given as an example of an agent that customers may readily try, particularly when it is easy to activate or offered at little extra cost.
• The speakers said incumbent offerings have generally started with information retrieval and simple processes, while more judgment-intensive, end-to-end work can be harder to pursue. They cited potential internal conflicts between selling workflow software and selling software that resolves the work itself.
• Actionable takeaway: The episode presents a competitive question, not a bullish or bearish stock call: incumbents may benefit from distribution, while focused startups may have room to own workflows that span multiple systems and departments. The transcript gives no Salesforce price target or stock recommendation.

SAP (SAP), Oracle (ORCL), and Coupa (private)

• SAP and Oracle are mentioned as enterprise systems that procurement agents can update or work alongside. The speakers’ point is that the system of record may show a final price or delivery date while omitting the emails, analysis, and exceptions behind it.
• Coupa is mentioned as an example of a legacy procurement system. The episode argues that many procurement tools have made established processes more efficient without fundamentally changing how people work.
• Actionable takeaway: The discussion suggests that systems of record may remain important infrastructure even if new AI applications take on more of the work around them. It does not provide a company-specific outlook, price target, or recommendation for SAP, Oracle, or Coupa.

OpenAI and Anthropic

• OpenAI and Anthropic are mentioned as model providers that incumbents may partner with. Leo’s CEO said the company uses models from multiple providers and views general-purpose models as a commodity for some tasks.
• The proposed source of differentiation is the surrounding product: integrations, workflows, evaluation processes, proprietary or industry-specific data, and the ability to deliver a reliable business outcome.
• Actionable takeaway: The discussion favors examining the application and workflow layer, rather than assuming that access to a leading general-purpose model alone creates a durable advantage. No investment recommendation or valuation view on either provider was stated.