Investors should prioritize MongoDB (MDB) as a high-conviction infrastructure play, as its vector search capabilities provide the essential "plumbing" for AI agents to function. Microsoft (MSFT) remains a dominant force through bundling, but its moat faces long-term risks if the industry shifts from per-seat licensing to consumption-based pricing. For enterprise software exposure, look toward Asana (ASAN), which is successfully transitioning from a simple interface to a workflow "operating system" for human-AI collaboration. The most significant cost-saving trend is the rise of open-source models, which are expected to handle the majority of enterprise workloads within three years due to a 10x cost advantage over OpenAI. While Anthropic is expanding its ecosystem through its Model Context Protocol (MCP), the underlying "intelligence" layer is rapidly commoditizing, shifting value toward platforms like Glean that own the company-specific data context.

Glean (Private)

• Glean is an enterprise AI platform that started as a "Google for your work life," providing semantic search across a company's internal systems (e.g., Slack, Jira, Drive).
• The company has evolved into a co-worker agent that combines models like GPT-4, Claude, and Gemini into a single interface connected to company context.
• Glean acts as a "model router," selecting the most cost-effective model (including open-source) for specific enterprise tasks to help customers control AI spending.

Takeaways

• Context is the Moat: The value in enterprise AI is not the model itself, but the ability to connect that model to a company’s specific data and workflows (RAG - Retrieval-Augmented Generation).
• First-Mover Advantage: Glean benefits from being the first to bring semantic search and RAG to the enterprise, creating a brand that allows them to compete with giants like Microsoft.
• Cost Management: As enterprises hit budget ceilings for AI, platforms that offer cost-optimization by routing tasks to cheaper models will see higher adoption.

Microsoft (MSFT)

• Microsoft is identified as Glean's most formidable competitor due to its "bundling" strategy with Microsoft 365 Copilot.
• The primary competitive threat is the "path of least resistance" for enterprises; many large companies prefer one approved vendor over managing 15 different AI startups.
• However, the guest argues that Microsoft's bundling advantage may weaken as the industry moves toward consumption-based pricing (paying per task) rather than per-seat licenses.

Takeaways

• Bundling vs. Best-of-Breed: While Microsoft has a massive distribution advantage, there is still significant room for "best-of-breed" software like Glean that performs specific tasks (like internal search) better than a general bundle.
• Pricing Shift: Investors should watch if AI shifts entirely to consumption models, as this could disrupt the traditional SaaS "bundle" moat that Microsoft relies on.

Frontier Model Providers: OpenAI & Anthropic

• Enterprises are increasingly skeptical of frontier model providers due to "operational dependence"—if an AI agent learns a company's workflows, the company becomes tethered to that provider.
• Anthropic is moving into vertical applications (Legal, Health, Finance), but the guest views these as "shallow" implementations that currently expand the market rather than cannibalizing specialized tools like Figma.
• Frontier models are currently "absurdly expensive" for certain automated tasks, leading to a push for cheaper alternatives.

Takeaways

• Commoditization of the Model Layer: 90% of enterprise use cases can be handled by non-frontier or open-source models. This suggests the "intelligence" itself is becoming a commodity, putting pricing pressure on OpenAI and Anthropic.
• Ecosystem Play: Anthropic is successfully building an ecosystem via its MCP (Model Context Protocol), allowing developers to build automations directly on their platform.

Open Source & Chinese AI Models (DeepSeek/GLM)

• A major inflection point has been reached where open-source models are now within three months of the capabilities of top-tier proprietary models.
• Chinese models (like GLM 5.2) are now considered high-quality enough for majority workloads, though "sovereignty" and security concerns (backdoors) remain a barrier for US enterprises.
• The guest predicts that the majority of enterprise workloads will run on open-source models within three years due to the 10x cost advantage.

Takeaways

• Cost-Efficiency: For investors, the "winner" in AI may not be the most intelligent model, but the one that provides "good enough" intelligence at the lowest cost.
• Geopolitical Risk: There is a growing tension between using superior/cheaper Chinese open-source models and the desire for "sovereign" US-based technology.

MongoDB (MDB) & Asana (ASAN)

• MongoDB is highlighted as a critical infrastructure layer for AI agents, providing the "vector search" and real-time data reasoning necessary for agents to have memory.
• Asana is positioned as the "operating system" for human-agent collaboration, focusing on integrating AI into actual team workflows rather than just providing a chat interface.

Takeaways

• Infrastructure over Apps: Companies like MongoDB that provide the "plumbing" for AI (storing and searching data) are currently seeing high conviction from developers.
• Workflow Integration: The next phase of AI ROI will come from companies that bridge the gap between "cool AI demos" and actual business workflows.

Investment Themes & Sector Insights

The "Land Grab" vs. Discipline

• The AI sector is currently in a "land grab" phase where market share is prioritized over immediate profitability. However, the guest warns that the "overabundance of capital" is leading to unsustainable startup structures (e.g., paying $500k for junior engineers).

Team Sizes and Productivity

• Bullish on Large Teams: Contrary to the popular "1-man unicorn" theory, the guest believes the most successful companies will continue to have large teams. AI will be used to do 10x more work rather than just cutting 90% of staff.
• Role Generalization: Expect a rise in "composite roles" (e.g., an engineer who also does design and PM work) and the disappearance of specialized "analyst" roles (data entry, basic dashboarding).

AI ROI (Return on Investment)

• Current AI ROI is strongest in Customer Support (measurable productivity) and Coding (speed of writing).
• Risk Factor: While coding speed has increased, "shipping speed" of products has not necessarily followed, as human review and maintenance remain bottlenecks.