Investors looking to capitalize on the enterprise artificial intelligence boom should consider positioning themselves in hardware and infrastructure providers that supply the open-source community, as companies increasingly shift 90% of their production workflows away from closed-source alternatives to lower costs. While private startups like Decagon and Sierra prove that application-layer companies can build strong moats by solving complex business logic, public investors should view foundational labs like OpenAI and Anthropic as R&D leaders rather than monopolistic threats to vertical software. Instead of betting on general-purpose frontier models for long-term scale, target enterprise software firms that implement a "glass box" approach for rapid workflow customization and deployment. Watch for upcoming initial public offerings in the enterprise AI space over the next 12 to 24 months, focusing particularly on product-led businesses that effectively transition custom client solutions into scalable core software products.

Decagon (Private / Startup)

• Decagon is an enterprise AI company specializing in AI agents for customer support, sales, and operational business processes.
• The company has shifted 90% of its AI workflow to open-source models to optimize for latency, fine-tune specific tasks, and lower costs.
• They utilize frontier models (from providers like OpenAI and Anthropic) for the remaining 10% of their workflow, specifically for new, exploratory, or complex auxiliary tasks like their Duet Autopilot feature.
• Decagon operates on a "glass box" approach rather than a traditional "black box" enterprise AI deployment, allowing enterprise clients to build, test, and iterate rapidly on custom workflows.
• The business model is product-led rather than pure consulting, utilizing forward-deployed engineers to productize custom client solutions into a scalable core software product.
• Decagon competes in the enterprise AI space against peers like Sierra.

Takeaways

• Enterprise AI adoption is transitioning toward open-source models for mature, high-volume, and production-scale workflows because they offer better speed, cost-effectiveness, and task-specific fine-tuning capabilities compared to general-purpose frontier models.
• Companies building application layers on top of AI foundation models can carve out strong moats by solving complex business logic, integration challenges, and compliance needs rather than just acting as thin user interfaces.

OpenAI (Private / Startup) and Anthropic (Private / Startup)

• OpenAI and Anthropic are recognized as the primary "Frontier Labs" providing state-of-the-art, general-purpose closed-source foundation models.
• Their models are widely used for initial experimentation, complex reasoning tasks, and new product launches across the AI industry.
• The transcript addresses and refutes the narrative that Frontier Labs will completely take over the application layer, arguing instead that specialized application companies and enterprises will continue to build customized vertical workflows using open-source models for scaling.

Takeaways

• While Frontier Labs dominate initial model intelligence and general capabilities, specialized enterprise application providers can successfully unseat or complement them at the production scale by optimizing for cost, latency, and vertical-specific business processes.