How Enterprise AI Really Gets Deployed

This section explores Decagon's transition from relying primarily on frontier, closed-source models towards adopting open-source AI models. Jesse Zang details that their initial use of large, proprietary models was driven by immediate delivery of value, but as latency and control issues arose, they shifted focus to open source models. These smaller models offered better performance on specific tasks, faster response times, and lower costs, enabling Decagon to optimize enterprise AI applications such as voice agents and operational workflows. The discussion emphasizes that the biggest innovations in enterprise AI are happening around productization and tuning of models rather than just the models themselves. Jesse also discusses the importance of evaluating models along costs, intelligence, and latency, and how open source models can outperform larger models for specific tasks when properly fine-tuned.

This section centers on how Decagon develops AI systems tailored for large organizations. Ashwin Srinivas underscores that deploying AI at scale requires significant infrastructure—such as collaboration tools, testing, monitoring, and security protocols—to make models enterprise-ready. Jesse describes how Decagon’s approach involves building tools and procedures that turn complex, manual work into automatable processes via AI agents. The narrative highlights that successful enterprise AI deployment involves more than just powerful models; it requires embedding these models within entire systems that integrate legacy tech, ensure compliance, and facilitate continuous improvement. The discussion also touches on the ongoing evolution of AI capabilities and the need for continuous model training and evaluation to stay ahead in the market.

This section discusses the evolving landscape of AI in enterprise settings, including the role of application companies versus labs. Jesse and Ashwin debate whether deep, specialized vertical solutions or broad, horizontal platforms are more sustainable long-term. They emphasize that enterprise AI will be an ongoing process, with AI agents acting as interfaces (‘AI concierge’) integrating with CRM and other legacy systems. The conversation also covers the importance of fast iteration and product-led growth, with Decagon focusing on delivering core products that empower customers to build and adapt their own AI solutions. Additionally, they express optimism about AI creating new jobs and expanding capacity for enterprise teams by automating mundane tasks and enabling more strategic work.

In this practical section, Jesse and Ashwin share insights into how large enterprises are adopting AI faster due to the clear value proposition in cost reduction and improved customer experience. They describe their successful sales strategies, including rapid iteration, controlled deployment, and close customer collaboration. The podcast highlights that AI’s rapid progress demands agile, highly responsive teams, including founders actively involved in sales and product iteration. They reflect on how AI is transforming roles, not eliminating careers—by automating routine tasks, AI allows enterprise employees to focus on higher-value work. Judgments are also made on geographic expansion, the importance of building a scalable product, and balancing global growth with local market needs.

This section delves into Decagon’s company culture and operational practices, especially as they scale globally. Jesse and Ashwin emphasize a culture of hard work, camaraderie, and a deeply integrated team environment, where sales and engineering teams collaborate closely. They state they avoid the 'grind culture' narrative, viewing hard work as a natural part of their mission-driven approach. The company invests heavily in onboarding new offices internationally by immersing new hires in the core culture. They also highlight the importance of maintaining high accountability, continuous communication, and aligning the team around a shared vision for AI’s role in enterprise transformation.