On December 3rd, National Taiwan University's Global MBA program hosted Ms. Vivian Liu from Oracle as part of the Industry Mentorship Program's Guest Speaker Series. Ms. Liu, an NTU EMBA alumna from the 2004 cohort, delivered an insightful presentation on enterprise AI transformation titled "From Prompt to Profit."

Ms. Liu's central message was clear: organizations must shift from merely experimenting with AI to generating measurable business value. She highlighted a sobering statistic: 87% of AI projects fail to reach production, attributing this not to technological limitations but to strategic and architectural challenges. The core problem, she explained, is "pilot purgatory," where companies remain stuck in endless experimentation without achieving enterprise-scale deployment.

The presentation was structured around two major transformations. The first, "From Prompt to Profit," addressed the production challenge. Ms. Liu introduced the concept of "data gravity," explaining that moving massive datasets to AI models is costly, slow, and risky. Instead, Oracle's approach brings AI to the data through what she called the "converged database": a unified system handling vectors, graphs, documents, and relational data within a single secure environment. This architecture eliminates the "integration tax" that enterprises pay when managing multiple specialized databases.

The second transformation, "From Reporting to Reinvention," focused on the rise of agentic AI. Ms. Liu distinguished between chatbots that simply answer questions and AI agents that can plan, reason, and execute multi-step business processes autonomously. She introduced concepts like RAG (Retrieval-Augmented Generation), which grounds AI responses in enterprise data, and MCP (Model Context Protocol), an emerging standard for AI agent interoperability.

Ms. Liu concluded with strategic takeaways for future business leaders: think "prompt to profit" by demanding clear paths to production and ROI; move AI to the data rather than vice versa; recognize that context is king: models are commodities while proprietary data creates competitive advantage; and build agentic systems rather than simple chatbots.

The Q&A session was particularly engaging, with students asking about Oracle's multi-cloud strategy, on-premise versus cloud deployment considerations, and how enterprises can escape the experimentation phase. Ms. Liu emphasized that successful AI projects require executive-level sponsorship, clearly defined business impact, and domain-specific use cases rather than generic knowledge-based applications.

The session concluded with a certificate presentation, recognizing Ms. Liu's valuable contribution to the GMBA community.

(Written by GMBA student - Kevin Lin 林愷)