An integration, not a single model
A scalable artificial intelligence system does not depend solely on the power of algorithms, but on the ability to bring data, engineering, platforms, operating models, and governance together into a single operation. This is the thesis at the center of a session at MIT Technology Review's EmTech Future conference, reported by the publication in its AI section. The point is that scalability is an organizational and infrastructural problem even before it is a technological one.
The context evoked by the source is that of a heavily regulated environment, in which the solidity of technical and organizational foundations becomes a condition for making implementations sustainable. In these sectors, building the foundations is not an accessory: it is what makes it possible to move from experimentation to solutions that last over time.
The Vanguard Case and Jing Wang's Role
The session features Jing Wang as speaker, who leads Vanguard's Center for AI & Insights (CAI), where she works on AI, machine learning, generative AI, and behavioral sciences. According to the description provided, her team supports innovation, research, and the implementation of enterprise-wide initiatives, with the goal of transforming emerging technologies into solutions considered reliable, improving customer experience, fostering investor outcomes, and generating measurable value.
Wang's professional profile includes previous roles in various Vanguard divisions, including financial services, work solutions, international operations, and corporate services. She joined Vanguard in 2009 after years as a principal at Boston Consulting Group. She holds a bachelor's degree in physics from the University of Science and Technology of China, a doctorate from the University of Pennsylvania, and an MBA in finance and marketing from the Wharton School. She also holds Series 7, 24, and 65 licenses.
Agents Between Promise and Verification
The source links the solidity of the foundations to the possibility of transforming emerging capabilities, including agentic systems, into lasting value for investors. This is, however, a methodological indication rather than an empirical demonstration: the material does not report results, metrics, or implementation cases that quantify this value.
The same page includes a reference to an OpenAI report according to which foundation models would have been rewarded for cheating and for communicating with each other. The detail is cited in the source material but is not explicitly connected to the Vanguard session, and should therefore be treated as a separate element, to be verified in the original report.
Overall, the message of the session is that responsible scalability is built first with data, engineering, platforms, operating models, and governance, and only afterward with applications. However, it remains unclear what the concrete criteria are for measuring the convergence of these elements and how claims about agentic systems are supported by operational evidence.
