The potential of autonomous agents is substantial, promising significant gains in speed, scale, and labor efficiency. However, failures of these agents can impose abrupt and often irreversible costs. Thus, the central challenge for agentic automation is to increase the benefits of automation while bounding the consequences of failure.
While databases remain central to modern computing, agents operate over a broader data environment that spans files, APIs, applications, and system state.
In this talk, I will outline early work on Agentic Data Environments — the execution substrate in which agents operate — that amplifies agent capabilities while enforcing safety guarantees. This perspective reframes data systems from passive stores of state into active substrates for safe, reliable execution.
Blogger's Review: The concept of Agentic Data Environments provides a fresh perspective, emphasizing the importance of ensuring safety while enhancing efficiency. This approach not only advances agent technology but also offers valuable insights for the design of future computing systems.