Uber's Uday Kiran Medisetty discusses AI cost metrics at Summit
Summary
At the Agentic Infrastructure Stack Summit on October 28 in San Francisco, Uber’s Uday Kiran Medisetty emphasized that raw token costs are not the right metric for evaluating AI agents, advocating instead for the measurement of cost per successful task. This aligns with a growing trend in enterprise discussions, which prioritize success-oriented metrics over traditional compute expenses when assessing AI agents. The summit aims to address critical areas such as orchestration and economic frameworks necessary for implementing agentic AI in production environments, reflecting the industry's focus on improving operational efficiency in this rapidly evolving field.
Analysis
Uber: Uber Technologies operates a global platform for ride-sharing, food delivery, and freight services with expanding investments in AI for operational efficiency and automation. Uday Kiran Medisetty, representing Uber, is addressing cost measurement practices for AI agents at the upcoming summit. This highlights how large enterprises like Uber are refining approaches to AI deployment in production settings. VentureBeat: VentureBeat is a technology media and research firm covering AI, enterprise software, and infrastructure topics while organizing specialized industry events. It is producing the Agentic Infrastructure Stack Summit to deliver proprietary research and facilitate direct discussions among technical leaders. The event underscores practical insights from companies deploying AI agents at scale. Uday Kiran Medisetty: Uday Kiran Medisetty serves as a practitioner focused on AI infrastructure and agentic systems at Uber. He is scheduled to speak at the Agentic Infrastructure Stack Summit on practical alternatives to raw token cost metrics for evaluating AI agent performance. His contribution centers on tracking cost per successful task as a more relevant measure in enterprise AI operations. Agentic Infrastructure Stack Summit: The Agentic Infrastructure Stack Summit is a VentureBeat-hosted, invitation-only event for senior enterprise IT and AI leaders taking place on October 28 in San Francisco. It focuses on real-world challenges in running agentic AI systems, including inference economics, memory architectures, and workload orchestration. The program features practitioner talks such as the one from Uber on AI cost metrics. AI Cost Metrics: Enterprise discussions increasingly emphasize success-oriented measures such as cost per completed task over raw token or compute expenses when evaluating AI agents. Agentic Systems Focus: Industry events are highlighting the need for robust orchestration, memory management, and economic frameworks to support agentic AI moving into production environments.
Categories
aimachine_learningai_agents