Particle.news

Download on the App Store

Agentic AI Enters Build Phase as Google and OpenAI Roll Out Toolkits, Enterprises Tackle the ‘Plumbing’

Fresh frameworks promise operational autonomy, yet executives warn that measurable ROI depends on integration, governance and reworked workflows.

Overview

  • At DevSparks in Hyderabad, Google Cloud showcased its Agent Development Kit, the Agent2Agent protocol and Vertex AI Agent Builder to standardize, secure and scale multi‑agent systems.
  • OpenAI highlighted agentic capabilities with its Responses API for tool use and code execution, new multimodal reasoning models, and the open‑source Codex CLI for autonomous repository tasks.
  • Enterprise leaders at a Google Cloud roundtable described a persistent “pilot problem,” urging workflow redesign, clear guardrails and human handoffs to move high‑impact use cases into production.
  • New reporting stresses the 90/10 reality of agentic deployments, with most effort going to integration, state and error management, monitoring and UX rather than model selection.
  • Surveys cite accuracy, oversight and privacy as top concerns, Gartner projects agentic features in one‑third of enterprise apps by 2028, and healthcare pilots target claims, care coordination and utilization review.