Particle.news
Download on the App Store

AI Deployment Accelerates as Banks Report Gains and Agent Infrastructure Launches

Hosted MCP connectors begin linking agents to real tools, reflecting a turn toward governed, interoperable AI.

Overview

  • Executives at major U.S. banks cited measurable productivity gains, with JPMorgan saying output rose from 3% to 6% and Citi reporting a 9% lift in coding, as Wells Fargo noted doing more work without current headcount cuts but flagged potential for fewer-staffed operations.
  • A Bloomberg Intelligence survey found over 70% of financial-services leaders expect higher operating costs in the first three years of AI adoption and two-thirds foresee initial staff increases, with efficiency benefits expected to materialize later.
  • The Linux Foundation launched the Agentic AI Foundation as a neutral home for agent standards, with initial contributions including Anthropic’s MCP, Block’s Goose framework, and OpenAI’s AGENTS.md, and members such as AWS, Google, Bloomberg, and Cloudflare.
  • Google introduced a hosted Model Context Protocol server in public preview that lets agents access services like Maps, BigQuery, Compute Engine, and Kubernetes Engine via a managed endpoint, with Cloud IAM and Model Armor protections and general availability targeted early next year.
  • Bloomberg reported that Meta is developing a closed-source flagship model codenamed Avocado for a possible spring debut, with training reportedly drawing on capabilities from Google’s Gemma, OpenAI’s gpt-oss, and Alibaba’s Qwen.