Databricks：91% 企业已部署两套以上 AI 编码工具，管理碎片化成新难题
Databricks 这篇讲企业 agent 怎么管：91% 公司都在跑两套以上 AI 编码工具，它给了个选型、上下文、管控三件套框架，搞团队落地可以参考。
Databricks 发文称 91% 的企业已在同时使用两套以上 AI 编码工具，多工具并行带来的碎片化管理比选型更难。
文中提出可扩展的智能体策略需要三个条件：可在任意模型、harness 或框架上构建的选择权，让所有 agent 共享同一受治理上下文的 Context，以及把模型、工具、MCP 服务器、skills 和 agents 统一管控的 Control。
配套给出了降低 agent 集群碎片化的实践框架。
91% of enterprises already have two or more AI coding tools active today. Managing that sprawl is the harder problem.
A scalable agent strategy needs three things:
Choice: to build on any model, harness, or framework Context: so every agent works from the same governed context Control: to govern models, tools, MCP servers, skills, and agents in one place
Get the practical framework to scale agent fleets with less fragmentation: https://t.co/AlOCgIdQWn
This report is categorized under AITOP's "How-to" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Databricks, MCP/工具, 智能体.
Source attribution: curated from Databricks (@databricks).
