Josh Bersin 发布 HR 多智能体 AI 架构指南，提出六项上线前治理决策
Introducing The Leader’s Guide To Multi-Agent AI Architecture
- 1Josh Bersin 发布 HR 2030 研究新报告，核心主题是 HR 多智能体架构如何治理
- 2报告主张六项业务决策属于 HR 领导层而非 IT，且须在智能体上线前敲定
- 3报告绘制多智能体 HR 技术栈，映射 8 个超级智能体与 150 多个 AI 智能体
- 4案例涵盖 ISS North America、Costa Coffee、DBS Bank、Mastercard 等 6 家企业实践
多智能体进 HR 的关键已不是用不用，而是上线前由谁定治理规则，该报告把这类决策明确归给 HR 业务层而非 IT。
- 影响：编排、监控、数据规则、人工控制等六项决策的质量，决定 HR 智能体是协同系统还是高成本的无序扩张；决策权划给 HR 领导层，意味着 HR 需在上线前建立规则与问责；同业案例可作参照，但属报告方自选实例。
- 参考：可考虑对照六项决策逐项自查本企业 HR 智能体的治理现状；值得在上线前先明确哪些环节必须保留人工控制；先盘点已部署智能体的数量与归属，评估是否已经失控蔓延。
- 依据：论点出自 Josh Bersin 的 HR 2030 研究报告，属研究机构口径，暂无独立验证；ISS 北美、DBS Bank、Mastercard 等为报告方所选案例。
by joshbersin · Published October 8, 2026 · Updated October 8, 2026
Today I’m excited to share a new additions to our HR 2030™ research: The HR Leader’s Guide to Multi-Agent AI Architecture.
Getting a multiagent HR architecture right is like conducting an orchestra. The instruments are powerful, the musicians are skilled—but without a score and a conductor, what you get isn’t music. It’s noise.
Our newest report tackles one of the most urgent questions facing HR leaders today: not whether to deploy agentic HR, but how to govern it—deliberately, and before the agents go live.
The message is clear: the organizations pulling ahead are not just adding AI tools—they are making six deliberate business decisions. And those decisions belong to HR leadership, not IT.
The report includes our HR 2030 Multiagent HR Technology Stack, which maps how 8 superagents and more than 150 AI agents connect to power the entire HR function. Against that backdrop, the report walks through six decisions that will determine whether those agents work as a coherent system or descend into costly, ungoverned sprawl: how to orchestrate and govern agents, how to monitor and secure them, what rules and data they reason from, where humans must stay in control, what capabilities to build into a shared library, and what to build, buy, or extend.
Importantly, these are not configuration questions for IT. They are business decisions about how HR operates, what the organization values, and who is accountable when an agent gets it wrong. Real-world examples from ISS North America, Great Wolf Resorts, Costa Coffee, DBS Bank, ASICS, and Mastercard show what getting them right looks like in practice.
Short on time? The two-page executive summary delivers the essential argument for leaders who want the big picture first. You can also read my in-depth discussion of this topic on my substack page.
You can also listen to a detailed discussion of the AI safety issues and why AI “agent specialization” is so important in your architecture in this week’s podcast.
For more details on our ongoing HR 2030 research program, click here. All these materials are available to Galileo® users as well as our corporate clients.
Share
来源：Josh Bersin · joshbersin.com