How-to

Harrison Chase:数据定义不一致是企业智能体的真正瓶颈

Why we picked it

LangChain 的 Harrison Chase 和 Sridhar Ramaswmy 就企业知识管理吵了一小架:一个说要建 knowledge factory 集中指标定义,一个说先把五个版本的"revenue"定义统一再说。给智能体接公司数据的开发者尤其该看,这坑很常见。

LangChain 创始人 Harrison Chase 在 X 上提出,企业需要搭建"knowledge factory",把散落在各系统中的指标定义和取数逻辑集中起来,让全公司统一使用。他举例称,LangChain 内部就维护了大量指标定义和从不同系统提取数据的技能。Sridhar Ramaswamy 回应指出,多数公司其实不需要 knowledge factory,而是要先解决一个更基础的问题:同一个"revenue"在公司里存在五个互相矛盾的定义。RamaswmySridhar 补充说,智能体不会制造新问题,只是让这类数据定义混乱以更快的速度暴露出来、付出更高代价。

Body

这段讨论指向企业落地智能体的一个常见误区:在数据治理尚未完成时急于部署自动化。两方观点并不完全对立——Chase 强调把隐性知识显性化并集中管理的价值,RamaswmySridhar 则提醒治理应从最小集开始,比如先统一核心业务指标的定义。对正在给智能体接内部数据系统的团队来说,指标口径不统一会让同一个问题的答案因取数路径不同而不同,这类故障在自动化场景下会被成倍放大。

Original · Harrison Chase

having inconsistent ways to define revenue is exactly why you'd need knowledge factory for your company! centralize tacit or implicit knowledge, make those readily available to everyone a big part of ours at langchain is a bunch of definitions of metrics/skills for fetching data from different systems sridhar @RamaswmySridhar agree, one edit: most companies don't need a knowledge factory. they need to stop having five inconsistent definitions of "revenue." agents just make that problem expensive faster. 🔗 View Quoted Tweet📊 13 ⚡

Industry Coordinates & Source Verification

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 LangChain, knowledge management, agents, data governance.

Source attribution: curated from Harrison Chase (@hwchase17).