从救火到建设：AI agents 如何恢复我们团队的核心生产力
Grab’s Analytics Data Warehouse (ADW) team supports over 1,000 users each month. These users support an extensive repository of more than 15,000 tables, which powers approximately 50% of all queries within our data lake.
However, the manual process of addressing “quick questions” is time-consuming and labor-intensive, thus creating a bottleneck in our operations.
Grab’s Analytics Data Warehouse (ADW) 团队每月支持超过 1,000 名用户。这些用户支持一个包含超过 15,000 个表的庞大存储库，该存储库为我们数据湖中大约 50% 的所有查询提供动力。
然而，手动处理「快速问题」的过程耗时且劳动密集，从而在我们的运营中造成了瓶颈。
The team was drowning in repetitive requests, spending approximately 40% of their time or an equivalent of roughly 2 days every week, on tasks like:
团队淹没在重复请求中，大约花费 40% 的时间，或每周相当于大约 2 天，在诸如以下任务上：
- Answering the same questions about data definitions
- 回答关于数据定义的相同问题
- Tracing data sources and troubleshooting
- 追踪数据源和故障排除
- Running quality checks to verify data integrity
- 运行质量检查以验证数据完整性
- Basic enhancement requests
- 基本增强请求
We deployed a multi-agent AI system that autonomously answers simpler questions and collaboratively addresses more complex requests. This led us to reclaim significant engineering bandwidth and unlock hundreds of hours of productivity monthly.
我们部署了一个 multi-agent AI system，它自主回答较简单的问题，并协作处理更复杂的请求。这让我们重新获得了大量的工程带宽，并每月解锁数百小时的生产力。
Introduction
引言
It’s 5:00 PM on a Friday. You’re wrapping up for the week when you receive a Slack message: “The vehicle_id in our production table looks gibberish. Is the pipeline broken?”
这是周五下午 5:00。你正在结束一周的工作时，收到一条 Slack 消息：“我们生产表中的 vehicle_id 看起来是乱码。pipeline 坏了吗？”
We tracked the anatomy of these ‘simple’ questions. It involved a fragmented journey through data catalogs, manual lineage tracing, SQL validation, and log diving. By the time a stakeholder received an answer, hours of high-value engineering time had been diverted into investigative overhead.
我们追踪了这些“简单”问题的剖析。它涉及通过 data catalogs、手动 lineage tracing、SQL 验证和 log diving 的碎片化之旅。到利益相关者收到答案时，数小时的高价值工程时间已被转移到调查开销中。
This process consumed nearly half of our team’s bandwidth. We recognized that while problems differed, the pro...