From Natural Language Queries to Data Agents: What Changed and What Didn’t
Details
In April 2025 we showed a natural language interface to enterprise data: a carefully engineered pipeline that found the right tables, generated SQL, checked it and ran it.
Since then, the major data platforms have all started shipping data agents instead. A data agent is AI that explores your databases and documents on its own, writing and running dozens of queries to reach an answer.
This talk looks at what changed under the hood, and what didn't. We are putting a general-purpose coding agent, with no purpose-built pipeline, through public data agent benchmarks.
We also look at where the field stands. Model intelligence is rising quickly, but the new "data intelligence" benchmarks remain stubbornly hard, and vendor accuracy claims are almost all internal and unverifiable.
The questions this raises land squarely on data management:
- Which metadata helps an agent.
- What it does with real-world data quality.
- What it takes to govern a system that writes its own queries.
We bring the technology; you bring the domain. The questions we end on are yours to answer.
Join DAMA SA and guest speaker - Ndivhuwo Makonda, to learn more!