DeploymentCognitive automation2024-09-19
Uber put QueryGPT, a natural-language-to-SQL tool, into production for operations and support teams, reporting query authoring time down from about 10 to about 3 minutes (~300 daily users)
Data analystoccupation page →
Event date / reported
2024-09-19
Evidence stage
DeploymentAn employer has put it into production. Can move the baseline — weighted by scale and how similar the setting is.
Tasks this bears on
Writing queries and building dashboards
Translating 'how many users did X last month' into SQL, and wiring the result into a chart someone can refresh.
Automating✓ Evidence-backed
Where this applies
One company, limited release; Uber's platform runs about 1.2 million interactive queries a month, so this is a small share. The post itself flags hallucinated tables and columns and prompt quality as open problems.
What this means
A deployment record on the query-writing task from inside the workflow: Uber's natural-language-to-SQL tool is in production for operations and support teams, with authoring time reported down from about 10 to about 3 minutes for around 300 daily users. It supports the task page's mechanism on the deployment layer — and the post names the condition that made it work: curated 'workspaces' of tables, the semantic layer someone had to build first.
What it does not yet show
300 users against roughly 1.2 million interactive queries a month is a small corner of one company, and the users are ops and support staff, not analysts; the timing figure is self-reported. Uber's own post lists hallucinated tables and columns as open problems, so this is not evidence the tool is reliable on an uncurated warehouse, and nothing in it concerns analyst headcount.
What you can check
Take the five queries you wrote most recently and try each as a plain-English question in whatever SQL assistant your BI tool or warehouse offers, recording which ran correctly without edits; the failures will almost all be on tables with no documented definitions, and that list is your team's missing semantic layer.
Does it change the assessment?
No. The impact index is never moved by a single event. What this record did: the 1 linked task judgement above now rest on evidence instead of inference.
Source
Uber — engineering blog · verified 2026-09-10 · Claude (VOLO agent) — source text fetched and cross-checked · interpreted 2026-09-10 · Claude (VOLO agent)