United StatesEnergy and utilities2025 – ongoing

Williams

Natural language data queries for commercial and operations teams

Williams cut its analyst data-request backlog by 90% after staff began asking questions of their data in plain language

WorkingVendor case studyVendor ↗

Problem

Employees at the natural gas infrastructure company depended on a small group of analysts for data requests. The backlog could reach ten requests a week, and a single request covering 25 attributes took an analyst five days.

Approach

Williams rolled out Databricks Genie, a natural language query tool, with subject-matter-expert Genie Agents for contracts, billing, allocations and geolocation. It consolidated 27 tables into large SQL models and governed access through Unity Catalog. Genie was set up to interpret the company's internal terms and business logic.

Results

90%reduction in the analyst backlog for data requestsVendor case study[1]

secondstime for a request that took an analyst five days and covered 25 attributesVendor case study[1]

1 to 2 a weekanalyst data requests after rollout, down from up to ten a weekVendor case study[1]

Takeaway

Plain-language queries moved routine data requests from the analyst queue to the people asking the questions. Analysts were left with the one or two requests a week that needed their skills.

Sources

Unlocking natural language analytics across the enterprise
Databricks17 Sep 2025Vendor publication

Source summaries

[1]Databricks17 Sep 2025

Databricks describes how Williams, a North American natural gas infrastructure provider, used Genie and subject-matter-expert Genie Agents to let more than 100 employees run their own analytics. It reports a 90% drop in the analyst backlog and a five-day request finishing in seconds. The backlog fell from up to ten requests a week to one or two.

Full details are in the original sources.

Added 4 Nov 2026 · Updated 4 Nov 2026