MexicoConstruction and engineering2025 – ongoing

Cemex

LUCA Bot financial-analysis agent for executives

Cemex's LUCA Bot handled 400 to 500 financial queries monthly, according to its executives

WorkingVendor case studyVendor ↗

Problem

Executives searched reports or contacted colleagues for internal financial information. Answers could take hours or days.

Approach

Cemex launched LUCA Bot in 2025 after a trial. Built with Microsoft Foundry and Azure OpenAI, it answered natural-language questions about internal financial data. Access followed each executive's authorized business scope.

Results

400 to 500queries handled monthly, per Microsoft's account of Cemex's resultsVendor case study[1]

82%reported accuracy for analysis answersVendor case study[1]

92%reported accuracy for data answersVendor case study[1]

Takeaway

Reported accuracy differed between data answers and financial analysis, leaving analysis as the weaker task.

Sources

Faster decisions: How an AI agent is redefining executive workflows at one of the world’s largest building materials companies
Microsoft12 Feb 2026Vendor publication

Source summaries

[1]Microsoft12 Feb 2026

Microsoft reported LUCA Bot's monthly query volume and separate accuracy figures for analysis and data. Cemex assessed responses against a weekly benchmark of 500 predefined questions.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026