Posted in

Faster decisions: How an AI agent is redefining executive…

by ailona•

[LUCA Bot is an Azure OpenAI–based financial agent at Cemex giving 100+ executives instant access to granular KPIs, real-time financial analysis and self-service insights. It integrates Azure services, enforces access controls, processes 120+ KPIs, and accelerates data-driven executive decisions.]

Cemex has deployed LUCA Bot, an AI-powered financial agent for senior executives. The agent centralizes confidential financial and operational data for fast, reliable decision support.

Main feature/change and impact

LUCA Bot integrates over a decade of granular financial and operational data into a natural language interface. It serves 100 senior leaders with secure, role-limited access to 120+ KPIs. Built on Azure OpenAI and Microsoft Foundry, the agent reduces time spent searching reports. Executives receive near-real-time financial analyses and KPI comparisons across region, country, and plant.

Practical implications

Executives now replace phone chains and report hunting with instant chat queries. The agent returns structured analyses and benchmarked KPIs within seconds. This improves controllership visibility and speeds activation of operational levers. Monthly updates and daily training on executive feedback maintain data relevance and model accuracy. Access controls and session logging preserve confidentiality and auditability.
“It’s an improvement in financial information visibility that enables you to activate operational levers.”
Cemex plans staged expansion to lower management and a separate employee-facing agent using verified external data. Next steps include broader rollout, additional KPI inclusion, and continuous accuracy benchmarking. Technical teams should prioritize governance, prompt engineering, and secure data pipelines before extending access.

Key points from the article:

  • Deployed to 100+ senior leaders for executive decision support.
  • Built on Azure OpenAI, Azure AI Search, Cosmos DB, App Service.
  • Processes over 120 KPIs with regional, country and plant granularity.
  • Access controls restrict data by user region and business line.
  • Handles 400–500 queries monthly with ongoing accuracy benchmarking.

    • Related Coverage:

    From the Source