Retail

Unilever’s Multi-Agent AI System Automates Customer Orders

Unilever is scaling a multi-agent AI system that automates email order processing, reduces handling time, and improves onboarding and refund workflows.

Unilever has developed NOVA, a network of five AI agents built on Microsoft Foundry to process customer orders received by email. Four agents read and interpret messages and extract order details, while a fifth checks the results. In 11 pilot markets, the system reduced the first order-to-cash stage from 15 to 20 minutes to about 90 seconds. Microsoft says 95% of orders are now processed without human intervention, with employees handling exceptions. Refund claims are processed twice as quickly, and new-customer onboarding has been reduced by 90%. NOVA has processed thousands of orders in production, and Unilever plans to expand it to more than 50 markets. The case shows how agentic AI is being applied to retail-related consumer-goods administration rather than only customer-facing experiences.

Original source: Microsoft

Related RAQI solutions

In-depth analysis on RAQI Insight

Work with RAQI