From Information Access to Controlled Transactions: A Practitioner Case Study of Bilingual Conversational AI in B2B Manufacturing

This practitioner case study examines the design and deployment of a bilingual, transaction-capable Customer Portal and AI conversational assistant in an anonymized B2B manufacturing enterprise operating in a multi-system digital-transformation environment. The deployed solution provides authenticated customers with Arabic and English text and voice access to selected informational and transactional workflows. Governed application interfaces connect customer intent with authorized enterprise systems. The paper distinguishes author-reported deployed controls—identity and authorization, explicit confirmation, duplicate prevention, audit logging, and contact guidance—from a broader recommended assurance model that also includes transaction reconciliation and live human-escalation capabilities. Internal operational reporting, not independently audited for this study, indicated sustained portal adoption, broader multi-feature use, and a directional increase in tracked order initiation through the customer-facing digital channel. These portal-wide observations cannot be attributed specifically to the conversational assistant. No customer counts, transaction volumes, percentages, organizational identifiers, system names, or underlying dashboards are disclosed. The paper offers a reusable Governed Conversational Transaction Pattern, a synthetic test matrix, and a measurement framework for customer-facing AI in B2B manufacturing. Five permission-aware retrieval-augmented generation design propositions are presented while preserving authorization and transactional isolation.