Automation of order validation and email processing with agentic AI

Contrôles Laurentide, a well-established Quebec company in the field of industrial automation, is managing a growing volume of customer orders and emails. With more than 16,000 purchase orders processed since 2019 and a client base of over 1,000, each order validation requires significant human effort.

Customer service handles a high volume of uncategorized emails, which must be opened, read, and manually redirected, hampering productivity. At the same time, purchase order validation relies on manual comparison between customer POs and ERP data, involving numerous checks (addresses, parts, quantities, prices, terms, etc.) from often unstructured documents. This manual process is time-consuming, error-prone, and ties up resources on low-value tasks.

To address these challenges, with the support of Videns, two AI solutions were deployed: automatic email classification and document processing automation, to streamline critical validations and handle requests more efficiently without compromising quality.

Since the launch of these two projects leveraging the full potential of agentic AI, testing phases showed performance levels exceeding 50% on incoming email classification, with an accuracy rate of 80%. This allows the company to speed up processing and significantly improve customer service responsiveness. In addition, nearly 50% of purchase order checks can now be automated, cutting processing time in half and reducing associated costs by 40%. The time saved enables teams to refocus on higher-value customer interactions. This transformation is set to improve operational efficiency while maintaining the quality standards that have built Contrôles Laurentide’s reputation.

“This project has transformed the way we work thanks to AI, especially agentic AI. We managed to find solutions to automate complex tasks without compromising quality, giving our teams back the time and space to focus on what truly matters to us: creating value for our customers,” emphasized Minh Dat Nguyen, Director of Operations at Contrôles Laurentide.

For incoming email management, the chosen approach aims to gain an overall view of interaction types, automatically classify messages with high confidence, and minimize human intervention. Regarding purchase order validation, the goal is to process a large portion of POs autonomously, drastically reduce the time required for data matching, and cut the costs tied to this process. Both areas are driven by the same intent: to refocus human efforts where they create the most value, while enhancing the company’s operational agility.

The main challenge lay in the wide variety of document formats sent by clients, often personalized and handwritten, carrying risks of errors. Matching this data with information from distinct systems such as Salesforce or IFS is complex, especially when documents are numerous and unstructured. Moreover, incoming emails are neither sorted nor categorized, preventing any automation without a fine-grained understanding of the content.