Day & Ross leverage Pythian’s AI consulting services and Google’s Vertex AI 

Founded in 1950, Day & Ross is a North American transportation and logistics firm and one of the largest in Canada. With over 7,500 team members, the company specializes in cross-border, truckload, and less-than-truckload (LTL) (which consolidates smaller shipments from multiple shippers to the same destination), as well as dedicated fleets and residential delivery.

Day & Ross was determined to overcome manual data entry tasks and inconsistent freight document formats

With a multi-year technology and business transformation underway, Day & Ross had an urgent need to ensure the smooth implementation of their new Transportation Management System (TMS). Any need for manual processing of paperwork posed a challenge for the real-time data feature of the new TMS.

The team recognized the need to adjust their approach for the full trailer loads of shipments from customers who didnʼt provide advanced electronic visibility to the trailer contents. The time-consuming and complex manual data entry required for those shipments would create a backlog of freight in their terminals and delay drivers.

These full trailer loads contain between 20 and 120 Bills of Lading (BOL), each representing a shipment. The manual processing of BOLs—which contain vital shipment data such as ship-to address, tracking number, total weight and number of pieces—required significant data entry into the companyʼs TMS to label, unload and track shipments.

The time required to manually enter the data from these BOL documents would accumulate quickly—with each document taking one minute or more to enter.

The inconsistent formatting of most BOLs—which arrive at shipping terminals in a variety of formats, including handwritten or upside-down notes—led to large numbers of variations within each document. This made document processing automation a huge challenge, which typically relies on clean source data with fixed and predictable formats.

Automated freight documentation processing with Google Vertex AI

Pythian is a Premier Google Cloud Partner with expertise in Vertex AI implementation and rich experience in automated data extraction within the transportation business. Discovering the challenges that Day & Ross were facing, Pythian suggested the implementation of Google Cloud and key Google AI tools, technologies and infrastructure to shrink the driver dwell time and improve shipment throughput via data extraction automation. Based on years of process design expertise along with deep experience combining various Google solutions to tackle unique challenges, the Pythian team brought an innovative approach that directly addressed Day & Rossʼ challenges and complemented their agility, responsiveness, and tradition of innovation.

Pythian used Googleʼs Gemini 1.5 Pro multimodal large language model (LLM), which can accommodate a range of inputs, such as video and images, and not just predictable text. Vertex AI was used to create an all-in-one solution to automate data extraction from scanned BOLs, persist with the extracted data, validate the data against a labeled dataset provided by the client, and interface with the companyʼs TMS to validate accuracy and create real-time shipment data.

• The LLM was prompted with a set of text and formatting instructions, business context, and sample BOL images, then asked to perform a set of tasks.
• These tasks include scanning documents to locate specific pieces of information such as delivery status or tracking number, regardless of their location in the document image.
• After Gemini processes the BOLs, they're then converted into structured data and ingested into the TMS using Cloud Functions to automatically create a freight bill and either populate an existing record or create a new record.

After running a successful proof of concept (PoC) for Day & Ross, the Pythian team then created a new Google Cloud landing zone—a scalable configuration of various Google Cloud products into a coherent, effective, business-ready system—to transform it into a production-ready application.

This successful engagement and production-ready environment allowed Day & Ross to explore many future AI use cases.

• Cloud Operations Suite
• Google Vertex AI (Gemini 1.5 Pro multimodal AI model)
• Cloud Functions
• Cloud Run
• Cloud Storage
• Identity and Access Management (IAM)
• BigQuery
• Pub/Sub
• Eventarc
• Secret Manager
• Cloud Logging
• Cloud Monitoring