The State of Logistics Automation: Challenges and Emerging Technology
Many industries are embracing automation technology to improve productivity and efficiency, decrease labor costs, and minimize human error. The supply chain industry is no different, but logistics process automation presents unique challenges since global networks are notoriously fragmented, with numerous stakeholders involved at every step. Emerging technology in logistics automation provides solutions for international freight forwarders to increase efficiencies by reducing redundancies and eliminating tedious manual data entry tasks.
This article explores the evolution of automation and artificial intelligence (AI) technologies for freight forwarders, challenges and limitations to early versions, and discusses how emerging AI solutions are driving efficiency in supply chain data management.
First wave automation technologies brought exciting ideas for digitizing the paper-driven world of freight forwarding, but ultimately fell a little short of that goal.
Early attempts to digitize unstructured documents and automate logistics data entry utilized optical character recognition (OCR). This technology recognizes text from scanned documents or images and converts that text to an editable file. OCR solutions for detailed, data-heavy documents use programmed templates to read and grab relevant data from an image, then convert the data into a spreadsheet or table. OCR technology was intended to simplify manual data-entry, but did not perform well with complex international freight documents.
The biggest problem with OCR solutions for unstructured document processing is low practical accuracy rates for complex numerical data. While OCR technologies generally boast accuracy levels of 98% or higher, it is important to put this into context. Consider an OCR system that claims 99% accuracy – processing a 10,000 character invoice at a 1% error rate would produce 100 errors. These errors are mostly in minute details, like mistaking a comma for a period or the number 0 for the letter O. These errors are relatively insignificant when reading words or written documents, but when applied to invoices and the transfer of money, these errors quickly become costly and require manual human corrections. OCR may claim 99% actual accuracy, the practical accuracy is substantially lower.
An additional challenge to OCR is that most solutions are template-based. There is not a standard invoice format for freight forwarders and their vendors, so a template has to be created for each individual vendor in order for OCR to read their documents. Most freight forwarders have thousands of vendors in their network, so the amount of manual hours required to build individual OCR templates is inefficient and does not provide cost-savings.
In recent years, artificial intelligence (AI) has been introduced to address problems with OCR templates. This AI is able to learn and understand unstructured documents like bills of lading and vendor invoices, classify them by document type, and map fields to extract relevant data. With the development of AI, multiple companies came to market with software as a solution (SaaS) systems for logistics automation, but sadly most of these also fell short.
The challenge of these AI-driven SaaS systems is that they did not deliver the cost-savings they promised. The AI platforms had to be trained to read documents manually, so staff was still spending time on data management, while also being tasked to learn new programs. Additionally, most of the systems were not designed to integrate easily with existing TMS platforms, so exception management and resolution processes were still handled manually. There was also a large amount of variance in data extraction accuracy with these AI systems, meaning people were still spending time tracking down errors and making corrections. Overall, the AI SaaS solutions failed to create immediate value in labor costs and efficiencies.
How can AI technology be optimized for logistics process automation? The answer lies in strategically putting humans back in the loop.
Technology companies like Uber and Google often outsource their AI training to other vendors rather than assigning it to internal engineers. These vendors are tasked with teaching the company’s AI model to recognize whatever data and patterns it is designed to process, while big tech engineers stay focused on new products and developments. Similarly, freight forwarders are turning to the emerging technology company Expedock to train AI for data extraction and automate supply chain data management workflows.
Expedock lends their proprietary AI to freight forwarders and trains the AI to read unstructured documents and extract data. Their platform integrates with Cargowise, so document data is automatically linked to existing shipments, without any manual labor from the freight forwarder. After training the AI and integrating with existing systems, Expedock employs freight operators who serve as “humans in the loop” to monitor data extraction accuracy and ensure consistent results. The combination of advanced AI and human-in-the-loop operations delivers 99.97% accuracy and labor savings of 30-80%, since removing tedious data management means staff can be reallocated to more productive tasks and projects.
Expedock is the AI-powered solution for logistics automation and creates immediate and exponential cost-savings for freight forwarders. See for yourself - get in touch with their experts today for a consultation and demo.