Supply Chain Scale and Efficiency with AI

In the last five years, the world has seen a marked growth in eCommerce and related logistics since the world experienced COVID in 2020. It’s safe to say that everyone is a lot more comfortable ordering things online and getting them delivered to their home than they were 5 years ago.  

According to data from the US Census Bureau in 2019, e-commerce accounted for 10.6% of all US retail sales - by the end of 2023 that figure stood at 15.4%. That 15.4% represents 1.1 trillion dollars, nearly double the 570 billion figure of 2019. Volumes are up, and with it so are the number of players in the market and the competition among them.  

Technology is enabling this competition as parties in the supply chain have new opportunities to automate, perform better, and operate with more information than ever before. 

Customer expectations are high and driving improvements in speed of service, quality of service, and efficiency across the logistics landscape. Companies who have executed effective logistics operations such as Amazon have created customer expectations for timing and experience that impacts expectations for all parties in the logistics industry. The stakes have been raised, and all businesses must ensure that their supply chains are able to meet them.  

This can be particularly challenging for startups, businesses with complex manufacturing or assembly scenarios, those with perishable products - or those with large fluctuations in customer demand.  So how do companies meet these expectations, while recognizing that not every business has the resources and expertise of a major multinational brand?  

‍

One answer is: Elastic Logistics

Elastic Logistics is essentially the ability to grow or shrink your logistics footprint on demand.  This approach favors using automation over manual labor, rentals, and contractors versus ownership and employees, and in many cases, third party providers.  

The core concept is the ability to scale to the demand that exists at any given time, and to only pay for what you need. This is a concept that has taken a stronghold in the IT industry as companies seek the ability to ramp up or ramp down the amount of IT resources that underpin their applications and websites. Elastic Logistics is implemented in order to meet peak demand when usage is high, but not pay for maintaining that level of service availability when traffic is low.  

‍

On the IT side, it was achieved by migrating hosting and application workloads to the cloud, where third party providers can manage and scale the infrastructure as a managed service. Providers such as AWS, Azure, and Google Cloud Services now perform this function for many businesses to give them seamless resource scaling and more, while offering a pay-as-you-go model to keep costs in check. This allows businesses to meet their customer experience needs for speed and availability even in a very dynamic environment.

In Elastic Logistics it’s much the same. In both cases, some barriers to entry for smaller businesses are reduced by lowering the capital investment required, and making the service an operational expense that is incurred over time and scaled to the need at that time.  

In both the cloud computing and 3PL services arenas this ability to scale, transparency of service and costs, and the related customer experience is 100% enabled by data.  For 3PLs the availability of data regarding inventory, usage patterns, and location of goods and materials continues to improve.  

This is in part due to the wealth of information that is now available from warehouse management systems, scanning devices, and various sensors including GPS that allow greater tracking and transparency of the end-to-end logistics processes.  

Using this data, machine learning and other statistical and analytical tools can be used to drive much more accurate predictions of future demand and other key performance indicators.  End-to-end supply chain performance management is enabled by end-to-end supply chain data visibility.  

Advanced Analytics and Data Science

The supply chain is concerned with all the logistics it takes to meet customer demand from the acquisition of raw materials through to the delivery of the final product to the consumer.  

The core data sets involved may include data from manufacturing, assembly, transportation, inventory, sales, and labor management systems to name a few. It’s a complicated web of requirements and constraints that presents unique optimization challenges to professionals in this industry. The problem is too big to be solved as a whole by any technology, including advanced analytics, so we need to look at some of the specific metrics and areas where data science techniques can be used to unwind and simplify individual challenges.  

Let’s start with predictive analytics, which is using available data to predict future trends and events, often achieved using Machine Learning. 

This approach requires a lot of high quality historical data that describes both the result that you are attempting to predict and the related factors that may contribute to that result, referred to as features. Data Scientists create machine learning models that use that data to attempt to discover patterns and relationships between the features and the actual results and through this, learn how to accurately predict them. 

A prime target for this type of analysis is sales demand, which if accurately forecasted in detail, tells a company an incredible amount about what the end targets should look like for the supply chain to deliver.  Other common areas for predictive analytics in the supply chain include performance metrics such as Turnaround Time, Inventory Accuracy, Inventory Stocking Requirements, Equipment Maintenance Events, Bottleneck and Disruption Prediction, and Backorder Risk.  Using advanced analytics against these use cases, companies are able to take both internal and external data to more accurately predict and respond to future events.  

The area of AI perhaps getting the most attention right now is Large Language Models, or LLMs, with ChatGPT from OpenAI and Gemini from Google being the most well known two.  

These are pre-trained deep learning models that are trained on top of massive text data sets.  These models are used for Generative AI, which allows the model to produce original written content based on text prompts which it can process and understand. These models have greatly improved the accessibility of AI as the interactive prompting processes and the delivery of results can occur in a very natural and conversational way. 

Due to this flexibility there are a lot of use cases emerging. In 3PL, this technology is currently being used for things like Voice Picking, where warehouse associates can receive information about order status, inventory levels or locations using only their voice. On the technical side, software Developers are able to use code companions built on generative AI that write code based on requirements, or that can assess and QA written code to ensure that it is efficient and hasn’t drifted from the requirements.  Additionally customer service groups are able to use this technology to reference documentation and prepare professional written responses to inbound tickets.  

The customer service use case speaks to the idea of training an LLM against internal company information.  An LLM with access to your company’s policies and procedures, fulfillment history, customer service ticketing, customer SLAs, and emails for example, would be a very powerful way to reference and give access to the whole picture of how your organization, including its supply chain, is operating.  

This would give you the ability to ask and answer questions that improve the findability of information, and the AI could draw insights from this source material that deliver immense value to the business. 

MyContentScout

The Virtual Forge has created a tool called MyContentScout, to do something similar. MyContentScout gives AI permissioned access to company files and information including text documents, audio, image, video files, dashboards, e-learning courses, and more.  

‍

This is targeted at creating an internal company search interface that delivers results from your own document stores based on the meaning of what you are searching. The search works across media types, and languages, and delivers convenient summarization as well as links to source materials.  

With the proliferation of digital assets at all companies these days, finding things within that sprawl can be challenging, so AI is helping to ease that pain point by providing a very natural way to search and find content.

Use Cases 

Product Supply Optimization

One instance where The Virtual Forge was able to deliver hard savings back to a client was in the medical device segment. We were able to do this using existing data to optimize and automate a supply chain process. 

This particular client had a large field salesforce with several hundred members that they needed to keep adequately stocked with product samples. This was originally done manually as a monthly process where all resupplies were done as a batch.

The Virtual Forge worked with this team to improve inventory availability and product mix for the salespeople, by automating the allocation and distribution functions. 

This was a technical solution that leveraged supply and demand signals from both the field sales force (demand) and the warehouse (supply) and leveraged an algorithm that used that information to smartly generate resupply orders for individuals, and determine the optimal ship timing.  

Proactively replenishing inventories facilitated dramatic cost savings in shipping due to the reduction of rush orders, saving $1 million in carrier costs within the first year.  Sales representatives and management stakeholders also benefited from the change as their time spent managing the process was reduced, and the availability of inventory was much improved.

‍

Parts Inventory Exchange

The Virtual Forge innovated for a client in the supply chain space, a large distributor of automotive parts. Their business had a segment where they would purchase existing parts sellers that operated as an integrated part of a larger business. Consider, for instance, a parts counter catering to the needs of a bus fleet or a municipal parts depot.

The issues that they faced were excess inventory and dead inventory - large inventories of parts that sold too slowly or might not sell at all. Everyone knows that large stores of unsaleable inventory is highly inefficient to carry,  and this came about frequently because each of the sites where they operated using this business model was standalone; so if an individual client no longer needed to buy a certain type of part, there was no means to monetize that inventory.  

To resolve this, The Virtual Forge designed and developed a centralized inventory exchange.

The inventory exchange utilized a customized algorithm that searched for the most cost efficient inventory exchanges between sites using inventory, geographic location, inventory consumption, and other data points.  We then created a user interface for the counterpeople to be able to see their most beneficial inventory opportunities. Sites with excess could find ranked best fit sites who may like to purchase that inventory, and alerts when those sites were likely nearing a buying cycle. Sites with shortages were prompted to request a transfer from sites with excess of their required products versus purchasing the product elsewhere. The user interface also managed the site-to-site transfer process transactions, as that feature was not included in the inventory application used by the sites.  

The net impact to the business was millions of dollars of savings as their excess parts inventories were able to be purchased and used rather than disposed of or sold externally for pennies on the dollar.  Additionally, the carrying inventory efficiency of all of the sites in the network was improved as they gained both an outlet for excess inventory, and a cost-effective means of sourcing critical parts inventories.  

‍

Debit Memo OCR

One of the workstreams we support for a major client at The Virtual Forge is reverse logistics. Part of this process requires managing Debit Memos, which are documents that list details about products that a customer is requesting to return for credit. These documents contain critical details that must be faithfully entered into our returns management system for processing, but typically arrive in PDF format, and with a lot of variation in layout as they come from various returning parties.  Because of this, our data entry team was required to spend a lot of time manually extracting information from these PDFs. It was an ideal scenario to apply Robotic Process Automation (or RPA) to automate and streamline the process..  

To achieve this task, we built an Optical Character Recognition (OCR) pipeline in Azure. OCR is an area of Artificial Intelligence, Pattern Recognition, and Computer vision focused on the extraction of text from images containing text, either printed or handwritten.

Using this process we created a Robotic Processing Automation application to ingest, identify, process, and output the most common types of debit memo forms our team receives.  We have increased the accuracy of the data entered into our system, as well as achieved a greater than 50% decrease in manual data entry efforts. In addition to the internal benefits, this benefits our customers as it has allowed us to increase speed of the returns process, and enhanced our ability to scale when additional returns activity is needed.

Looking Ahead

The landscape of logistics and supply chain management is evolving rapidly, driven by technological advancements and the need for businesses to adapt to changing consumer expectations. 

Concepts like Elastic Logistics, fueled by data-driven insights and automation, are reshaping how companies manage their supply chains. Moving forward, the integration of advanced analytics, AI, and innovative solutions will continue to shape the industry, enabling businesses to optimize their operations, enhance customer experiences, and achieve greater efficiency. Embracing these developments will be crucial for companies seeking to thrive in the dynamic and competitive world of modern commerce. 

With innovative solutions from The Virtual Forge, businesses are empowered to navigate industry-specific complexities, optimize their resources, and ensure the needs of their consumers are met. In the ever changing landscape of logistics and supply chain, we help businesses to apply technology that helps them compete and win.

How Can The Virtual Forge Help

At The Virtual Forge, we aim to build data-rich, intelligent platforms that solve real-world problems. We use a collaborative approach that lets us know what our clients need so that we can provide efficient and unique solutions for their business problems.

If you are looking to understand the intricacies of when and how to effectively apply AI & ML technologies in your business, learn more about our Artificial Intelligence Services here. 

If you want to learn how our technology solutions can optimize your supply chain management, explore our Custom Manufacturing Software Development Services.