Many things can go wrong in the last miles, part of the supply chain, which involves transporting goods from warehouses to consumer homes. The package may end up at the wrong address. Transportation may delay shipments, or thunderstorms may damage parcels left in the rain.
“You're dealing with the human and the real world, with trucks and traffic,” said Fred Cook, co-founder and chief technology officer of last-mile delivery company Veho.
In regions where carriers such as UPS, FedEx and US postal services have long been dominated, Veho and many other software providers are trying to solve the challenges that permeate this infamous, expensive part of their supply chain. Using AI, we design more efficient delivery routes, improve accuracy and customer experience, and predict errors before they occur.
Erik Mattson, a partner in manufacturing and operational practices at consulting firm Alixpartners, believes that “AI is a great opportunity to help this industry catch up with other industries.”
E-commerce sales continue to rise, reaching a new high of $300 billion over the past two quarters. This will make your last mile busier than ever and ripen into the technological mess. According to McKinsey's report, Over the past decade, roughly $80 billion in venture capital has been sent to logistics startups, with the on-demand last-mile distribution platform gaining the largest share of these funds.
Ai from the road to the main entrance
The last mile route usually includes multiple stops and individual small packages, rather than a single truck delivering one truck to a single warehouse. According to the Capgemini Research Institute, last mile delivery accounts for an estimated 41% of all logistics costs in the supply chain.
One of the previous applications of routing technology at the last miles was a machine learning application that was launched in 2013 with a UPS called Orion, or on-load integration optimization and navigation. Four years ago, the parcel company deployed an upgrade to Orion. This reduced the route average of 2-4 miles per driver, rerouteing the driver based on changing conditions.
“The historical technology is static and will be implemented the night before,” Mattson said. If an order is changed or construction begins, the technology will not consider these changes.
Meanwhile, today's AI models adjust in real time.
“Compared to pre-AI methods that rely on static routing rules or dispatcher intuition, our platform now dynamically responds to large real-world conditions,” said Andrew Leone, CEO and co-founder of last-mile delivery platform Dispatch.
Dispatch uses AI to plan routes based on factors such as traffic, delivery windows, estimated time per outage, and driver capacity. A more efficient route can reduce fuel costs, improve density, allow for more delivery in one day, and increase the profits of the provider.
Jett McCandless, founder and CEO of Project44, a supply chain software platform, said Amazon is on the forefront of introducing AI into its final miles. Last month, Amazon announced an initiative called Wellspring. Wellspring uses the generated AI to analyze satellite images, apartment layouts, street images, consumer instructions, and photos from past delivery. We can recommend parking or apartment entrances that drivers use to unload their luggage. In tests this fall, the technology has identified parking lots at 4 million home addresses.
Veho uses AI to assure delivery quality. In an ideal world, employees who are dedicated to quality assurance tasks will look at geocodes where the parcel remains, look at delivery photos, gather feedback from drivers, and determine whether it should be changed for future delivery.
“It's completely ineffective to do that with millions of deliveries, but these are the types of use cases that I think AI is ideal for in a very near-term period,” Cook said.
Additionally, delivery data allows last mile providers to continue to provide information to consumers. Last mile delivery service Derveright has dropped 80% customer service calls due to real-time tracking and more accurate ETA, while customer service calls fell 80%, according to CEO.
Veho said the large-scale language model, created in-house, answered 60% of customer and driver questions, reducing average response times from 2.5 minutes to 15 seconds.
Predicting and preventing package disasters
Vehicles use AI to identify commonalities in accidents that occur during the logistics process, such as the same warehouse associate process that links multiple packages that result in error, or mid-mile trucking companies that have damaged items.
The company predicts the possibility of a particular route or delivery issue. You then make decisions based on the pattern, such as moving your packages to different facilities or increasing your rates on a specific route. That's why drivers are incentivized to pick them up earlier in the day.
“We're now taking that a step further and moving on to where we're trying to predict flaws,” Cook said.
According to a USPS WatchDog report, swiped packages are a major issue with last mile delivery, with 58 million parcels stolen from the gateway last year to a loss of $16 billion.
UPS has created DeliveryDefense, an AI-based software. This analyses historical factors such as loss frequency and delivery attempts. AI discovers areas that could become targets for future porch pirates.
McCandless said that AI can predict the times of high-risk areas and times, allowing businesses to plan delivery schedules and routes to minimize the chances of packages being stolen.
“AI can play an important role in identifying patterns and help prevent theft before it occurs,” McCandless said.
