©sturti via Canva.com
October 11, 2024
Can Artificial Intelligence Speed Up Last-Mile Delivery?
Innovations in artificial intelligence (AI) are promising to transform last-mile logistics, making it more efficient, predictable, and responsive to consumer needs.
Giant Eagle in early October announced an expanded partnership with Flybuy, an AI-driven location platform. The goal of their collaboration is to cut avoidable delivery driver wait times by 50% to speed delivery to customers and enable drivers to maximize their earning potential, since most gig workers are paid by the order rather than by the hour.
Giant Eagle noted that across the industry, the average wait time for grocery delivery drivers to pick up an order is less than 10 minutes, but unexpected delays can result in reorders and late delivery for the end customer. Flybuy’s technology, which pinpoints which driver is arriving for specific orders and sends alerts to its main dashboard and picking app, has been shown to reduce wait times to less than 2.5 minutes on average.
Heather Feather, senior director of strategy, e-commerce, and innovation at Giant Eagle, said in a statement, “Using Flybuy for delivery has helped lower wait times for our guests, building their confidence that their fresh groceries will be delivered quickly and reliably.”
In an article and video released last year just before Cyber Monday, Amazon cited five ways AI is being used to deliver packages faster.
The benefits cited included:
- Stocking the right products in the right locations: Deep learning and massive datasets are tapped to decide which products to stock in which quantities at which Amazon facility while coordinating inventory shipments from millions of sellers worldwide. In recent years, transformer technology has been employed to “further improve long-term forecasting of what products Amazon should stock on its shelves.”
- Fulfilling orders with mobile industrial robots: AI-enabled robots equipped with machine vision recognize, sort, and inspect the quality of hundreds of millions of diverse goods before they are packaged and loaded on delivery trucks and shipped to customers. Once an order is placed by a customer, one of Amazon’s 750,000 mobile robots at a nearby fulfillment center fetches the item and takes it to an Amazon employee, who scans, labels, and prepares the item for packaging.
- Sorting packages for fast delivery: Once items are packaged and labeled, a robot, “Robin,” helps sort packages before they get loaded in a truck, using AI-enhanced vision to understand what the objects (different-sized boxes, soft packages, and envelopes) are and how they should be optimally lifted and packaged in the truck.
- Expecting the unexpected on the road: With more than 100,000 packages arriving every day at some facilities during the holidays, AI-supported predictive technologies help Amazon “predict the unpredictable,” such as a truck arriving early or weather disruption, to make adjustments.
- Picking the best route: More than 20 machine learning models are used to determine optimal delivery routes.
Scot Hamilton, VP for last-mile delivery routing and planning technology at Amazon, said AI “is meant to be something you don’t see or feel, but it’s as critical as oxygen. When you don’t realize it’s there, that means it’s working perfectly.”
A recent survey of 110 North American transportation, logistics, and supply chain executives from AlixPartners found that 85% of respondents said reducing total cost per order is the top priority for last-mile delivery.
Among the methods shippers are using to improve performance:
- 40% shifted volume away from UPS and FedEx to other providers in the last year.
- 49% increased the spending threshold for free shipping over past 12 months.
- An increasing number of merchants are tightening return policies.
- More executives are utilizing artificial intelligence to optimize performance, although only 7% indicated they’re investing in robotics or delivery equipment like drones and unmanned delivery bots, compared to 16% in 2023.
Discussion Questions
What do you think about the potential of artificial intelligence to speed up last-mile delivery?
Where are the biggest opportunities where AI can reduce costs and drive efficiencies around last-mile delivery?
Poll
Loading …BrainTrust
Ricardo Belmar
Retail Transformation Thought Leader, Advisor, & Strategist
Oliver Guy
Global Industry Architect, Microsoft Retail
Bob Amster
Principal, Retail Technology Group
Recent Discussions
More Discussions
How Can Retailers Crack the Gen Z ‘Riddle’?
Gen Z is becoming an increasingly important consumer demographic, but an elusive one to map. How…Does Bath & Body Works’ Scented Holiday Campaign Pass the Sniff Test?
Bath & Body Works is kicking off a new campaign with scent being at the fore…Which Tech Will Solve In-Store Inventory Woes?
Three major tech lineups are looking to solve retail inventory problems. Which is most suited?
The simple answer is yes. The longer answer is yes, both in the short term and the long term. In the short term AI is already playing a role in improving route planning, ensuring inventory is allocated closer to the shopper to reduce the distance travelled. Amazon’s visual assistance package retrieval is another innovative way of improving efficiency. In the longer term, I firmly believe that a lot more delivery will be made by automated vehicles which will reduce costs and improve efficiency dramatically.
The simple answer is no and the longer term answer is no. In the real world, there is too much traffic, not enough roads and the price of gas, oil and batteries is not going down.
we still have the same basic infrastructure problems we always have. Speeding up delivery by improving our infrastructure is going to take a very long time and a lot of money. AI will do better on all the other miles, just not the last one
It has already happened and it’s still happening. Amazon’s recent reorganization of logistics has allowed for faster delivery across a wider area of the country. And interestingly it’s way more efficient in terms of road usage because AI has helped get products closer to the customer. That’s a good thing for efficiency, cutting emissions, and so forth.
AI will definitely speed up last mile delivery. In fact, a paradigm shift in last-mile logistics has been brought about by the integration of AI, with new improved levels of efficiency, cost savings, and of course, customer satisfaction.
In the final stage of delivery, companies can use AI to optimize routes, forecast demand, and greatly enhance and tighten security. With AI, routes can be systematically adjusted to reduce travel times and fuel consumption, which spells super big savings.
Transparency and security are further enhanced by real-time tracking and monitoring of fraud. Last-mile delivery will continue to grow as AI’s impact on last-mile delivery increases, making it an essential tool.
I think there are some areas of caution as well, including regulatory and safety. It’s also critical to have human oversight and necessary interventions. Relying strictly on algorithms will get us in trouble.
David, you’ve brought up important topics that companies have to solve for the last mile but aren’t usually talked about – transparency and security. They can really create a lot of headaches and I agree minimizing those issues with AI could be a huge benefit!
Yes, AI will help with last mile logistics, and we’re already seeing this across retailers. The Amazon examples cited are great examples of this, starting with the distribution and staging of inventory in the right locations so products are as close to the customer as possible. Clearly this is something Amazon has a built in advantage with their scale, but even regional retailers can benefit from this. Route optimization and selection of carriers for last mile delivery is another area, and I have worked with multiple organizations providing these services to retailers today. I do believe in the examples cited there is a reliance on integration of robotics and AI combined to deliver on the promise of speeding the delivery process overall, but this is a natural pairing of technologies that we will see more retailers adopting as costs come down and more implementations result in additional learnings that can be replicated.
Yes, it is and will continue too. AI is already impacting route optimization, speeding up the pick, pack, and ship in DC’s, is optimizing for micro-fulfillment centers, improving demand forecasting, etc.
If you don’t believe me, you can ask FEDEX, UPS, AMAZON, MIT…
I seriously doubt if there’s an advantage. Last mile is last mile — so inherently complex that there’s a limit on what’s possible. I’m reminded of things like “100% quality” which are impossible yet companies waste far too much time obsessing over what can’t be improved. Besides, if there is any improvement possible it’s likely tiny. On the other hand, there’s a lot of money being spent trying to convince people AI has application where it doesn’t.
What is rarely discussed is how optimisation of deliveries could reduce the overall environmental footprint of deliveries. The fact we can press a button on our phone and have a $3 item delivered to our door the next day – for free – is mind boggling. Few of us consider the environmental and emissions implications but this is where AI has potential.
By predicting where product will be ordered from allows pre-placement of inventory but AI also has the potential to further optimise delivery routes.
Yes, but… aren’t we already here? Every bullet point on the list is already being done, and in most cases, we’ve been doing it for years, even decades. Sear’s home repair services were using route optimization, downstream notification, and load optimization- ie here’s the HVAC van, the plumbing van, etc and managing inventory across drivers ten years ago. AI won’t do anything different; it may enable a bit more efficiency in the delivery process, especially in the routing exceptions due to local conditions. But really, we do that already as well based on real time traffic data.
Sometimes I think we are just repackaging AI as a marketing slogan over what is traditionally called analytics. There are new tech like Gen AI and bigger use of Machine Learning, but from what I am reading a lot of what is done is traditionally called analytics
Bingo! NONE of this is “AI” and none of this is breakthrough news – routing and inventory planning algorithms & analytics have been worked on as soon as there were computers that could handle the data. We were talking about all this when I was in college in the early 1990s. There have been countless recursive learning loops and incremental improvements in the past four decades; getting real-time location and traffic data has been amazing but that’s old news. Slapping an AI label on this work is just stock-pumping lipstick.
At the risk of disagreeing with some of my colleagues on the is (illustrious) panel, I propose that most of the things that can be done to improve last -mile delivery for Bothe the drivers and fo the consumers can be, and will, accomplish through the application of innate, not artificial, intelligence. Having worked with delivery routing software, understanding the capabilities of robotics in a fulfillment center, worked with myriad inventory augment systems, and understanding the importance of being able to collaborate with the customer to insure deliver within a reasonable window of time, I can say with certainty that there is nothing artificial about anything stated here. As my colleague Paula Rosenblum points out, traffic is the biggest problem in urban (and even suburban) delivery and for that, we now have GPS navigation systems to aid us. If anything there is more AI built into navigation software than any of the others I mentioned.
Agree with you. We’re already doing this. The question in my mind – are we going to attribute everything to AI? We need to remember that in 2024 the most popular BI Tool in business is still Excel.
We are in the very early stages of AI use cases, and there will be a lot of trial and error (and mistakes and successes) for at least three years before we will see serious cost reduction opportunities as a result of AI rollouts.
So, no way in the short term. Come back in 2028 and we can talk then.
AI has significant potential to enhance last-mile delivery by optimizing routes, predicting demand, and managing inventory levels. Those that effectively integrate AI will – and already are – gaining a competitive edge, transforming logistics into a more responsive and customer-focused process.
AI is enhancing last-mile delivery, but most companies are still playing catch-up. The potential to cut costs and speed up deliveries is huge, especially in route optimization and managing unexpected delays.
However, I think businesses need to stop overhyping AI without actually investing in it. Sure, Amazon is ahead, but many retailers are barely scratching the surface. The biggest opportunities lie in smarter inventory placement, better demand forecasting, and real-time delivery tracking.
If companies don’t embrace AI quickly, they’ll fall behind in customer satisfaction and operational efficiency.
It has the potential, but comparing Amazon and/or grocery to the rest of retail is like comparing apples and steak.
Most companies don’t have Amazon’s infrastructure or data set to train a model in-house, so they rely on third party applications and models.
Using third party tools for things like route optimization is already happening. Because the companies providing those tools have access to lots of data to train a model. People drive and experience traffic everyday. So are AI enabled tools for demand forecasting. In this area, grocery is way ahead of other retail categories. Why? Because they track so many product attributes (e.g., 150 for a can of beans, 400 for a lettuce) that have lots of predictable value when it comes to demand forecasting. In short, they have a lot of data.
But most retailers outside grocery, don’t have anywhere near this kind of high quality that can be used to get the most value from AI. They don’t track hundreds of attributes per product (new AI tools may help them get there). They don’t capture the number of stock availability checks to orders (which can also have predictive value because it varies a lot outside commoditized essentials) They are just starting the journey of making sure they have the good, clean, expansive data set needed to take full advantage of AI.