March 2025

What technology advances are having the most significant impact on carriers delivering the last mile?  How are route optimization, real-time tracking, and AI impacting this segment of the supply chain today?  And what about tomorrow?

“Thinking Outside the Box” talked to two of the industry’s thought leaders, Nitin Gupta, CEO of Beans.ai, and Arthur Axelrad, CEO and Co-Founder of Dispatch Science, for their take.  Here’s what they told us.

QUESTION: How have recent advances in route optimization helped last-mile providers?

Gupta:  The most significant advancement in the last decade is that the data you can put into your route optimization systems has grown exponentially.  Today, you can track your drivers and your vehicles in real-time. You can also track a lot more in real-time, from exact location breadcrumbs to the temperature inside your vehicles to delivery time windows, delivery policies, and live picture proof of deliveries. That data flows into your route optimization system, helping inform better decisions. In addition to the increasing labor and fuel costs, courier companies have also become much more cognizant of what their contracts with shippers should look like.

Route optimization helps carriers adapt as the flow of deliveries changes.  Planning the day for a driver has always been a tedious process.  Carriers would try to optimize the routes in the morning based on what they knew.  But what would happen when an extra ten deliveries came in?  In the past, trying to fit them in was manual work.  Today, all that is automated.  It happens within a few seconds, so the driver gets on the road quicker.

These systems also allow you to add extra stops, respond if a driver is running behind, or take action if a vehicle breaks down. Today’s route optimization systems can handle all those things, which wasn’t possible five years ago. This means that these systems have made operations much more efficient.

Route optimization has also significantly impacted carriers’ ability to make the most of drivers’ time. Much of what happens in the final-mile is manual work, and the cost is very high as a result. Let’s take California as an example.  The minimum wage there is $18 an hour. That’s $0.30 a minute.  So, if you’re wasting a minute of a driver’s time, that impacts a carrier’s margin for that delivery. So, route optimization can be the difference between running profitably and not. Companies like Beans.ai run powerful optimization algorithms to squeeze the maximum benefit out of their day.

Axelrad: Route optimization has evolved extensively in the past few years. In its early days, it was much more like an independent, stand-alone tool. Today, it’s tightly integrated and embedded into modern transportation management systems.

Nitin mentioned that there’s a ton of prep of planning and manual steps that used to bog down dispatchers and staff. That’s been drastically reduced or even eliminated by today’s systems. Route optimization is at the heart of modern cloud-based transportation management systems like ours. It’s a must-have for today’s couriers.

For courier companies, the benefits of route optimization, like Dispatch Science’s, go way past putting stops in the best sequence. It can take all of the orders and tell the carrier how many drivers they’ll need to win the day.

It can then tell when and where drivers should start and what time they should end. It can also account for specific pickup and delivery windows. If a driver breaks down or has to add a last-minute pickup, it can adjust accordingly.

It’s perpetual. It adjusts, reacts, and moves.  It’s not a static one-time event. For courier companies, this means no more jumping through hoops to plan or worrying about formatting data manually. They don’t have to upload information to a third party, download it back, and inject it into their system.

Carriers no longer have to ignore the fact that the system doesn’t know about specific industry concepts like service levels, pickup and delivery windows, or parcel-specific information. Today’s systems tie all that together.

QUESTION: Talk more about real-time tracking

Axelrad: It’s all about efficiency. Nitin talked about real-time tracking.  I want to build on that and discuss the impact on the last-mile. Recent advances mean that it’s about much more than just seeing the exact position of a shipment. We’re using geofencing to automatically send an alert when a driver arrives in a preset zone. This could allow a warehouse manager to know ahead of time that they need to clear the dock because the driver is about to arrive. It lets a recipient know that the driver is just minutes away or has arrived. It gives a shipper total visibility on the movement of their goods. Think about that impact on, for example, the life science delivery space.  Real-time tracking could actually save lives here for those awaiting an organ transplant or life-saving medications. Knowing exactly when these deliveries will arrive has a dramatic impact. 

The most critical decisions in the world are based on real-time tracking information. Recent advancements in real-time tracking have revolutionized the last-mile by allowing for unprecedented levels of transparency. And that’s what everybody’s come to expect these days. It’s not just about receiving that shipment. It’s about tracking where it’s been and knowing when it will arrive.

QUESTION: What is the impact of route optimization on pricing jobs?

Gupta: Delivery companies can use this technology to price jobs more intelligently.  Better data is the key to answering RFPs or negotiating a contract with shippers. The system can consider more than just where the deliveries are going.  It can consider the load size so the carrier knows how many drivers they’ll need to handle them. Our route optimization system looks at historical data and 75+ parameters to determine how many drivers are required and how much time they will spend on the road. That gives courier companies an estimate of what they can pay the drivers. Those calculations can consider fuel costs or the time and cost of charging if they use EVs.  Then, the courier company can go to shippers and say, “This is my pricing based on this information.” These couriers can take the output from a model like ours and give it to the shippers, saying, “Here is the exact calculation of how we got to these numbers.” It lends transparency to the entire ecosystem, from the shipper to the carrier to the driver.

Carriers who use these systems have an advantage over their competitors.

 QUESTION: Let’s talk about AI and machine learning.  How do recent advances here help carriers improve their jobs?

Axelrad: Today, carriers are using AI to analyze partial photos to detect if there’s damage to a delivery. In this scenario, AI can automatically alert staff and streamline the remediating process for that problem.

AI can also transform a handwritten label to make it more legible or enable virtual assistants to take phone orders. AI and machine learning are already helping to improve dispatch decisions. AI can learn by using past data, including data on actual load times and delivery coordinates. That information can be corrected in the real world and used to improve plans for delivery. 

Gupta: AI has become a wide field in the last three years. Traditional AI has existed for about three decades. It was called machine learning, not AI. Around 2010, it started being called AI, and now it’s called generative AI, which is one facet of AI right now.

Today’s generative AI uses machine learning systems to make the system better.  Machines can run the numbers but don’t necessarily arrive at optimal outcomes.  For example, the system might tell a driver to deliver in a particular order.  But the driver says, “I know better. I’m not going to follow that sequence.  My experience tells me there’s a better way.”  We’ve been humbled when drivers are more efficient by not following the model’s suggestions.  Mathematically, those models are correct, but on the ground, they may not be.  Drivers have such local knowledge that they can move better. Machines don’t.  A driver may know there’s a broken gate or a temporary detour because of construction.  The machine doesn’t know that. Machine learning has allowed us to consider the driver’s past behaviors, combine them, and mirror what’s happening on the ground.  The machine will make minor tweaks based on real-world experience to make the driver move more efficiently. It improves over time by taking into account drivers’ intuition.  That’s a big part of AI today.

QUESTION: How does machine learning help with missed or incorrect deliveries?

Gupta: One of the biggest problems courier companies face is drivers dropping packages in the wrong place or in front of the wrong door. Traditional solutions have included geofencing, where the driver must be within 100 feet of the delivery location. But today, that’s not good enough.

At Beans.ai, we started with apartments. Everything is within 100 feet.  So, drivers could drop things in front of the wrong door without the system being aware. We knew there had to be a better way, and machine learning helped us there.  With machine learning, we can ensure that the driver puts it at the right door because it learns everything about that front door and the package.  We lean heavily on the use of photos.  We ask our Machine Learning models, “Do you see a package in that photo?” “Do you see the right barcode on that package?” “Do you see a door?” “Do you see a number on the door?” “Does that number match the input data?”  And our model gets trained on these questions with every picture that comes in. Now, after having seen two billion plus pictures, when the driver makes a delivery, our trained model can verify whether they were at the right door. 

QUESTION: What’s next for AI for those in the last-mile?

Axelrad: AI agents. Think of an agent as a layer of technology that sits on top of what’s already out there. It could sit on top of your TMS, add current weather or traffic information into the mix in real time, and automatically take action. So, when a weather event happens, that AI agent could automatically connect to the weather bureau, get that information, connect to the TMS, and redirect a load or an entire route.  This is not that far off.  I expect it will happen by this time next year. Companies like Dispatch Science will be starting to introduce prototypes of this.

As AI advances, carriers will do more without adding more people. The reality is that the tools that are coming will enable courier companies to maximize the use of their personnel. Some tasks will be completely automated, allowing companies to reassign their people to different functions. For example, a company may not need as many customer service reps to take telephone orders because AI will make it possible for a customer to have a conversation with a human-sounding voice.  They will tell the AI agent what they want to do, and it will take that order.  That will allow the staff to focus on more complex and specialized tasks.  Technology will handle anything repetitive, allowing humans to do what they do best. 

Gupta: Yes, I foresee these tools creating better efficiency for carriers because systems will perform more complex tasks quicker. There will likely be huge leaps, especially when it comes to route optimization and planning. Optimizations will happen quicker.  The processing units already coming into use can process millions of pieces of data, looking at all the possible solutions in seconds rather than minutes.  That has enormous implications for cost. It’s going to make operations way more efficient for carriers.  Couriers that start using machines the right way will survive in the market.  They will have to because their competitors will be using these systems, and they will be super-efficient. Carriers will be performing much better in terms of efficiency. Ultimately, their margins will be better.

We’ve seen many logistics companies close down in the last three years. This is primarily due to carriers’ inability to meet their sales goals. They’re losing contracts faster than they can get them. Those who use technology effectively will retain the contracts for much longer because their service level requirements will be higher than the clients’.

Anyone who doesn’t use the right kind of AI and load optimization cannot survive in the market. It’s like bringing a knife to a gunfight.  

Axelrad: Carriers have to be prepared to make the most of AI.  They need to have a compatible framework. They need to have software that is modern enough to take advantage of AI. It must be cloud-based software with APIs and the type of connectivity and interconnectivity that AI needs to be efficient.

QUESTION: How does a carrier keep up with all these changes?

One way to keep up is to avoid falling too far behind. It’s much easier to take continuous little steps than to let yourself fall too far back and then have a big wall to climb. I would use the word “iterate.”  Make sure you’re checking and making technological improvements every month, every year, and moving forward. It’s like your house. If you maintain your home and do the maintenance and some renovations as you go, you won’t end up with a fossil.

It’s easier to keep up if you never get too far behind. It’s important to continuously look for the little things and the little steps that you can take to stay modern as a business. This means paying attention to what’s out there every month or every year and implementing little bits of technology as you move forward.

One of the best ways to keep up is to get involved with organizations like CLDA.  That’s how you find out what’s new and get access to thought leadership in the business. It’s where companies can exchange information with colleagues and vendors.  That’s how you stay current and find out how to make the best of the technological advances in our industry.