How AI can improve last-mile deliveries this holiday season

The holiday season is here, bringing with it joy, cheer, and snow—as well as ice, traffic, and demanding customers. For fleets, add driver safety to the list of seasonal challenges, with the presence of increased driver fatigue due to driving for longer hours and more nighttime driving, said Dr. Suzannah Hicks, AI strategist at Merchants Fleet.

Data insights vs. AI algorithms

Unlike traditional data analytics and insights, AI takes operational efficiencies a step further, and Hicks explains how.

A vehicle can generate hundreds of data points a day. While a human can make decisions based on that data, a human can only analyze small amounts of data and review one report at a time, Hicks said. What’s more, when reviewing reports from telematics data, a fleet manager only reviews what has happened in the past.

“When you bring in AI and machine learning, you bring in a whole other layer, a deeper layer, of analytics,” she explained. “So, instead of just having descriptive or diagnostic [data], which is ‘what happened and when did it happen,’ you start bringing in the concepts of predictive and prescriptive analytics.”

The difference between descriptive and diagnostic data compared to predictive and prescriptive data is like the difference between viewing insights on “what happened and when did it happen” compared to “what could happen and how can I fix it,” Hicks said.

She said AI and machine learning make these predictions using advanced algorithms and analyze massive amounts of data, both historical and in real-time, to optimize workflows as they happen.

These AI models also work better when the longer fleets use them because they compare their predictions with what happened and “learn” from the difference.

Ultimately, Hicks said AI is a force multiplier, helping fleets do things that their managers and route planners can do but that they likely don’t have the time or manpower to complete.

Adding extra packages and deliveries to make can also be a stressor.

“Those vehicles you can depend on, they are going to be packed tight,” Hicks told FleetOwner. “So, you've got all of those issues on board, and ... the drivers are being asked to deliver more parcels in the same amount of time, so that can add a lot of stress.”

Fleets hope to alleviate some of that stress through AI implementations, according to a survey commissioned by Merchants Fleet. The study found that “88% of respondents agree that AI will help alleviate pressure on their overall business as this technology emerges as a valuable ally for busy fleets.”

AI can help accomplish this by optimizing tasks from package scanning to dynamic routing.

Scan and sort packages with AI

Some companies that started with more basic, legacy technology have developed AI algorithms in their products to give their fleet customers a better experience for their drivers. Scandit, which allows drivers to scan barcodes, identification cards, and text using smartphones, uses machine learning and AI algorithms to improve their driver workflows.

One example is what Scandit calls its “MatrixScan” capabilities, popular among last-mile delivery drivers. With MatrixScan, instead of scanning the barcode on each package, drivers can take a single picture of an entire pallet of packages, and Scandit’s augmented reality automatically provides dynamic routing and sorting information for the driver.

This is "changing the way delivery drivers can go and do their job, and then also speeding up things in the back of the van,” Christian Floerkemeier, Scandit co-founder and CTO, told FleetOwner.

This works by providing delivery information as an overlay on the image taken on the driver’s phone. With this technology, a driver can see which packages will be delivered at the start of their route through the end of their route, which packages are in a different zip code, and which packages a driver should leave for drivers of other routes.

Scanning these packages while they’re still on pallets at a hub or distribution center also helps ensure drivers load each package designated for their route, and it helps them ensure they load those packages in the most efficient way.

“Rather than scanning each [package] individually and then maybe detecting some exceptions too late,” Floerkemeier said, “our technology really allows you to confirm that you picked up all the packages and that they are the right packages.”

Floerkemeier also mentioned the “search and find” process that takes place in the back of the van when delivery drivers reach their destination. Some last-mile delivery companies require drivers to spend no more than 20 seconds in the back of a van looking for a package, he said, and if the driver can’t find the package within that timeframe, they’re required to move on.

“Making that process as efficient as possible to go and find the item that you’ll deliver at the current stop—and maybe there's multiple ones that need to be delivered at the current stop—that's one of the key things that we drive,” Floerkemeier said.

AI and dynamic routing

Along with the influx of packages throughout the holidays, last-mile delivery drivers must also contend with winter weather conditions and more traffic. This poses a challenge for routing.

Delivery companies with static routing—the same driver on the same route daily—might not operate most efficiently.

Perhaps one day a driver has a large amount of deliveries on their typical route, causing them to run behind schedule. At the same time, another driver might have very few deliveries on their route. With dynamic routing, the driver with fewer deliveries could aid the driver with multiple deliveries. Yet without having a system to analyze that data and to notify the managers and drivers of the more efficient option, the deliveries will be late, and the driver will likely be fatigued, which could lead to a safety concern.

Further, it isn’t uncommon for delivery fleets to hire extra drivers to combat driver fatigue and to help meet demand, but those extra drivers need direction and route planning, as well. This increase in packages and adding more drivers adds even more variables to the already complex route planning, Floerkemeier said. It’s these scenarios that Scandit’s customers are looking to improve using AI—it's also an area that, once improved, can help the bottom line.

“People are looking at trying to optimize this, and also, from an efficiency perspective, try to squeeze the margins,” Floerkemeier said.

Bringing AI into the mix can even help fleet managers make difficult decisions during emergencies. Hicks used the example of a delivery driver experiencing a breakdown. AI can help the fleet manager determine if there is another delivery driver close by who could take some of those packages and ensure on-time delivery.

With AI, “you can make changes on the fly,” Hicks said. “AI allows for a fleet manager to make those changes as they're happening in real-time, as new things are introduced.”

AI’s role in the 2024 holiday season

While AI and machine learning aren’t new, the trucking industry has only begun scratching the surface of its potential. Hicks predicts that last-mile delivery fleets will primarily use AI for routing optimization this holiday season, but by next year, AI will likely be much more integrated into fleet operations.

“This year was the year that we experimented [with AI],” Hicks explained. “This year was the year that we learned, that we understood, that we did some small-scale things so that we would become comfortable and know what we could do with the data we had and ... the art of the possibility with AI.”

Now, within this holiday season, fleets are going to begin implementing those experiments at scale, Hicks said.

With AI adoption increasing, next year’s holiday season could look very different for last-mile delivery fleets.