Separating Fact from Fiction: Real AI in Last Mile Delivery

AI is more than just hype—but current deployments of it cover the entire spectrum from life-changing to wildly unnecessary. On the one hand, two of the 2024 Nobel prize winners (the prizes for Physics and Chemistry) went to researchers who were leveraging AI to tackle complex problems in protein folding and artificial neural networks. On the other hand, you have McDonalds, who had to cancel their deployment of AI-assisted drive-thru ordering last year because the system was riddled with errors. In one instance, it wouldn’t stop adding chicken nuggets to a patron’s order and ultimately tried to saddle them with an order of more than 200 nuggets.

In all likelihood, your technology stack already has some AI-powered capabilities. And in the coming years there’s a good chance there will be more and more AI in whatever technology you adopt. So how do you separate the ideas that are potentially transformative (i.e. smarter data analysis through neural nets) from the ones that are destined to lead to tears—or worse, a pile of unwanted chicken nuggets?

What Is Real AI?

Not all AI is created equal. Just because a developer says its product is “AI-powered,” doesn’t mean that AI is being used to its full potential, or even that it’s doing useful work inside the platform. Like the word “organic” in food, “AI” can mean anything from “the engineers used ChatGPT to do some research” to having a significant amount of AI and machine learning (ML) processes baked in. Hype AI is just adding “AI” to the marketing materials even if the software makes little use of actual AI to improve results.

Real AI for the last mile uses sophisticated algorithms to perform its basic functions, from sorting stops into routes, optimizing fleet capacity, and predicting highly accurate ETAs to behind-the-scenes—but critical—functions such as planning optimal sales territories, analyzing driver behavior, and identifying patterns in performed routes to highlight opportunities for improvement. In doing this, the AI combs through vast amounts of data from every past route performed.

How to Tell Reality from Hype

As any route specialist will tell you, the complexity of managing the last mile can be overwhelming. AI is ideally suited to cutting through that complexity to produce simple, actionable plans, suggestions, and insights.

To do that, the software needs data—lots of data. Collecting, storing, and analyzing data points from every part of the process (i.e. capacity and equipment of every vehicle in your fleet, service time records of each driver, customer preferences, actual arrival times vs ETAs, updated traffic information etc.) is essential.

So the first question you need to ask is how many (and which) factors are used in computing routes and ETAs. Software that simply computes the shortest distance between stops is not going to optimize your fleet or keep your customers satisfied. Some of the important factors include:

• Capacities and equipment for each vehicle in your fleet
• Customer preferences for delivery windows
• Driver service times by job type for each driver and crew
• Evolving traffic conditions
• Shuttle routing to remote DCs or hubs
• Need to reload

Are routing results fixed or can you make manual adjustments without degrading efficiency? Sophisticated AI can take manual adjustment inputs and reconfigure a day’s worth of routes in seconds, not hours. Especially for B2B deliveries (from a food or beverage wholesaler or a building supplies vendor), defining a base route that satisfies customer order patterns and preferred delivery days/times that can then be updated when someone orders extra cases of canned tomatoes or wooden studs—instantly and efficiently—is essential.

Beyond routing, ask what other functions AI supports in the platform? The complexity of the last mile extends far beyond routing, embracing everything from processing orders to creating routes, assigning loads to vehicles to optimize fleet capacity, loading, performing the routes, maintaining communication with customers, delivery/installation services, obtaining proof of delivery and analyzing results. While standard software can do some of these things, real AI does them faster and extends capabilities, such as looking for patterns in results that can be exploited to achieve better optimization.

AI Chatbots vs Intelligent Two-Way Communication

AI chatbots are a novel development in last mile delivery, but are they an improvement or the latest shiny object?

While they may make us feel good, chatbots aren’t necessarily preferred by most consumers. A 2023 research study found that 86% of those surveyed preferred human agents to chatbots, and users reduced purchases by more than 79% when communicating with chatbots on e-commerce sites. Users said they believed that chatbots can’t provide high quality communications, leading to less loyalty to the platforms and increased complaints (source).

The reasons human agents connected via two-way chat are perceived as more capable is obvious: They can assess a customer’s situation (not just their request) and quickly provide a solution if there’s a problem. They may need to reach out to the driver, or connect the driver with the customer in real time, or call in the operations team for a quick fix. They can make a phone call to the consumer and speak with them if necessary.

Of course, there’s still a place for chatbots in the last mile. It’s just a matter of being intentional about how you use them, and of making sure that you’re deploying them in a way that augments real human interactions, rather than trying to replace them. 

It's also a matter of making sure that the underlying technology is really there. You want to make sure you're partnering with someone whose technology can actually make life easier for your customers or associates. 

Right now, the customer communications tools in DispatchTrack's platform are designed to be highly effective in increasing both first attempt delivery rates and customer satisfaction.

At any point in the process, the customer can engage directly with a human. This structure allows customers to set their own delivery windows (while AI ensures your fleet capacity remains optimized), resulting in far fewer not-at-homes. Continuous communication about that delivery gives the customer control over the process while also minimizing where-is-my-order calls. Customers are empowered, delivery rates go up, communication is simple, customer satisfaction increases and—importantly—costs are minimized. 

Simply put, this is the kind of structure that you want to integrate chatbot technology into. You want it to augment a best-in-class customer experience by helping customers get what they need faster and more easily. It's not a replacement for high-touch, automated customer communications. 

Learn how Morris Furniture Company improved their NPS by leveraging DispatchTrack for exactly the kind of experience we've been talking about: 

Delivering the Benefits of AI Now

AI is not just something that promises a better future: The AI and ML capabilities baked into DispatchTrack are already revolutionizing the last mile. By creating algorithms that dig into the hard complexities of delivery and finding workable solutions, DispatchTrack has been building AI reality for years. When you see a promise about what AI can do for you, look beneath the surface and make sure what’s being offered is real AI, not just hype.