AI in Last Mile: What it Takes for Technology to Reduce Risk

Editor's Note: This article originally appeared in the Winter issue of Customized Logistics & Delivery Magazine, a trade magazine published by the Customized Logistics and Delivery Association. Reprinted with permission.

As the ever-expanding capabilities of artificial intelligence (AI) continue to find more applications in the last mile delivery sector, it is giving providers new tools to help them operate and grow their companies more efficiently. There is no doubt that applications of AI in the last mile delivery sector should help companies better solve the issues of traffic congestion, routing, predictive analytics, and communication in the delivery process. But does integration and reliance on AI make a last mile delivery company safer? Will it translate to less incidents and claims? The answers to these questions remain to be seen, but as we have seen with prior applications of technology in the last mile delivery sector, it is usually a double-edged sword with respect to risk reduction and claims.

When telematics were first introduced in the early 2010s, the prevailing opinion was that dash cams would make for safer drivers and fewer claims. Dash cam providers promised customers that they’d improve performance and reliability and reap the benefits of savings on their auto insurance. While some transportation companies did realize savings, most did not. The stark reality was that for every instance where dash cam video provided proof that a driver was not at fault, it also provided proof, without a shadow of a doubt, that a driver was at fault.

As dash cams began offering additional AI-powered features, such as driver performance tracking, the thought was that driver coaching would help correct poor driving behaviors before an incident occurred. The reality was that most transportation companies did not apply the time and effort into coaching. Therefore, the incidents didn’t stop occurring. The failure to apply the coaching when the data existed only spelled more doom in the courtroom for transportation companies and insurers.

When Electronic Logging Devices (ELDs) were introduced, most in the last-mile industry believed that these devices would allow for greater oversight into driver hours of service and drastically reduce the number of drivers on the road operating vehicles while fatigued or tired. Even as the Federal Motor Carrier Safety Administration (FMCSA) mandate on ELDs took hold, requiring the use of these devices, we did not see a dramatic decrease in claims activity across the entire transportation sector. Once again, the failure to properly monitor the application of technology left many transportation companies at a disadvantage when defending themselves in a claim situation.

Examples like the two above are everywhere within the sector through the years. Technology has a funny way of making our lives easier and harder at the same time. What it provides in efficiency by streamlining data, it complicates by overwhelming us with data to consider. As we look at how AI can help improve operations in the last mile, we must also recognize the potential pitfalls it presents us in regard to risk.

Take for instance route optimization and dynamic routing — AI can and has already helped improve routing. It allows you to maximize efficiency while minimizing time on the road, adjusting routes in real-time based on congestion or weather factors. These are undoubtedly big factors in reducing the likelihood of an incident of auto and cargo claims.

Drivers have come to enjoy the AI-enabled routing software and use it daily. But what happens when the seasoned driver decides to disregard the routing software and take their own route, only to have that result in an accident where they are at fault? Your drivers have all operated without incident by utilizing AI, but an accident occurred when this driver did not. In this instance, the liability impact of this driver going off script and having an accident will not be lessened by AI, but instead will be exacerbated by AI. Even if this driver is an independent contractor and you can argue that you are not responsible for their actions, recent claims trends suggest that your company will still be held vicariously liable for the accident. What if this driver has a history of going off script and not listening to the AI routing? Have you failed to reinforce the need to follow the suggested route for safety reasons? Have you documented those discussions? Are you potentially complicating labor law issues by mandating the routing software be utilized and followed? Any last mile delivery company using AI routing software must thoroughly consider the risks, heavily document protocols, and act immediately when drivers veer from suggested routes.

One of the most powerful capabilities of AI is its ability to perform predictive analytics and give us a projection of future needs based on past data. In the last mile delivery sector, it becomes an incredibly valuable tool to assess the “peak season.” As you look to AI to review past peak seasons, or any defined period, the tools will predict what the seasons may look like in the future. Bear in mind that using predictive analytics does not come without risk. Many last mile providers approach the booms in business by securing additional drivers and/or short-term rental vehicles to bolster their ability to handle the increases.

These short-term solutions come with their own set of risks such as lack of familiarity with a driver or putting a driver in a rental vehicle that is larger than they are used to driving. This increases the risk to the last mile delivery company to meet the clients’ needs. In some cases, we have seen these peak times and operational changes completely overwhelm last mile delivery businesses with unintended incidents and claims given the lack of control. Growing too much too quickly leads to a lack of proper controls. Approach cautiously, and integrate measures that mitigate the risk whenever possible, including thorough training and vetting of drivers.

Predictive maintenance is one of the best examples and applications of AI in the last mile delivery sector. Utilizing performance data from telematics, it gives you the ability to see the maintenance needs of your fleet vehicles. This allows you to plan for preventative maintenance to keep vehicles mechanically safe and efficient. But what happens when the maintenance timeline doesn’t jive with the operational timeline? What happens when we can’t afford to pull a truck off the road this week to maintain it because we can’t get a replacement vehicle from the leasing company? You can imagine your increased risk and liability in this delayed maintenance scenario. Should you have a claim or incident, and the data is called into question, it will be difficult to argue why you aren’t liable when AI told you that the vehicle needed maintenance. When adopting predictive maintenance tools, be prepared to follow through on maintenance recommendations or risk paying the price in the courtroom.

As a consumer, the AI tools powering communication and transparent delivery expectations are my favorite. Automated updates and projections of when items will be delivered enhance the buying experience. As AI communication grows, delivery expectations could increase risk. Take for instance lab work, medical specimens, or critical care medicine/equipment. The healthcare industry expects that, based on communication and timelines, items will be delivered within a set time. What happens if that timeline cannot be kept, and a life hangs in the balance? While contractual liability with your customer is important, it will not keep the deceased's family or estate from suing you if your delivery does not make it in time despite what was communicated. It isn’t that you mismanaged the AI, but the AI has set an expectation that must now be met. I’m not suggesting that you completely disregard safety to hit established delivery timelines. Still, it is important to recognize that either way, you are potentially increasing your liability exposure while attempting to provide a more positive customer service experience for your clients.

There is no arguing that the growth and application of AI in the last mile delivery sector can help to improve many facets of your operation. However, we also must realize that the human factor still looms large in all organizations. The actions, or inaction, of the people in your organization, including you, will determine whether AI is successful in enhancing your operation and reducing risk or making it riskier.