How to Optimize Last-Mile Delivery in the Age of AI

By Colin Bonini

In the age of fast-paced e-commerce, customer expectations for product selection, convenience, and delivery speed have reached new heights. From next-day and same-day to within-hours deliveries, meeting these diverse demands requires intricate coordination and optimization in all supply chain areas, especially in the last mile. According to Forbes, last-mile delivery makes up 53% of total shipping costs, and 84% of customers are likely to abandon a company after one bad last-mile experience, making it the most costly and crucial step in the delivery process.

Hrishikesh Paranjape is a product management expert who has optimized and deployed tech-driven routing algorithms and spearheaded the development of industry-changing technologies used by hundreds of thousands of drivers daily. With extensive practical experience and degrees from the Indian Institute of Technology, the University of Oxford, and the MIT Sloan School of Management, Paranjape knows what it takes to orchestrate modern last-mile logistics. His efforts help companies fulfill their promises to end users and prioritize driver safety, customer satisfaction, sustainability, and cost efficiency.

Hrishikesh Paranjape

In this Q&A, Paranjape discusses the recent advances in last-mile delivery and the role AI and other technologies are playing in optimizing deliveries.

Related:AI Quiz 2024: Test Your AI Knowledge

Q: What are the recent advancements in the last-mile delivery space, especially in the areas of safety, accessibility, and sustainability?

Paranjape: Today's delivery applications and underlying routing technologies make it easier for drivers to deliver packages safely. For example, new experiences in driver applications alert delivery drivers when they transition from a low-speed to a high-speed road, or when routes require left- or U-turns on potentially busy roads. Moreover, these apps can consolidate historical data to let drivers know if a delivery destination is associated with a previous safety incident, like getting stuck in a driveway, blocking a narrow street, or parking issues.

Beyond safety, these experiences allow last-mile delivery to extend into complex, hard-to-reach areas like high-density urban zones and remote rural communities. In urban areas that are potentially difficult to navigate, some companies have integrated a hub-and-spoke strategy that leverages a van filled with parcels close to high-density areas. Parcels are distributed from the van to workers who walk separate routes to reach delivery destinations. This approach increases flexibility for delivery workers, reduces carbon emissions, and increases end-user satisfaction.

Related:AI Assistants: Picking the Right Copilot

The delivery industry is experiencing a monumental shift to prioritize safety, sustainability, and driver accessibility in addition to optimization and customer satisfaction objectives. As a result, last-mile delivery companies are developing and implementing new delivery experiences powered by advanced algorithms and optimization techniques that are now used by hundreds of thousands of delivery drivers in the United States daily. These systems help ensure safe, consistent, and efficient execution of last-mile delivery.

Q: How do generative artificial intelligence (GenAI), machine learning (ML), and other technologies optimize and advance last-mile delivery?

Paranjape: Technology is at the heart of all advancements in last-mile delivery. For instance, a typical map application gives the longitude and latitude of a building — its location — and a central access point. That isn't enough data when it comes to deliveries. In addition to how much time it takes to drive or walk from point A to point B, it's also essential for a driver to understand what to do at point B. At an apartment complex, for example, they need to know what units are in each building and on which level, whether to use a front, back, or side entrance, how to navigate restricted or gated areas, and how to access parking and loading docks or package lockers.

Before GenAI, third-party vendors usually acquired this data, sold it to companies, and applied it to map applications and routing algorithms to provide delivery estimates and instructions. Now, companies can use GenAI in-house to optimize routes and create solutions to delivery obstacles.

Suppose the data surrounding an apartment complex is ambiguous or unclear. For instance, there may be conflicting delivery instructions — one transporter used a drop-off area, and another used a front door. Or perhaps one customer was satisfied with their delivery, but another parcel delivered to the same location was damaged or stolen. GenAI can re-create the building's structure to fill the gaps not recorded in the original data. By leveraging GPS traces from successful deliveries and triangulating that information with other high-fidelity datasets like publicly available photos, satellite imagery, and third-party datasets about parking conditions, GenAI-powered apps can notify drivers of alternate building entrances, drop-off spots, and more intuitive, hyperlocal delivery routes.

With the specialized data generated by AI and ML, last-mile routing becomes quicker, the sequence of delivery stops improves, and the overall delivery experience becomes more intuitive. These technologies reduce instances where drivers are asked to expend extra energy and put themselves in unsafe or confusing circumstances, saving drivers and companies valuable time and resources.

Q: What are some other ways routing algorithms, last-mile delivery, and large-scale distributed systems can be enhanced?

Paranjape: One intuitive improvement is the comingling of different products in last-mile deliveries. Most unique products — groceries, pharmaceuticals, cold storage, and same- or next-day — are delivered separately. Combining deliveries would decrease last-mile complexity for the transporter and allow companies to optimize delivery costs while increasing delivery speed.

Sustainability is another area in which the delivery sector will improve. Many companies are investing in electric vehicles (EVs) for their last-mile fleets, and multimodal delivery methods like walking, e-bikes, and hub-and-spoke strategies will further optimize delivery times and costs while reducing the fleet's overall carbon footprint.

Q: How can companies leverage technology to benefit delivery drivers and end-use customers while growing their business in a cost-effective manner?

Paranjape: Product managers and business leaders have a primary responsibility to understand their users' needs and pain points profoundly. Working backward from a clear user need and exploring new and emerging technologies are crucial for solving real-world challenges. With better access to powerful learning models and deployment techniques, it's easy to innovate in the wrong direction — either over-complicating processes and systems or addressing issues that don't directly solve users' immediate needs. For these reasons, it's vital for product managers to prioritize solving unmet customer needs, regardless of whether this requires a new technology or legacy technique. By working backward from real-world problems based on drivers' last-mile experience, new products powered by GenAI are more likely to succeed in the market.  

With the rapid growth of e-commerce, the ability to deliver products quickly and effectively is foundational to business success. To accomplish this, it's imperative for companies to consider societal expectations around sustainability, ethical work practices, driver safety, and the capabilities of AI and ML. By leveraging innovative technologies to prioritize safety and sustainability, organizations can make last-mile delivery processes more intuitive and cost-efficient, running advanced optimization and GenAI on large-scale distributed systems.

When business leaders and product managers look beyond cost optimization and advocate for meaningful change and technology architecture improvements, it benefits end customers, drivers, and communities in which they operate.

About the Author:

Colin Bonini is a writer and college instructor from San Jose, California. In addition to creative projects, he also highlights innovative, goal-oriented research from leaders and professionals across various sectors. Connect with him at cpjude.com.