In its never-ending quest to get products to customers more quickly, Amazon has deployed three AI-powered innovations designed to speed up and streamline fulfillment and delivery processes:
- Wellspring, a generative AI mapping solution that draws data from sources including building footprints, customer instructions, information from prior deliveries and satellite imagery to create more comprehensive delivery solutions;
- A foundational AI forecasting model that adds time-bound data like weather patterns and holiday schedules to sales history data, to more accurately place the right products in the right locations; and
- Agentic AI to facilitate the next stop in Amazon’s robotics evolution, which will add the ability for warehouse robots to hear, understand natural language, apply reason and act autonomously.
Mapping 2.8 Million Apartment Addresses to the Right Buildings
Amazon began testing the Wellspring solution in October 2024, and the system has been able to map more than 2.8 million apartment addresses to their corresponding buildings across more than 14,000 complexes, while also identifying convenient parking locations at 4 million addresses.
The AI-powered mapping technology also detects building entrances and mailroom locations by analyzing proof-of-delivery photos and location data from past deliveries, helping drivers navigate unique environments — like multi-building apartment complexes and brand-new neighborhoods that haven’t yet appeared on navigation apps — with greater confidence.
Amazon also has been working to improve delivery in less populated areas. Last month the company announced it would invest $4 billion to enhance rural delivery services, following the September 2024 news that the retailer would invest $2.1 billion in its Delivery Service Partner (DSP) program.
Advertisement
Demand Forecasting Tool Boosts Long-Term National Forecasts for Deal Events 10%
Amazon has made a number of investments in regionalized demand forecasting in recent years. In May 2023 Amazon reported on its success in regionalizing where inventory is located, with the result that 76% of the products Amazon customers order are delivered from within their region.
Now, AI is making it possible for Amazon’s systems to incorporate time-bound data like weather patterns and holiday schedules, along with sales history, to guide inventory planning decisions. Analysis of regional differences, such as sunscreen sales in Cape Cod, Mass. in summer or ski goggles in Boulder, Colo. during peak ski season, have contributed to a 10% improvement in long-term national forecasts for deal events and a 20% improvement in regional forecasts for millions of popular items.
The results include faster arrival of packages (sometimes same-day rather than two days) and fewer miles traveled by delivery partners, reducing both traffic and carbon emissions. Operations networks in the U.S., Canada, Mexico and Brazil already are using the technology, with additional expansions coming soon.
Telling Robots What to do Using Natural Language
An agentic AI development team within Amazon Robotics is building an AI framework to facilitate adding the ability for robots to hear and understand natural language, reason about it and act autonomously. The goal is for operators to communicate directly with robots in fulfillment centers, with verbal directives such as “Pick all items in the yellow tote to your left and place them in the gray tote,” or “Load the trailer with all totes in the loading area.” Vision Language Models (VLMs) and policies that drive robotic actions would allow these instructions to be issued in plain speak.
The goal is to transform autonomous robot systems such as Amazon’s Proteus into versatile assistants that can move heavy objects in tight spaces, for example, while freeing up humans for problem-solving that requires critical thinking. Benefits will include safer jobs for frontline employees; faster delivery, as robots are rerouted to where they’re needed most; and greater efficiency, with one robot capable of performing multiple tasks.
