Amazon.com Inc.’s stealthy Lab126 organization has formed a new group that’s focused on artificial intelligence agents, the company said today.

Lab126 is a research and development organization within the company that’s known for its great secrecy. It has previously helped to develop devices such as the Amazon Kindle e-reader and Echo smart speaker, and it has a reputation for keeping many of its innovations under wraps until it’s ready to launch them as products.

The formation of the new group comes at a time when a growing number of technology firms are exploring the potential of AI agents. They’re looking to leverage advances in AI to go beyond simple text and image generators, and create digital assistants that can perform complex, multistep tasks and actions on their user’s behalf.

The new group at Lab126 will mostly be looking at how AI agents can enhance the company’s robotics. Amazon already employs a wide range of warehouse robots within its logistics operations, and it’s looking to increase their intelligence, so they can perform various tasks. That would be a big improvement, as today’s robots are generally designed to do only one job.

To that end, Amazon’s new group will try to develop an agentic AI framework for what are known as “physical AI” applications.

By embedding agentic AI capabilities into robots, the machines will be able to unload trailers and then retrieve any parts that need repairs, Amazon said at a presentation on Wednesday. The agentic AI software will make it possible for the robots to “hear, understand and act on natural language commands,” the company explained, transforming them into an army of flexible, automated workers with multiple skills.

Yesh Dattatreya, a senior applied science manager at Amazon Robotics, told Reuters the main benefit for Amazon’s customers will be faster deliveries. He believes they can be especially helpful during periods of high demand, such as holidays, helping to process orders faster and streamline warehouse operations by lifting heavy objects in confined spaces. They would also help to keep waste to a minimum and reduce carbon emissions, he said.

Amazon’s agentic AI robots are still at the very early stage, however, and the company has not yet decided what form factor they will take, or when and how many will be deployed, Dattatreya stressed.

Amazon already has experience with AI agents. Earlier this year, the company’s AI lab launched a web browser agent called Nova Act that’s capable of performing tasks online. It has also debuted an advanced version of its digital voice assistant, called Alexa+, which has some agentic capabilities.

AI-infused maps

In a separate announcement at the same event, Amazon said it’s working on the creation of more advanced mapping technologies for delivery drivers in an effort to improve their efficiency. The tech will provide more granular detail on things such as building shapes, and any obstacles drivers might need to navigate past on their way to dropping off a package.

“This innovation is making it easier for Amazon drivers to find the right delivery spot, especially in tricky places like big office complexes,” the company said at the event.

Amazon hinted that these advanced maps may be integrated with the specialized eyeglasses it’s developing for delivery drivers. Last year, it was reported that the company is working on glasses that feature an embedded display that can show a real-time map and provide real-time, turn-by-turn directions. The idea is that the drivers can keep their hands firmly on the wheel at all times, instead of stopping to fiddle with a GPS device – or worse, doing so without stopping.

Viraj Chatterjee, who is vice president and general manager of Amazon Maps and Geospatial, confirmed the company is working on such a device, the first time any company official has done so. He told Reuters that the new map technology may be used with such a device, but added that the hardware is still under development. He said that early tests show the software is especially helpful in large housing developments, where many buildings have an identical appearance.

Constellation Research Inc. analyst Ray Wang said Amazon’s robotics and mapping initiatives are a great example of how AI can help to drive exponential efficiencies in terms of supply chain optimization.

“The improved mapping capabilities really make it easier for faster and more accurate deliveries, and the agentic AI in physical robots will not only help with warehouse and transportation management, but maybe front-facing customer experiences in time,” the analyst said. “Amazon’s large and rich datasets give it a tremendous advantage when it comes to AI innovation and developing products that deliver on customer centricity.”

Image: SiliconANGLE/Microsoft Designer