A new review of AI use finds the postal industry is still sorting out what works at scale

Guest:	Rick Schadelbauer
Title:	Research specialist with the USPS Office of Inspector General
Summary:	Postal operators around the world are testing artificial intelligence in everything from package processing to customer service. A new review from the USPS Inspector General looks at where those efforts are producing measurable results and where the technology remains more promise than practice.

The Federal Drive with Terry Gerton provides expert insights on current events in the federal community. Read more interviews to keep up with daily news and analysis that affect the federal workforce. Reach out to Terry and the Federal Drive producers with feedback and story ideas at FederalDrive@federalnewsnetwork.com.

Interview transcript

Terry Gerton You have just done some really interesting research on how postal and logistics organizations are using AI around the world. I’m a former military logistician, so anytime I hear that word, it picks my attention up. And so I’m curious, why treat the U.S. Postal Service as a logistics function? When you did that, did it open up some insights into this research?

Rick Schadelbauer Yeah. You know, when you think about it, Terry, a lot of what they do entails logistics. Probably most of their function entails logistical details and planning and things like that. So we wanted to take a look at how they were using artificial intelligence to help them do that to do it either more efficiently or more effectively or at a lower cost or maybe all of the above.

        CPE Webinar: Join us Oct. 6 at 2 pm ET for a discussion on AI policy's role in shaping system risk, sponsored by Billington CyberSecurity.

Terry Gerton What were some of the logistics functions that you looked at for AI applications? Planning shipments, managing inventory, anything particular?

Rick Schadelbauer Yeah, we looked at how they sort parcels and mail. We looked at they plan their delivery routes, and we looked how they interact with their customers to provide an excellent customer experience.

Terry Gerton And as you looked at all of these different organizations around the world, what did you find about where the postal industry really stands today?

Rick Schadelbauer Yeah, so we, as part of our research, we contacted 20 different organizations, the foreign posts, logistics providers, consultants, as well as the postal service itself. And we did interviews with all of them. And what we found was that all the posts we contacted are using AI in some form, but they’re all at different stages of implementation. There are a lot of pilots and test programs going on, but we also found a lot established use cases. Most of the AI today in the postal and logistics sectors is used for discrete tasks. It’s basically assisting human workers, although we did find that it is slowly beginning to evolve toward more sophisticated applications. The postal service has been using optical character recognition or OCR and chatbots for decades now. AI is really not new to them and they’re continuing to expand their applications of artificial intelligence.

Terry Gerton Were there any places that you found in the early stages of AI deployment that really seemed promising? And if you did, where were they?

Rick Schadelbauer Yeah, so we found several really interesting and useful applications. One was in mail and parcel processing, where posts are using computer vision to correct addresses that might be illegible or missing, as well as to detect counterfeit postage, which is a very large problem for the postal sector. The USPS is using a parcel robotic sorting system that can actually sort between 3,000, 4,000 parcels a day, which really helps boost their efficiency. In delivery operations, many of the posts are using AI to develop their delivery routes that are most effective, most efficient, that allows them to hit all the addresses they need to hit in a relatively small short length of time. Then in the customer service sector, many of those posts are using AI to be able to better predict delivery times, Which means they can give their customers more information, a smaller window when their packages will be delivered. So they can make arrangements to have that package picked up when it does arrive.

Terry Gerton Rick Schadelbauer is a research specialist with the USPS Inspector General. Rick, as you’re talking about those functions, they seem like stand-alone parts of the process. Did you find anyone anywhere who was kind of looking at AI to optimize end-to-end processing?

        Sign up for our daily newsletter so you never miss a beat on all things federal

Rick Schadelbauer That is still somewhat off into the future. We identified a three-stage maturity curve in AI deployment and stage one is sort of what I described before where AI is assisting humans with discrete tasks and making their jobs easier. Stage two is where AI connects and coordinates data across different parts of the post-operations. And then stage three is the exciting future. It’s where AI becomes autonomous and makes decisions and triggers actions all by itself. What we found is that the USPS and the other posts we spoke with currently are somewhere between stages one and two. They’re moving into slightly more sophisticated applications. They’re not at the stage yet where AI can do things all by themselves, but they’re slowly but surely headed in that direction.

Terry Gerton For the organizations that you looked at that are moving forward, were there common characteristics? Was it all just about money? Did they need assets or leaders with a different vision?

Rick Schadelbauer Yeah, we found quite a few similarities among those posts that have been successful in AI implementation. Probably the most important one is having a culture that embraces change, not having a fear of failure. Letting employees, who are the ones, the workers who see the problems, letting them come to management and say, hey, here’s a situation I found. Let’s see if we can find ways to do it better. We also found the organizations that were successful had a very narrow strategic focus. They knew what they wanted to accomplish and were interested in finding the best ways to that. Equally important is having the financial resources because AI is not cheap. And that’s certainly been a challenge for the U.S. Postal Service, who is certainly not awash in cash. They need to have the necessary computing resources. AI requires quite a bit of computing power, and they also need to have staff that is trained in AI. Virtually all the posts we spoke with do have training programs to teach their employees how to use AI. Posts collect tremendous amounts of data. AI requires a lot of data, but we did find that some of the data is not of useful quality or is not in the same location, so it can be easily accessed. The ones that have done well have had data of robust quality. Finally, the successful posts had established policies and procedures for implementing AI as well as for AI ethics, making sure that they protect their customers’ data and things like that.

Terry Gerton One of the things I didn’t hear in that list was metrics. As they’re thinking about deploying these AI tools or systems in specific spots, are they thinking about ROI and what are those measures like efficiency or cost reduction or customer satisfaction that they would use to assess that?

Rick Schadelbauer Yeah, we did conclude that it’s critical that posts only pursue those AI applications that can present a positive ROI. Sometimes that’s not easily determined until after it’s been piloted, and then it can be measured that way. Certainly those three standards, efficiency, costs and customer experience are critical. It’s not always easy to measure them, particularly customer experience is difficult to quantify. But the posts are, especially when they’re in the stage of evaluating potential applications, they do look at the metrics, cost savings, efficiency savings, and then that sort of colors their decision as to whether to go forward with the application.

Terry Gerton I’m guessing you didn’t do the research just because you wanted to venture out to all of these different places. The hope would be that there are lessons here for the U.S. Postal Service. Where do you think they will take these lessons and move out to the next phase of AI deployment?

Rick Schadelbauer Yeah, well, one thing we heard was that the Postal Service is very interested in gaining the ability to use AI for dynamic route adjustments. So AI would be able to to look at road conditions in real time. And if there are problems with traffic or weather or construction, they could not only identify that, but they could suggest an alternative that could be taken to avoid those problems. They’re also very interested in looking at gaining the ability to more accurately predict shipment arrival time at their various facilities. That would allow them to have the resources ready to deal with those shipments when they do arrive. And then finally, one thing they’re already doing is a product called Advanced Expected Delivery, which gives customers a very narrow delivery window and allows them to know exactly when their parcels will be arriving at their doorstep.

Terry Gerton So if I’m a customer planning to ship my Christmas packages, AI might help me know when the best time is to put those in the mail?

Rick Schadelbauer Yes it would. I’m not AI, but generally it’s as early as possible. But yeah, certainly as a customer, you would have much more information about your shipments and when they arrive safely at the doorstep of the recipient.

        Read more: Artificial Intelligence

Copyright © 2026 Federal News Network. All rights reserved. This website is not intended for users located within the European Economic Area.