Pitney Bowes Inc. is a global shipping and mailing company founded more than 100 years ago. Staying in business for over a century requires the ability to adapt and evolve as needed — a capability it has mastered with ease.
Indeed, the organization has come a long way since its humble beginnings as a postage meter company in the 1920s. It has adopted technology to update its mail-processing offerings, added equipment leasing in the 1970s, offered management services in the 1980s, and opened a technology center in the 1990s. Pitney Bowes also launched software for digital document delivery.
In the past two decades, the company started its digital transformation journey, bolstered by the 2017 acquisition of Newgistics, which added digital commerce delivery, returns, and fulfillment services for retailers and e-commerce brands to the Pitney Bowes portfolio.
Today, Stamford, Conn.-based Pitney Bowes operates three business units: Sending Technology Solutions is the tech engine of Pitney Bowes providing a host of shipping and mailing solutions to help clients efficiently send, track, and receive letters and parcels.
Presort Services, a workshare partnership with the U.S. Postal Service, has over 30 operating centers. And, finally, Global Ecommerce includes strategically located shipping hubs powered by automation that can reach every address in the U.S. and over 200 countries worldwide.
Global Ecommerce manages hundreds of millions of shipments per year for some major brands, including Beard Club, Deckers Brands, Il Makiage, Uncommon Goods, and more. As a result, operations must always run efficiently.
Pandemic labor shortages lead Pitney Bowes to automation
But in January 2020, right as the COVID-19 pandemic worsened, the e-commerce group began to identify sortation problems related to the labor shortage. In response, Pitney Bowes was looking to automate that part of the business, which, in general, is a very monotonous job with high turnover.
It was during that time that one of its technology partners, Ambi Robotics, said it could solve piece-picking problem with artificial intelligence. The company‘s team designed a system using its AmbiSort A-Series, an AI-powered robotic sorting system. It implemented the small-parcel sorter in a pilot program at the end of 2020 at the Pitney Bowes e-commerce hub in Stockton, Calif.
Ryan Hannon, vice president of industrial engineering and collaborative innovation at Pitney Bowes, oversees the automation at the logistics facilities.
“It was a lot of testing and enhancing the solution until we rolled out a full suite of eight AmbiSort A-Series doing 325 pieces per hour, which is about double what a person could do,” he said.
Ambi Robotics officials added that the AmbiSort A-Series systems are now sorting at throughput speeds exceeding 500 sorts per hour. This is due to continuous improvements to company‘s AmbiOS hardware and software, and by learning from commercial production environments, including collaborative innovation with the Pitney Bowes teams.
In 2021, the AmbiSort system was deployed to other sites. Today, Pitney Bowes uses more than 60 sortation robots across its logistics network.
“Now, all of our facilities are at least semi-automated to 80% of the way, from induction onto the sorters and down to the Ambi [robots] sorting to the final mile.” -Ryan Hannon, vice president of industrial engineering and collaborative innovation at Pitney Bowes
Ambi Robotics controls costs with modular components In addition, Ambi Robotics can keep costs down for customers because it has its own robot operating system that works with third-party hardware. It is a modular design that uses existing commercially available components.
For example, AmbiSort’s barcode scanners are from Keyence or Cognex, and it uses a robot arm from Yaskawa.
“We’re combining all of these modular components into proprietary end-to-end systems,” Ruck explained. “So, if you look at our AmbiSort A-Series, it’s deep bin picking with that industrial Yaskawa robot arm, but then handing that package off to our gantry system, which we design in-house. That gantry system then drops that package into the final-mile stack.”
“We like to have two things always working together, like human robot collaboration,” Ruck said. “But then if you look at our solution – it’s two robots collaborating into an end-to-end full-stack solution.”
For Pitney Bowes, this turnkey system meant minimal integration work with the network. This was helpful for another automation effort. In 2022, Hannon approached the Ambi Robotics engineering team about solving a separate problem in the warehouse associated with the middle-mile, intra-facility sortation, which is gaylord sortation.
The Ambi Robotics and Pitney Bowes teams collaborated to find a way to unload pressure on the main sortation equipment using the AmbiSort B-Series, which they said reduces operating costs by inducting and sorting parcels into gaylord destinations. This addressed challenges of labor-intensive manual sorting throughout the high-speed operations.
“I think three months later, we were back in Berkeley looking at a pilot solution in-person built in their lab in Northern California,” Hannon recalled. “And a month later, we were piloting one in Southern California. By that August, we had three more systems plus the pilot system in place sorting packages for us that allowed us to be successful in 2023.”
Partners integrate AI, vision systems, robot controls, and more
The ability to see and analyze what’s going on across facilities provides a way to optimize operations. To that end, the Ambi Robotics systems are giving data back to Hannon and team to be more efficient in how they operate.
“The best thing about Ambi is, the machines already know what they are getting into before they show up,” Hannon said. “We install them, and on day one, they are hitting their throughput rate.”
Ambi Robotics can go to market quickly because it is based on simulation-to-reality (sim2real) AI technology, born from breakthroughs at UC Berkeley. Where traditional AI is based on data that already exists, sim2real AI creates scenarios virtually before deployment.
“In the world of parcel sortation, you never know what the package is going to look like, so you don’t have those datasets readily available to train AI,” said Joseph Ruck, vice president of marketing and communications at Ambi Robotics. “So, we create synthetic packages in the simulated world and
position any use case possible that a robot might see.”
“We speed up that training process 10,000 times faster than training robots in the real world, [since] there are no physical world constraints in that training process because it’s all done in simulation,” he added. “So anytime a robot gets deployed, it’s already pre-trained and successfully hits those KPIs [key performance indicators] determined with that customer.”
In addition, because Pitney Bowes scaled the Ambi Robotics systems across its network, it can take real-world data and feed that into simulation.
“Now [that] they have bulletproof parcel-sortation capabilities, the robots can pick up anything they haven’t seen, but they’ve also likely seen everything now. I like to say that robots make mistakes, but never the same mistake twice, because they learn from it and share it with every other robot out in the world.” -Joseph Ruck, vice president of marketing and communications at Ambi Robotics
The HMI and analytics difference
Of course, after pilots and deployment, users need visibility into the automated processes going on in the warehouse. Using Ambi Robotics’ AmbiAccess dashboard, Hannon can see real-time analytics of all the facilities from a cloud-based business intelligence platform.
“The dashboard allows us to look at all the robots as a whole, but we take that data internally and then build out dashboards and metrics to compare the entire building operation from end to end so we can make sure we follow the proper processes,” Hannon said. “But also, my team can dive in and create more efficiencies where bottlenecks are or where something is not being done correctly.”
The ability to compare each facility’s output via a common dashboard has even led to friendly competition internally around the best bag-fill rates and the most volume through the system, keeping everyone engaged, he noted.
Furthermore, the proliferation of sortation robots has been an opportunity to upskill operators. According to Hannon, the human-machine interface (HMI) built into Ambi’s sortation system is easy for employees to learn and use. And now, instead of sorting parcels, they’re robot operators as well as robot leads who run all of the systems in the facility and are also responsible for day-to-day maintenance.
“This allowed associates to do more customer-facing activities, whether it’s handling exceptions or just making sure that that product is getting out the door at the time of need instead of sorting it,” Hannon said.
Editor’s note: To read more about how to optimize warehouses with automation and controls, download our e-book.