AI Applications in Customs Still Need Human Touch: Panelists

“You want their fingers off the keyboard so they can be thinking about what they're doing and not just trying to get the data entered into your system,” said panel moderator Sandy Coty, director of operational development for A.N. Deringer. While the integration of artificial intelligence won’t and shouldn’t replace years of experience in the customs realm, the application of AI can help free up time and improve brokers’ compliance, Coty said.

Cindy Allen, CEO and managing director of Trade Force Multiplier, said AI relies on humans to feed AI models with good data and give context to analyses carried out by AI.

“Giving that context is really important,” and companies like Allen’s, as well as CBP, want to work with good data that’s been sourced and vetted, Allen said.

AI also can allow smaller brokers to compete with larger companies -- a trend preceded by initiatives such as importer self-filing and automated broker interface.

Allen said she joined the board of Altana AI, an AI supply chain management company, because she saw the potential uses for AI tools when she worked as a broker.

“Now as an independent consultant, I want tools that I consult to my clients. I want tools that they can implement to do auditing, to do self-assessments, to do risk assessments," Allen said. "And that's, I think, important in what we all want these tools to do."

The types of AI tools that brokers may consider using include automation tools that help with customs filings, more sophisticated supply chain mapping tools, or tools that aid compliance management with the Harmonized Tariff Schedule.

AI “really frees staff off to talk to customers about more important things, more strategic things,” said Vincent Iacopella, president of trade and government relations for Alba Wheels Up. There are even discussions brewing about using mapping and AI for verifying credentials, as well as risk management tools for forced labor, carbon emissions and sustainability.

“How do I integrate these shiny objects into my regular ERP [the enterprise resource planning] system? ... How do I get them more into my environment? These are all solvable things, but these are questions as a broker that you should be asking when you start to support this,” Iacopella said.

Companies and brokers also should consider how much and what kind of AI integration they want to take on, said Mark Ketcham, senior vice president of sales for Raft AI.

The integration of AI tools follows a similar trajectory to self-driving cars, Ketcham said, which can include semi-automation to automation with the assistance of a driver.

“As you delve into AI solutions [and] how that will affect your business, really consider determining what you're going to embrace,” Ketcham said.

For brokers, the stages of automation can start with data extraction, which involves “driving that data into our systems of record. How do you take a bunch of data, disseminate what's important, and pull that to the forefront to make your entry, to create your shipments,” Ketcham said.

“The longer-term value becomes a transformation of the daily life of your team members, the quality of life, taking away pages of data, phone calls, emails,” Ketcham said.

He continued: “We've all heard you can't improve what you can't measure. How do we measure things ... that are done manually? Now you can see this data, this flow, what it takes to actually project how much time an entry should take, which is shocking for all of us that have been around for a long time. But what should this take now? We can start measuring. We can actually improve.”

While the range and scope of AI tools might seem intimidating, brokers should start by assessing what their needs are, said John Pennypacker, vice president of sales and marketing for Deep Cognition.

While there is software out there that can do impressive tasks, it won’t help a business if the company doesn’t realize that it needs to get its database in order first, Pennypacker said.

“The first thing that you need to do is figure out the problems and then figure out the solutions, and then actually create a roadmap for yourself to say, hey, the first thing that we're going to try to do is low-hanging fruit, low effort, high value: product classification for what we need, or document automation or supply chain intelligence, whatever it is,” Pennypacker said.

A company should determine its problems, as well as the problems of its clients, and then build around that, he continued.

“You don’t want to put AI on top of a bunch of broken or not well-documented processes,” Pennypacker said.