GrubMarket, a South San Francisco, California-based software-as-a-service (SAAS) company working with food producers, wholesalers, distributors and retailers, is expanding its use of artificial intelligence (AI) with a new AI agent. The company designed its new AI agent to turn vendor invoices and receipts into purchase orders for food distributors. The tool, called the “AI Orders – Purchase Order Agent,” was made to reduce the amount of manual data entry required from produce and other fresh food distributors.
“The food supply chain industry is undergoing unprecedented transformation driven by artificial intelligence,” said Mike Xu, CEO of GrubMarket, in a release statement. “With our new AI Orders – Purchase Order Agent, we are bringing powerful AI capabilities to another critical function within food distribution.”
The new agent follows other recent AI rollouts from GrubMarket. In June, the company announced the launch of an AI sales agent. Its other AI deployments have included a reporting AI agent on its GrubAssist AI platform. GrubMarket also created its Monitoring AI Agent that tracks financial and operational data for customers.
Now, the technology provider is turning to purchase orders.
Why GrubMarket launched its new AI agent
Xu said that by automating purchasing workflows, GrubMarket is helping distributors operate more efficiently. In turn, he believes that they can gain better visibility into their businesses, making better-informed decisions and improving profitability.
“This launch represents another important step in our broader vision of building specialized AI agents that automate and optimize every major function across the food supply chain industry,” Xu said, adding that the company remains committed to leveraging artificial intelligence to help businesses modernize their operations and unlock new levels of efficiency and growth.
Industry observers said the real test will be in the details.
Amy Dean, vice president of operations at SC Codeworks, a Columbus, Ohio-based warehouse management software company, said that every company handles purchase order entry differently. However, she sees it is most likely being driven by spreadsheets, ERP systems and planners.
“AI has the potential to reduce repetitive data entry, which would free up employees to handle exceptions, vendor relationships, and more strategic purchasing activities,” Dean said.
She noted AI will need to navigate special challenges that come with the food business.
“Food distribution especially is not always straightforward,” she assessed. “There are many variables that change often, such as recalls, substitutions, and vendor-specific requirements.”
Dean, who is an operations and IT expert with over a decade of experience in logistics, believes that it will be “critical that all of that is accounted for in the logic that is used with an AI agent.”
Moreover, she sees access to accurate data as a vital part of the new agent’s success.
“The ability for an AI agent to read invoices, receipts and photos could be beneficial for organizations that don’t currently utilize integration tools,” Dean stated. “However, that AI agent is only as effective as the data it has access to. This is where inventory accuracy becomes critical and current business processes can make or break the success.”
Larger challenges for the new agent
Dmytro Klymentiev is an AI systems engineer in Chicago who builds AI automation and ERP systems for businesses. He said reading a photo of an invoice and pulling out the line items is no longer the hard part. In his view, that is because most current AI models can do it. What GrubMarket adds, he said, is the connection. The extracted data becomes a structured purchase order inside the distributor’s own system, matched to its catalog.
“That is useful, but I would call it an incremental step rather than a new way of purchasing,” Klymentiev said.
In his view, the time savings are narrower than they look. He pointed out that while nobody needs to type out an order, someone may still need to check each draft to avoid knock-on effects from errors.
“A misread quantity or a wrong product match turns into a wrong order, then a wrong delivery, a failed invoice match and wrong stock levels,” Klymentiev stated. “The work moves from data entry to review, and the review has to be careful.”
Before adopting a tool like this, he recommends that distributors ask whether it flags a line or guesses when it’s unsure, how it matches products to their catalog, what happens with new or renamed items, and whether they can measure its error rate on their own documents before it touches live orders.
Sign up
Sign up for a complimentary subscription to Digital Commerce 360 B2B News. It covers technology and business trends in the growing B2B ecommerce industry. Follow us on LinkedIn, X (formerly Twitter), Facebook and YouTube.
Favorite