ShipBob said on November 10 that it has made Bobby, its in-dashboard AI agent, available to every brand on the fulfillment platform. In a press release issued on PR Newswire, the Chicago company said the agent lives inside the ShipBob dashboard, answers questions about a brand’s fulfillment in plain language, and can act on those answers, from flagging a stockout risk to clearing a batch of stuck orders. Every change, the company said, waits for the brand’s approval before it runs.
“Everyone has an AI story right now. The difference is whether the AI can actually do the work and that comes down to the stack underneath it,” said Dhruv Saxena, co-founder and chief executive officer of ShipBob. “Bobby can because it runs on the systems we built and operate ourselves. The same ones our teams use to fulfill more than 100 million orders annually. We believe any brand, at any size or complexity, should be able to run a global supply chain just by talking to it. As of today, every ShipBob customer can.”
ShipBob said Bobby runs on a fulfillment stack the company has built and operated in-house since 2014, including the dashboard, a fulfillment decision engine, the company’s warehouse-management system, and its transportation logic. Because it owns those systems, the company said, the agent can take action in a brand’s account, such as reassigning a shipment or updating a stock-keeping unit, on the same systems ShipBob uses day to day.
According to the release, a brand can ask for on-time delivery rates, orders shipped within a service-level agreement, and average fulfillment cost; see stockout risk over the next 30 days and break out stock for a stock-keeping unit by fulfillment center, which the company said has been the most common request so far; find orders stuck in exception and substitute an out-of-stock item across affected shipments; reassign a batch of shipments to another fulfillment center and upgrade a shipping option; create a replacement or expedited order; hand a conversation off to a ShipBob support agent without starting over; and upload a file for the agent to read or download a report. Sagar Kohli, a distinguished engineer at ShipBob who leads the team building Bobby, said in the release that when a supply chain leader asks the agent to reroute shipments or clear a stuck order, “it’s working on the real operation.”
The company said the agent shows exactly what it is about to do and waits for a yes before changing anything. Read access is on for every user by default, it said, and account owners decide who can make changes, can set a monthly usage limit, or can switch the agent off. Every action, ShipBob said, is logged and visible in the company’s AI Hub.
Melissa Mosheim, logistics director at Aroma360, is quoted in the release as saying the company’s team uses ShipBob’s AI tools every day. “I’m able to quickly pull the shipping reports and metrics I need without spending time digging through data,” she said. “Having that information readily available has made a real difference in how quickly we can understand performance, make decisions, and take action immediately.”
The product claims in the announcement come from ShipBob. The release does not report independent measurements of how often recommended actions are accepted, how much time brands save, or how stockout and exception rates change after the agent is turned on. The “more than 100 million orders annually” figure is ShipBob’s own account of the volume its network fulfills.
What it means for manufacturers
ShipBob’s customers are mainly consumer brands, but the failure modes it describes are familiar on a factory floor and in a finished-goods warehouse: an order is promised against stock that is not where the system says it is, and the exception is found only after a shipment is already stuck. Among this week’s supply chain news, the more useful test is whether inventory by location, open exceptions, and service-level performance can be pulled without a spreadsheet hunt — a data question that also sits underneath any supply chain AI tool that tries to flag a stockout or reroute an order.
A practical next step is to pick a week of late or short shipments and write down whether the cause was a location-level shortage, a substitution that never reached the next person, or a manual workaround that never made it into the system of record. That review is closer to supply chain consulting and to tightening inventory records than to adding a chat window. Our recent notes on DHL’s warehouse robotics and AI orchestration pilot and our guide to warehouse automation for midsize manufacturers are a starting point.
Questions about inventory accuracy, exception handling, or where AI belongs in a midsize fulfillment process?
Contact usSources
Additional sources
- Rachel Hand, “ShipBob Fall ’26 Release: Robotics, Returns, Bobby, and More for Peak,” ShipBob (updated Nov. 3, 2026), which already described Bobby as generally available to every merchant in the dashboard and, in a frequently asked questions section, distinguished Bobby from ShipBob’s Model Context Protocol connection to tools such as Claude and ChatGPT. The same post said ShipBob MCP had more than 2,200 active users and quoted Saxena as saying Bobby confirms changes before it makes them and logs every action as the user’s.
- Rachel Hand, “Ask Bobby Anything: Inside ShipBob’s In-App Fulfillment Agent,” ShipBob (updated Sept. 30, 2026), an earlier product guide published while Bobby was in beta with more than 150 merchants. It described 37 read actions and 17 write actions as of September 2026, listed file uploads among capabilities not yet available at that time, and said general availability was planned for the fall. The November 10 release includes file uploads among the agent’s functions.
- The November 10 release also said ShipBob’s Model Context Protocol had handled more than 1 million requests from thousands of active users, that the company fulfills more than 100 million orders a year from more than 50 fulfillment centers, and that it reports 99.97 percent order accuracy and more than 250 sales-channel integrations. Those figures were not independently verified for this report.
Photo: FEMA, public domain, via the National Archives and Wikimedia Commons (illustrative)
Try the Statadventure live tools
Inventory tools for US manufacturers: reorder point, safety stock, EOQ and landed cost, with the inputs and assumptions shown.
Open the live tools