In this Partner Power Up, we sat down with OneStock to talk about practical AI in DOM and OMS, from agentic shopping and post-purchase support to faster operations for customer service and stores. We also covered the data foundations that make agents trustworthy, plus OneStock’s approach to extensibility and fulfillment orchestration.
The conversation centered on how AI is moving from hype to practical retail outcomes, with OneStock positioning “AI first” as both an internal operating model and a product strategy tied to measurable results. Key threads included agentic commerce across pre- and post-purchase experiences, the importance of trustworthy data like availability, promise, and order lifecycle status to make agents useful, and the fastest ROI coming from operational efficiency for customer service and store teams. It also covered how OneStock aims to enable both embedded and external agents through MCP tooling, accelerate configuration and implementation with agents, and differentiate fulfillment orchestration with competitive allocation, ending with a forward-looking view on connected agents working together behind the scenes.
Q: What does “AI first” mean at OneStock in practical terms?
A: OneStock describes “AI first” as using AI to solve real operational problems, not adding AI labels to features. In the session, Karthik shared that OneStock also uses internal agents (like OneBot) to help teams answer RFP questions, build assets, and resolve product questions faster, which informs how they design customer facing agents.
Q: What did the OneStock team see at NRF 2026 regarding AI in retail?
A: The main shift was that retailers and vendors are talking more about practical AI use cases, not just concepts. Karthik also highlighted growing momentum around shopping agents, even though many ecosystems are still early and not fully open for broad integrations.
Q: What is agentic commerce, and how does it work before and after purchase?
A: Agentic commerce is using AI agents to guide shoppers and customers through product discovery, purchase decisions, and post-purchase service in a conversational way. OneStock’s examples included a pre-purchase shopping assistant that recommends products based on preferences plus real availability and delivery options, and a post-purchase agent that summarizes order status, line item updates, and tracking without forcing users to stitch together multiple screens.
Q: Why are availability, delivery promise, and order lifecycle data critical for AI agents?
A: AI agents are only as useful as the accuracy of the data they can access in real time. If availability, promise dates, and order status are wrong or delayed, agents will give confident answers that cause missed SLAs, failed pickups, higher cancellations, and customer frustration.
Q: How does OneStock enable retailer owned agents and external agents like ChatGPT?
A: OneStock discussed using an MCP server and MCP tools to expose trusted commerce data, such as inventory availability, fulfillment status, and delivery options, so agents can query the OMS and DOM reliably. The goal is to make it easier to plug these capabilities into retailer experiences, whether the agent is embedded onsite or comes from an external platform.
Q: What is OneStock’s “competitive allocation,” and what results did they claim?
A: Competitive allocation sends an eligible order to multiple stores and lets the fastest store claim it, with guardrails to reclaim orders that sit too long and de-prioritize stores that frequently decline. In the session, Karthik cited an example with a 13 minute average claim time and a significant drop in cancellations after adopting the approach.
