Inspiration
Small businesses often rely on multiple people and disconnected tools to manage sales, inventory, reports, and purchasing decisions. This makes daily operations time-consuming and increases the risk of missed stock shortages and delayed decisions. We built OpsPilot to bring these workflows into one intelligent system where AI helps business owners understand what needs attention and what actions to take next.
What it does
OpsPilot is an AI-powered business operations assistant that analyzes sales and inventory data, identifies low-stock products, highlights operational risks, and recommends next steps. It coordinates workflows through tools and presents actionable insights in a centralized dashboard. Importantly, purchasing decisions remain under human control: the system can prepare a restock proposal, but human approval is required before proceeding.
How we built it
We built OpsPilot using Next.js, React, TypeScript, and a modular operations layer. WebMCP tools connect the assistant to business functions such as retrieving sales, checking inventory, identifying low-stock items, and generating reports. We integrated Nebius Token Factory to access NVIDIA Nemotron for AI-powered operational reasoning. The model receives business telemetry and produces structured summaries, risk assessments, and recommendations. The dashboard brings these results together with workflow activity, reliability indicators, and human approval controls.
Challenges we ran into
One of our main challenges was integrating AI reasoning into an existing operational workflow without disrupting the application's current functionality. We also had to consider incomplete data, API failures, and the risk of AI generating unsupported recommendations. We addressed these concerns by keeping business operations separate from the reasoning layer, handling failures explicitly, and maintaining human approval for purchasing actions.
Accomplishments that we're proud of
We are proud to bring multiple business operations into one workflow instead of treating sales analysis, inventory monitoring, and reporting as separate tasks. OpsPilot combines tool-based execution, AI-generated recommendations, activity tracking, and human oversight in one dashboard. We also extended the project toward NVIDIA Nemotron through Nebius Token Factory, creating a foundation for more intelligent and reliable business operations.
What we learned
We learned how to connect AI models with real application workflows, structure business data for model-based reasoning, and design safeguards around AI-generated recommendations. We also gained experience with WebMCP tool integration, API handling, modular architecture, and the importance of keeping humans involved in consequential business decisions.
What's next for OpsPilot
Our next steps are to improve the accuracy of operational recommendations, expand integrations with business platforms, add configurable restocking policies, and strengthen evaluation of model outputs. We also plan to improve failure recovery, add more comprehensive audit trails, and support additional business workflows. Our long-term vision is to make OpsPilot a reliable AI operations partner that helps small businesses save time, reduce operational risks, and make better-informed decisions.
Log in or sign up for Devpost to join the conversation.