Business brain

• 

  02-amazon-order-tracking-link.png

• 

  01-business-chat-ai-reply.png

Inspiration

Small online sellers manage orders, customer messages, inventory, and payments across several platforms. Switching between these tools makes it easy to miss a follow-up or delay an order.

We built Business Brain to bring these scattered workflows into one workspace and help sellers decide what to do next.

What it does

Business Brain combines a business chat interface with a prioritized To-Do List.

Users can select specialist agents and ask questions about sample business tasks. Business Chat now calls a language model through OpenRouter to provide recommendations based on the supplied task data.

The To-Do List demonstrates an Amazon fulfillment workflow. A seller can move from printing a shipping label to deciding whether to send a tracking link, then complete the task without leaving the card. Task completion and shipping steps remain saved after a refresh in the same browser, separately for each signed-in account.

The prototype also includes platform-specific connection forms and a dynamic connection counter.

How we built it

We built the frontend with React, TypeScript, Vite, Tailwind CSS, and React Router. Firebase Authentication supports Google and email sign-in.

A Node.js backend handles OpenRouter requests so the API key stays on the server. It passes conversation history, selected agent specialties, and sample business tasks to the model. It also validates requests and handles timeouts, rate limits, and provider errors.

The current prototype stores connection state and To-Do progress in browser storage.

Challenges we ran into

One challenge was keeping the fulfillment workflow understandable as an order moved through several steps. Updating the same card helped preserve context.

Another was preventing progress from disappearing after refresh. We added account-specific browser persistence and validated saved task IDs and states before restoring them.

When adding live AI responses, we also needed to keep credentials out of the frontend and handle unavailable models or failed requests clearly.

Accomplishments that we're proud of

We connected the chat interface to a real model API and built a working fulfillment demo with persistent progress.

An earlier version of Business Brain earned third place at the Next Founders Hackathon. That version used scripted chat responses. The latest update replaces those responses with an OpenRouter integration.

What we learned

We learned that a useful business assistant needs both relevant recommendations and a clear path for taking action. Reliable state handling and honest communication about the available data matter as much as the interface.

Current limitations

AI recommendations use sample business tasks rather than live platform data. Platform connections, label printing, tracking messages, and billing remain simulated.

The current AI backend runs locally through OpenRouter. This version does not integrate Nebius Token Factory or Nebius AI Cloud.

What's next for Business Brain

We plan to connect authorized commerce APIs, move operational data to a backend database, and introduce authenticated background workflows for fulfillment and notifications.

We also want feedback from small online sellers to identify which daily tasks would benefit most from automation.

Built With

• firebase
• node.js
• openrouter
• react
• tailwind-css
• typescript
• vite