We are looking for an n8n automation developer to build a small but functional AI-powered finance expense review proof of concept. The system should receive an expense request, process it through an n8n workflow, use an AI model to classify and summarize the request, check it against one sample expense-policy document using a basic RAG process, save the result in Supabase, and send a review notification. This is a limited proof of concept using test data. It is not a production accounting platform or a complete financial management system. Required Workflow 1. Receive expense data through an n8n webhook. 2. Validate the required input fields. 3. Clean and transform the submitted data. 4. Send the expense description to an AI model. 5. Return a structured AI response. 6. Retrieve relevant information from one sample expense-policy document. 7. Compare the expense request with the retrieved policy information. 8. Flag missing information or possible policy conflicts. 9. Apply basic rule-based approval routing. 10. Save the expense and AI-review result in Supabase. 11. Send an email notification containing the review summary. 12. Log workflow failures through an error-handling branch. Input Data The webhook should accept test data containing: • Employee name • Employee email • Expense title • Expense category • Expense amount • Transaction date • Expense description • Receipt availability • Request ID A basic form, Postman request, or minimal frontend page may be used to submit the data. A complete authentication system or advanced dashboard is not required. AI Review Requirements The AI component should: • Summarize the expense request • Suggest an expense category • Identify missing information • Highlight suspicious or unclear information • Review the request against retrieved policy content • Recommend manual approval, rejection, or additional review The AI output should use structured JSON containing fields similar to: requestSummary suggestedCategory missingInformation policyStatus policyExplanation riskFlag recommendedAction The AI recommendation is for demonstration purposes only. It must not automatically make an irreversible financial decision. Basic RAG Requirements Use one short sample expense-policy document. The implementation should: • Load the policy content • Divide it into searchable sections • Generate and store embeddings • Retrieve the most relevant policy sections • Pass the retrieved context to the AI model • Generate a short policy-compliance explanation Supabase pgvector or another simple vector-storage method may be used. Only one policy document and one basic retrieval workflow are required. Rule-Based Routing • Expenses below $100 follow the standard review path • Expenses from $100 to $500 require administrator review • Expenses above $500 are marked for additional approval • Requests without receipt confirmation are flagged • Requests with missing fields are sent to a manual-review branch • Possible policy conflicts are flagged for manual review The rule values may use test data and should be easy to modify. Supabase Requirements Store the following information: • Original expense data • Workflow status • AI-generated summary • Suggested category • Policy-review result • Risk flag • Recommended action • Processing timestamp • Error status, when applicable A simple, understandable table structure is sufficient. Notification Requirements Send one email notification after successful processing. The email should contain: • Request ID • Employee name • Expense title • Amount • Suggested category • AI-generated summary • Policy status • Risk flag • Recommended action A separate failure notification or error log should be created when the workflow cannot complete. Error Handling The workflow should include branches for: • Missing required fields • Invalid webhook input • AI API failure • Supabase connection failure • Empty RAG retrieval result • Email delivery failure Failed executions should not stop silently. Preferred Technology • n8n • OpenAI API, Claude API, Gemini API, or another suitable model • Supabase • PostgreSQL • Supabase pgvector or another basic vector store • REST APIs • Webhooks • JavaScript where custom n8n logic is required Deliverables • Working n8n workflow • Webhook-based expense input • AI classification and structured output • Basic RAG policy-check process • Rule-based approval routing • Supabase integration • Email notification • Error-handling branches • Exported n8n workflow JSON file • Supabase table schema or SQL setup file • One sample expense-policy document • Sample test data • Environment-variable list • Brief setup instructions • Short demonstration video or screen recording • One minor revision for functional issues Scope Limitations The following are not included: • Production financial compliance • Banking integrations • Payment processing • Advanced OCR • Complete authentication system • Full employee or administrator dashboard • Multi-agent architecture • Multiple policy documents • Advanced analytics • Enterprise deployment • Real financial or customer data All development and testing must use sample data. Budget and Timeline Fixed budget: $50 Expected delivery time: 3 to 5 days Please apply only if you can complete the defined proof of concept within the stated fixed budget. Application Requirements When applying, include: 13. A brief explanation of your proposed n8n workflow 14. Your experience with n8n and API integrations 15. Your experience connecting n8n with Supabase 16. Your proposed approach for the basic RAG policy check 17. One relevant automation example 18. Confirmation that you accept the $50 fixed budget 19. Your estimated delivery time Begin your proposal with “Finance AI Workflow” to confirm that you reviewed the complete requirements.