Claimo

Claimo uses AI vision and agentic workflows to transform expense management—upload a receipt, get instant extraction, policy validation, and fraud detection in under 10 seconds.

Comment

Claimo - AI-Powered Expense Management

Inspiration

Manual expense reporting is one of the most tedious tasks in corporate life. Employees spend hours photographing receipts, manually entering data, and waiting days for approval. Finance teams waste valuable time reviewing submissions, checking policy compliance, and detecting fraudulent claims. We envisioned a world where AI could handle this entire workflow autonomously—where you simply snap a photo and the system does everything else. Claimo was born from the belief that agentic AI can transform expense management from a painful chore into an instant, seamless experience.

What it does

Claimo is an intelligent expense management system powered by AI agents that automates the entire expense approval workflow:

  1. AI Vision Extraction: Upload a receipt photo and Claude's vision API instantly extracts vendor, date, amount, items, and category with high accuracy
  2. Smart Policy Validation: Automatically validates expenses against company policies stored in Sanity CMS, checking spending limits, required fields, and category-specific rules
  3. Fraud Detection: Advanced AI analyzes patterns to detect duplicate submissions, suspicious amounts, and policy violations—including attendee tracking for meal expenses to prevent fraud
  4. Instant Decisions: The AI agent makes autonomous approve/reject/review decisions in under 10 seconds, complete with detailed reasoning and policy citations
  5. Beautiful UX: Modern, animated interface with real-time processing logs showing exactly how the AI reached its decision

How we built it

Tech Stack:

  • Frontend: Next.js 14 with TypeScript, Tailwind CSS for a responsive, animated UI
  • AI Agent: Anthropic's Claude 3.5 Haiku with custom agentic workflow and tool-calling capabilities
  • Vision AI: Claude's vision API for receipt OCR and data extraction
  • Policy Management: Sanity CMS for flexible, real-time policy configuration
  • Data Storage: Redis (Upstash) for fast expense data caching
  • PII Security: Skyflow integration for tokenizing sensitive attendee information (optional)

Architecture: We built Claimo using an agentic architecture where Claude acts as an autonomous agent with access to specialized tools:

  • analyze-receipt: Extracts data from receipt images using vision AI
  • fetch-policy: Retrieves relevant company policies from Sanity
  • detect-fraud: Analyzes expenses for suspicious patterns
  • store-redis: Persists expense data

The agent orchestrates these tools, makes decisions, and provides transparent reasoning for every action—creating a fully autonomous expense processing pipeline.

Challenges we ran into

  1. Skyflow Integration Complexity: We initially planned to use Skyflow for PII vaulting but encountered authentication and permission issues. Rather than block the demo, we made it optional with graceful fallbacks—a real-world engineering decision that kept the project moving.

  2. AI Vision Accuracy: Receipt images vary wildly in quality, format, and structure. We had to carefully engineer prompts and confidence scoring to handle edge cases like faded receipts, handwritten notes, and non-standard formats.

  3. Agentic Workflow Design: Building a reliable AI agent that makes consistent decisions required careful tool design, error handling, and reasoning transparency. We iterated extensively on the agent's decision-making logic to ensure it could handle complex scenarios like meal expenses with multiple attendees.

  4. Real-time Policy Validation: Integrating Sanity CMS for dynamic policy management while maintaining fast response times required careful caching and query optimization.

  5. UI/UX Polish: Creating an interface that makes AI decision-making transparent and trustworthy required innovative design—we built animated processing logs that show users exactly how the AI reached its conclusion.

Accomplishments that we're proud of

  • End-to-End Automation: Built a fully functional AI agent that handles the entire expense workflow autonomously—from receipt upload to approval decision in under 10 seconds
  • Fraud Detection Innovation: Implemented attendee tracking for meal expenses, a real-world fraud prevention feature that many expense systems lack
  • Transparent AI: Created beautiful, animated processing logs that demystify AI decision-making and build user trust
  • Production-Ready Architecture: Used enterprise-grade tools (Anthropic, Sanity, Redis) in a scalable, maintainable architecture
  • Stunning UI: Designed and implemented a modern landing page with smooth animations and a polished expense management interface
  • Graceful Degradation: Built the system to work even when optional services (like Skyflow) fail—real-world engineering at its best

What we learned

  1. Agentic AI is Powerful: Claude's tool-calling capabilities enable truly autonomous workflows that can replace entire manual processes
  2. Vision AI is Production-Ready: Modern vision models like Claude can extract structured data from images with impressive accuracy
  3. Transparency Matters: Users trust AI more when they can see the reasoning—our processing logs were key to building confidence
  4. Iterate Fast: When we hit the Skyflow blocker, we quickly pivoted to make it optional rather than letting it derail the project
  5. Design Sells: A beautiful, animated UI transforms a technical demo into a compelling product experience
  6. Tool Integration: Combining multiple services (Anthropic, Sanity, Redis) requires careful error handling and fallback strategies

What's next for Claimo

  1. Mobile App: Native iOS/Android apps for on-the-go expense submission
  2. Batch Processing: Upload multiple receipts at once for bulk processing
  3. Advanced Analytics: Dashboard showing spending patterns, policy violations, and fraud trends
  4. Multi-Currency Support: Automatic currency conversion and international expense handling
  5. Integration Ecosystem: Connect with accounting software (QuickBooks, Xero), corporate cards, and travel booking systems
  6. Custom Workflows: Allow companies to define custom approval chains and escalation rules
  7. Receipt Matching: Automatically match receipts to corporate card transactions
  8. Audit Trail: Complete compliance logging for financial audits
  9. Machine Learning Improvements: Continuously improve fraud detection and extraction accuracy based on usage data
  10. Enterprise Features: SSO, role-based access control, and multi-tenant support for large organizations

Built with ❤️ using Anthropic Claude, Next.js, and modern AI technology

Built With

  • anthropic
  • nextjs
  • redis
  • sanity
Share this project:

Updates