An interactive, front-end demonstration of an AI-assisted expense review workflow based on GitLab's publicly available Global Travel & Expense Policy.
Live demo: juliannafan.github.io/ai-expense-auditor
Chinese photo demo: Open the online demo, or open dist/zh.html locally. On a phone, tap 拍照 to capture a receipt. The page also supports desktop camera access where available, image upload, local preview, and a simulated review flow. Receipt images stay in the browser; OCR and audit results are illustrative and do not use a backend. Access to GitHub Pages may vary on mainland China networks.
- HTML5 for the application structure and accessible workflow views.
- CSS3 for the responsive light enterprise interface, layout, and audit animations.
- JavaScript (ES6+) for scenario state, policy evaluation, risk scoring, uploads, exports, and interactive UI behavior.
- GitHub Pages for production hosting and GitHub Actions for automatic deployment from the
mainbranch.
- Three interactive expense-review scenarios: compliant travel, a government-recipient compliance hold, and an outside-Navan exception.
- Simulated document extraction, policy matching, risk scoring, approval routing, and evidence-backed decisions.
- A policy workspace containing 12 executable controls derived from the public policy snapshot.
- Expense-report search, audit history, U.S. tax-evidence views, and downloadable demonstration exports.
- Responsive desktop and mobile layouts.
No build step or API key is required.
cd dist
python3 -m http.server 8000Then open http://localhost:8000.
dist/
index.html # Application markup, data, and interactions
light-theme.css # Light enterprise UI theme
This independent simulation is based on GitLab's publicly available Global Travel & Expense Policy, using a policy snapshot dated September 13, 2026.
This project is not affiliated with, sponsored by, or endorsed by GitLab. GitLab and related marks belong to their respective owners. The demo is illustrative only and does not provide legal, tax, accounting, or compliance advice. Organizations should validate policy interpretations and configure their own approval controls before production use.
The current release is a presentation-ready prototype. Policy reasoning, OCR results, risk scores, and workflow outcomes are simulated in the browser. It does not yet connect to a production OCR pipeline, language model, vector database, accounting system, or persistent data store.
No license has been granted yet. Add an appropriate open-source license before allowing third-party reuse or redistribution.