Invoice Agent (X12 to ERP)

Invoice Agent automates EDI 810 invoices by parsing, validating, enriching, scoring anomalies, and posting to ERP with a live dashboard and voice approvals, cutting errors and manual work.

Comment

About the Project

Inspiration I was inspired by the pain points I have seen in finance and operations teams that deal with supplier invoices in EDI 810 format. Processing them manually is slow, error prone, and often requires a lot of back-and-forth for approvals. I wanted to create a solution that could simplify this process while showing how AI and automation can fit into traditional enterprise workflows.

What I Learned

Through this project I learned the details of how EDI 810 invoices are structured and how they map to ERP workflows. I got hands-on experience with Redis for anomaly detection, Apify for vendor enrichment, and Gladia for voice transcription. I also learned how to tie all of these pieces together in a pipeline that goes from raw file ingestion to a user-friendly dashboard.

How I Built It

I built an ingestion pipeline that parses EDI files into structured JSON, validates them against business rules, enriches vendor information, and computes anomaly scores. I used SQLite to store state and created a Streamlit dashboard to visualize invoice status, errors, enrichment details, and line items. I added ERP posting with both a live client and a dry run mode for demos. To make it more interactive, I included a voice action feature so I can upload an audio note and automatically approve or reject an invoice.

Challenges I Faced

Working alone, one challenge was managing the entire stack from parsing and validation logic to UI and integrations. Another was connecting to an ERP endpoint since I did not always have one running, which I solved by implementing a dry run ERP mode. Designing validation rules that flagged real-world issues like totals mismatches or missing invoice numbers also required careful thought. Finally, making the dashboard both clear for business users and detailed enough for debugging pushed me to refine the UI multiple times.

Future Scope

Direct integrations with ERP platforms like SAP, Oracle, and NetSuite Cloud deployment with Docker/Kubernetes for scalability Smarter AI for anomaly detection and richer voice intents Collaboration features such as Slack/Teams alerts and multi-user workflows

Built With

  • apify
  • dotenv-(for-environment-configs)
  • erp-client-(dry-run-+-local-posting)-tools:-docker-(for-packaging)
  • gladia-speech-to-text-api
  • gladia-stt-databases:-sqlite-(state-tracking)
  • pandas
  • python
  • redis
  • redis-(anomaly-scoring
  • regex-cloud-&-platforms:-redis-cloud-(essentials-on-aws)
  • requests
  • sql
  • sql-frameworks-&-libraries:-streamlit
  • sqlite
  • streamlit
  • tempfile
  • vector-search)-apis-&-integrations:-apify-vendor-enrichment-api
Share this project:

Updates