Using LLM in e-invoicing application

tags

E-Invoicing

LLM Use Cases in invoicing-apis #

axioms #

• llm is good at reasoning and understanding human queries
• information:
  • zatca documentation
  • backend apis
  • JavaScript functions,
    • security concern
      • allow llm to execute only available javascript

Marrying LLM to specific use cases has more value then creating generic rags #

• value is more explicit to the business users
• easier adoption
  • they use LLM from the app instead of changing their whole approach
• easier to see how LLM can enhance ands save time by improving the work intelligently

Use Cases #

• e-invoicing AI BOT

  • it helps you in understands and implementing ZATCA complaint e-invoicing system
    • source of data
      • zatca pdfs, sites, EN 16931 standard pdfs and docs and references
  • Types of user
    • new users
      • The aim of this bot is to help new users understand the requirements to implement it
    • aware user
      • help understand details of system
    • advanced users
      • explain complex parts simply

• in app AI with LLM

  • asking frontend application to perform a task enabled by JS functions reddit, use agents for this

  • asking frontend to call backend api directly