Table of Contents

AI

What is ERP.net AI?

ERP.net AI brings smart assistants and functionalities directly into your ERP, so people can:

  • Ask questions in chat instead of clicking through menus.
  • Let AI read and explain project data, tasks, and documents.
  • Automate routine work – from creating leads to validating business rules.

Under the hood, ERP.net connects securely to external AI providers (currently OpenAI). You choose which models to use, what they are allowed to know, and where they can help in your processes.

AI Models in ERP.net are hierarchical, allowing each model to inherit context and knowledge from parent models.
This structure supports both company-wide master models and specialized assistants for specific departments or workflows (for example, Marketing AI or Sales AI).

The models based on OpenAI’s technology can also be fine-tuned and trained directly within ERP.net, using the company’s own data, procedures, and best practices.
This enables the AI to understand the unique rules, terminology, and business logic of each organization, providing tailored and context-aware assistance.

By embedding this technology throughout ERP.net, users can:

  • Chat with an AI Assistant directly inside the application.
  • Involve AI as a Chat Arbiter in group or private conversations.
  • Use AI in form panels to analyze, explain, or generate business data.
  • Automate and validate operations through AI-enhanced business rules.
  • Employ AI-powered tools such as document ingestion, lead creation, and intelligent task generation.

The goal is to make AI a natural and context-aware part of everyday ERP processes, improving accuracy, speed, and decision-making across all departments.

Note

💡 Using OpenAI with ERP.net

ERP.net AI features use the OpenAI API behind the scenes.

  • Each customer creates their own OpenAI account and API key.
  • OpenAI bills you directly based on your usage.

This keeps ERP.net transparent about costs and allows you to control your AI budget.

Architecture

ERP.net AI is a modular system. You plug in an external AI provider, define one or more AI models (assistants), teach them with your data, and then use them in chats, forms, and business rules.

Key building blocks

Providers
Think of this as the “connection” to an AI service.

  • Stores the API credentials and base model (e.g., OpenAI GPT model).
  • Can be configured once per AI provider and used by all models.

Models
Your actual AI assistants.

  • Each model represents a specific assistant (e.g, "My Assistant", “Project AI”, “Sales AI”, “Master AI”).
  • Models can inherit settings and knowledge from a parent model.

Model QA's
Structured Q&A knowledge base.

  • Store question–answer pairs with your company rules and best practices.
  • Used to fine-tune how the model answers typical questions.

Training Conversations & Training Conversation Messages
Real chats used as training examples.

  • Capture real user ↔ AI interactions (via “Train AI”).
  • Used during compilation to teach the model how your users talk.

Compilations & Compilation Assets
Turning training data into a ready-to-use AI

  • Compilations bundle instructions, data, and context.
  • A successful compilation produces a fine-tuned AI model version.

Assistant Conversations
Day-to-day user chats with AI

  • Store the conversation history between users and AI assistants.
  • Ensure auditability and data security.
Suggest improvement to this page