AI in Odoo 19 Enterprise

Odoo 19 introduces a revolutionary AI framework that transforms the ERP into an intelligent, LLM-powered platform. This is the most significant AI integration in Odoo's history, providing native AI features across all major applications.

What's New in Odoo 19 AI

Odoo 19 moves beyond simple IAP-based OCR to a comprehensive AI agent system with:

Native LLM Integration: Direct support for OpenAI (GPT-3.5 through GPT-5) and Google (Gemini 1.5/2.5)
Vector Database & Semantic Search: PostgreSQL pgvector for RAG (Retrieval-Augmented Generation)
AI Agents: Configurable assistants with custom knowledge sources and tool access
AI-Computed Fields: Fields that auto-populate using AI inference
AI in Automation: Server actions and workflows powered by LLM intelligence
Cross-Module Integration: AI assistance in CRM, Documents, Knowledge, Livechat, Website, and Accounting

System Requirements

Critical Requirements for AI Features

PostgreSQL with pgvector extension: Required for vector similarity search (embeddings). The module auto-checks and enables this at installation.
API Keys: OpenAI API key OR Google AI API key with sufficient quota
Internet Access: For API calls to LLM providers (no offline mode)
Disk Space: Embeddings consume ~6KB per 1000-token chunk (1536 dimensions × 4 bytes)

Supported LLM Providers

Provider	LLM Models	Embedding Model	Best For
OpenAI	GPT-3.5 Turbo, GPT-4, GPT-4o, GPT-4.1, GPT-4.1 Mini, GPT-5, GPT-5 Mini	text-embedding-3-small (1536 dim)	Most features, best tool support
Google	Gemini 2.5 Pro, Gemini 2.5 Flash, Gemini 1.5 Pro, Gemini 1.5 Flash	gemini-embedding-001	Alternative provider, competitive pricing

API Key Configuration

Path: Settings → General Settings → AI section

OpenAI API key: Get from platform.openai.com
Google AI API key: Get from makersuite.google.comAlternative: Environment variables ODOO_AI_CHATGPT_TOKEN or ODOO_AI_GEMINI_TOKEN

AI Agents - The Core of Odoo AI

What is an AI Agent?

An AI Agent is a configurable AI assistant with its own personality, knowledge base, and capabilities. Each agent can:

Answer questions using its knowledge sources (RAG)
Execute actions via assigned tools (server actions)
Chat with users through the Discuss interface
Handle customer queries in Livechat
Generate content for emails, documents, and more

Agent Configuration Fields

Field	Purpose	Options
Name	Display name for the agent	Free text (e.g., "Support Assistant")
LLM Model	Which AI model to use	GPT-4o (default), Gemini 2.5 Flash, etc.
Response Style	Controls AI "temperature" (creativity)	Analytical (0.2), Balanced (0.5), Creative (0.8)
System Prompt	Custom instructions for agent behavior	Free text - tells AI how to behave
Topics	Groups of instructions + tools	Natural Language Search, Information Retrieval
Restrict to Sources	Only answer from knowledge base	Boolean - prevents hallucination

Example: Creating a Support Agent

Scenario: Create an AI agent to answer product support questions.

Go to AI → Agents → Create
Name: "Product Support Bot"
LLM Model: GPT-4o (best for complex queries)
Response Style: Analytical (factual answers)
System Prompt: "You are a helpful product support assistant. Answer questions about our products accurately. If you don't know, say so."
Enable Restrict to Sources
Upload product manuals to Sources tab

Technical: Agent Model Structure

Model: ai.agent | File: Enterprise/ai/models/ai_agent.py

python

class AIAgent(models.Model):
    _name = 'ai.agent'

    name = fields.Char(related='partner_id.name')
    llm_model = fields.Selection(default='gpt-4o')
    response_style = fields.Selection([
        ('analytical', "Analytical"),   # temperature: 0.2
        ('balanced', "Balanced"),       # temperature: 0.5
        ('creative', "Creative"),       # temperature: 0.8
    ])
    system_prompt = fields.Text()
    topic_ids = fields.Many2many('ai.topic')
    sources_ids = fields.One2many('ai.agent.source')
    restrict_to_sources = fields.Boolean()

The "Ask AI" Button (Natural Language Search)

How Ask AI Works

The Ask AI feature lets users search Odoo data using natural language:

User types: "Show me all sales orders over $10,000 this month"
AI parses: Identifies model (sale.order), filters (amount_total > 10000), date range
AI builds domain: [('amount_total', '>', 10000), ('date_order', '>=', '2024-12-01')]
AI opens view: Navigates to filtered list view automatically Available from: Command palette (Ctrl+K) → "Ask AI"

RAG Sources - Building Knowledge Bases

What is RAG?

RAG (Retrieval-Augmented Generation) allows AI agents to answer questions based on your specific documents rather than general knowledge. This prevents hallucination and ensures accurate, context-specific responses.

Source Types

Type	Supported Formats	Use Case
Binary (File)	PDF, DOC, TXT, HTML, images	Product manuals, policies, contracts
URL	Web pages (auto-scraped)	Knowledge base articles, documentation sites
Document	Odoo Documents app files	Internal documents already in Odoo
Knowledge Article	Odoo Knowledge app articles	Wiki content, procedures

How Indexing Works

mermaid

graph LR
    A[Upload File] --> B[Chunking]
    B --> C[Embedding]
    C --> D[Vector Search]
    D --> E[Retrieve Chunks]
    E --> F[AI Response]

Source Status Lifecycle

Status	Meaning	Action Required
Processing	Being chunked and embedded	Wait for cron job to complete
Indexed	Ready for use in RAG queries	None - source is active
Failed	Error during processing	Check error details, retry

Technical: Vector Embedding Architecture

Model: ai.embedding | Uses PostgreSQL pgvector extension

python

# Vector field definition (1536 dimensions for OpenAI)
embedding_vector = Vector(size=1536)

# IVFFlat index for fast similarity search
_embedding_vector_idx = models.Index(
    "USING ivfflat (embedding_vector vector_cosine_ops)"
)

# SQL for similarity search (cosine distance)
"""
SELECT id, 1 - (embedding_vector <=> %s::vector) AS similarity
FROM ai_embedding
ORDER BY similarity DESC
LIMIT 5
"""

AI Fields - Auto-Fill with AI

What are AI Fields?

AI Fields are model fields that compute their values using AI. Instead of writing Python code or relying on user input, the field value is generated by an LLM based on a prompt template.

Supported Field Types

Field Type	Auto-Update	Example Use Case
Char, Text, HTML	Yes (on save)	Product descriptions, summaries
Selection	Manual trigger	Priority classification, categories
Many2one	Manual trigger	Auto-assign salesperson, category
Many2many (Tags)	Manual trigger	Auto-tagging based on content
Integer, Float, Monetary	Manual trigger	Scoring, estimated values
Date, Datetime	Manual trigger	Suggested deadlines
Boolean	Manual trigger	Yes/No classification

Field Reference Syntax

Use {{field_name}} to reference other fields in your prompt:

{{name}} - Current record's name
{{partner_id.name}} - Related partner's name
{{order_line.product_id.name}} - Products in order lines

Example: AI Product Description

Scenario: Auto-generate marketing descriptions for products. Field: marketing_description (Text, AI-computed) Prompt:

text

Write a compelling product description for "{{name}}".
Key features: {{description_sale}}
Category: {{categ_id.name}}
Keep it under 100 words, professional tone.

Result: AI generates unique descriptions based on each product's attributes.

Technical: Creating AI Fields in Code

python

class ProductTemplate(models.Model):
    _inherit = 'product.template'

    # AI field with prompt in field definition
    ai_description = fields.Text(
        string="AI Description",
        ai="""Write a marketing description for {{name}}.
Features: {{description_sale}}
Category: {{categ_id.name}}"""
    )

Cron Job: ai_fields.ir_cron_fill_ai_fields processes empty AI fields in batches.

AI Server Actions - Intelligent Automation

The "AI" Action Type

Odoo 19 adds a new server action type: "AI". This action uses an LLM to intelligently execute operations based on context.

AI Tool Execution Flow

Trigger: Automation rule fires (on create, update, scheduled, etc.)
Context: AI receives record data, available tools, and prompt
Decision: LLM decides which tool(s) to call based on instructions
Execution: Tool runs with AI-provided parameters
Loop: Can call multiple tools in sequence until __end_message

Action Types That Can Be AI Tools

Action Type	What AI Can Do
Execute Code	Run Python code with AI-provided parameters
Create Activity	Schedule activities with AI-determined details
Create Record	Create records with AI-populated values
Update Record	Modify fields based on AI analysis
Add Followers	Add relevant users as followers
Send Webhook	Trigger external APIs with AI parameters
Post Message	Post AI-generated messages to chatter

Example: AI Lead Qualification

Scenario: Automatically qualify leads based on description and source.

Create Server Action: Type = "AI"
Prompt: "Analyze this lead and determine quality (hot/warm/cold). Consider: {{description}}, {{source_id.name}}, {{expected_revenue}}."
Add tool: Update Record (priority field)
Create Automation: Trigger on lead creation Result: New leads automatically get priority assigned by AI analysis.

AI Livechat - Automated Customer Service

AI Chatbot in Livechat

AI Agents can be assigned to Livechat channel rules, allowing them to handle customer queries 24/7 without human intervention.

Setting Up AI Livechat

Go to Website → Configuration → Livechat Channels
Select your channel, go to Rules tab
Create or edit a rule
In AI Agent field, select your agent
Configure URL pattern and conditions Note: System agents (internal) cannot be assigned to livechat.

AI Livechat Features

Feature	Description
RAG Responses	Answers from uploaded knowledge sources
Lead Generation	Creates CRM leads from chat (ai_crm module)
Human Escalation	Transfers to human operator when needed
Scripted Bot Fallback	Can forward to chatbot script

Predictive Lead Scoring (PLS) - Statistical AI

Not LLM-Based AI

Predictive Lead Scoring uses Bayesian statistics (classical machine learning), not LLM/GPT technology. It's been in Odoo CRM (Community) for several versions and works differently from the new AI modules.

How It Differs from ChatGPT-Style AI

Aspect	Predictive Lead Scoring (PLS)	LLM AI (Odoo 19 AI Modules)
What it is	Statistics-based prediction	Language understanding (like ChatGPT)
Learns from	YOUR historical won/lost leads only	Billions of internet documents + your prompts
What it outputs	A number: "37% chance to win"	Text: emails, summaries, classifications
Needs training data?	Yes - needs your past leads	No - works immediately with prompts
Data privacy	100% local - nothing leaves your server	Sent to OpenAI/Google APIs
Cost	Free (included in Community)	Pay per API call (Enterprise only)
Best for	Prioritizing leads, forecasting revenue	Writing content, understanding text

Two Probability Fields on Leads

Probability (%): Can be manual OR automated - shows current win likelihood
Automated Probability: Always calculated by PLS - the statistical prediction

If you manually change Probability, it detaches from automated updates. Reset by setting it equal to Automated Probability again.

AI Documents - Intelligent Filing

Automatic Document Sorting

AI can automatically sort uploaded documents into the correct folders, add appropriate tags, and trigger workflows based on content analysis.

Document Sorting Flow

Document arrives: Upload, email, or scan
AI analyzes: Reads content, identifies document type
AI decides: Matches to folder based on configured prompts
AI executes: Moves to folder, adds tags
Workflow triggers: Downstream automations fire

Example: Invoice Auto-Filing

Go to Documents → Configuration → Folders
Select or create "Vendor Bills" folder
Enable AI Auto-Sort

Configure prompt:

text

Analyze this document. If it's a vendor invoice or bill:
- Move to folder: [Vendor Bills folder ID]
- Add tags: "invoice", vendor name
If it's not a vendor invoice, do nothing.

Enable tools: move_to_folder, add_tags

Other AI Features

AI Website Page Generation

Module: ai_website Generate website page content with AI:

Multi-tone support: Concise, Professional, Friendly, Persuasive, Informative
Multi-language: Generates in website's default language
Placeholder replacement: AI fills template snippets with relevant content How to use: Website → Create Page → Check "Generate with AI" → Enter instructions

AI Knowledge Article Drafting

Module: ai_knowledge Draft Knowledge articles with AI assistance:

Click AI button in article editor
Describe what you want to write
AI generates structured article content
Articles can also be used as RAG sources

VoIP Call Transcription

Module: voip_ai Automatic call transcription and summarization:

Transcription: Uses OpenAI Whisper API for audio → text
Summary: AI generates one-liner summaries (max 60 chars)
Status tracking: pending → queued → done/error Configuration: VoIP Provider → Transcription Policy = "Force for all users"

Web Studio AI Fields

Module: web_studio_ai_fields Create AI-computed fields directly in Web Studio:

No coding required
Supports all 11 field types
Configure prompts with /field syntax
Visual prompt editor with placeholder insertion How to use: Studio → Add Field → Select "AI Field" → Configure prompt

Practical AI Use Cases - Try These!

Use Case 1: Smart Lead Qualification

Goal: Automatically score and prioritize incoming leads Setup Steps:

Create AI Field on crm.lead: "AI Priority Score" (Selection: Hot/Warm/Cold)

Prompt:

text

Analyze this lead. Company: {{partner_name}}, Revenue: {{expected_revenue}}, Source: {{source_id.name}}, Description: {{description}}. Rate as hot (ready to buy), warm (interested), or cold (just browsing).

Create automation to notify sales team when AI marks lead as "Hot" Result: Sales team focuses on high-value leads first, improving conversion rates.

Use Case 2: Product Description Generator

Goal: Auto-generate SEO-friendly product descriptions Setup Steps:

Create AI Field on product.template: "Marketing Description" (HTML)

Prompt:

text

Write a compelling product description for "{{name}}". Category: {{categ_id.name}}. Features: {{description_sale}}. Price: {{list_price}}. Include benefits, use cases, and a call-to-action. Format with bullet points.

Use on website product pages via Website Builder Result: Consistent, professional product descriptions without manual writing.

Use Case 3: Customer FAQ Chatbot

Goal: 24/7 customer support using your knowledge base Setup Steps:

Create AI Agent: "Customer Support Bot"
Upload FAQ documents, product manuals as RAG sources
Enable "Restrict to Sources" to prevent hallucination
Assign agent to Livechat channel rule for website visitors
Configure human escalation for complex queries Result: Instant answers to common questions, human agents handle complex cases only.

Configuration & Best Practices

API Key Setup

OpenAI (Recommended): platform.openai.com → API Key. Best for most features and GPT-4 model support.
Google Gemini: makersuite.google.com → API Key. Good alternative, competitive pricing.

Cost Considerations

Embedding calls: Cheap (~$0.02/1M tokens)
LLM calls: Varies ($2-20/1M tokens). GPT-4o is more expensive than GPT-4o-mini.
Whisper transcription: $0.006/min. Recommendation: Use GPT-4o-mini or Gemini Flash for high-volume tasks, GPT-4o for complex reasoning. Set spending limits.

Security Considerations

Data Usage: Prompts and context are sent to the AI provider.
Access Rights: AI respects Odoo access rights. It only "sees" what the user has permission to see.
Secrets: Keep API keys secure. Use environment variables.

Complete AI Module Reference

Module	Feature	Auto-Install	Dependencies
`ai`	Core AI framework	No	mail
`ai_app`	Agent management UI	No	ai, attachment_indexation
`ai_fields`	AI-computed fields	Yes	ai
`ai_server_actions`	AI in automations	Yes	ai_fields
`ai_livechat`	AI chatbot	Yes	ai, im_livechat
`ai_documents`	Document sorting	Yes	ai, documents
`ai_documents_source`	Documents as RAG sources	Yes	ai, documents
`ai_crm`	Lead generation	Yes	ai, crm
`ai_website`	Website content generation	Yes	ai, website
`ai_knowledge`	Article drafting	Yes	ai, knowledge
`voip_ai`	Call transcription	Yes	ai, voip
`hr_recruitment_ai`	Rejection emails	Yes	ai, hr_recruitment
`sign_ai`	Signature request emails	Yes	ai, sign
`ai_account`	Invoice emails	Yes	ai, account
`web_studio_ai_fields`	Studio AI field creation	Yes	ai_fields, web_studio

AI in Odoo 19 Enterprise ​

System Requirements ​

Supported LLM Providers ​

AI Agents - The Core of Odoo AI ​

Agent Configuration Fields ​

The "Ask AI" Button (Natural Language Search) ​

RAG Sources - Building Knowledge Bases ​

Source Types ​

How Indexing Works ​

Source Status Lifecycle ​

AI Fields - Auto-Fill with AI ​

Supported Field Types ​

Field Reference Syntax ​

AI Server Actions - Intelligent Automation ​

AI Tool Execution Flow ​

Action Types That Can Be AI Tools ​

AI Livechat - Automated Customer Service ​

Setting Up AI Livechat ​

AI Livechat Features ​

Predictive Lead Scoring (PLS) - Statistical AI ​

How It Differs from ChatGPT-Style AI ​

AI Documents - Intelligent Filing ​

Document Sorting Flow ​

Example: Invoice Auto-Filing ​

Other AI Features ​

AI Website Page Generation ​

AI Knowledge Article Drafting ​

VoIP Call Transcription ​

Web Studio AI Fields ​

Practical AI Use Cases - Try These! ​

Use Case 1: Smart Lead Qualification ​

Use Case 2: Product Description Generator ​

Use Case 3: Customer FAQ Chatbot ​

Configuration & Best Practices ​

API Key Setup ​

Cost Considerations ​

Security Considerations ​

Complete AI Module Reference ​

AI in Odoo 19 Enterprise

System Requirements

Supported LLM Providers

AI Agents - The Core of Odoo AI

Agent Configuration Fields

The "Ask AI" Button (Natural Language Search)

RAG Sources - Building Knowledge Bases

Source Types

How Indexing Works

Source Status Lifecycle

AI Fields - Auto-Fill with AI

Supported Field Types

Field Reference Syntax

AI Server Actions - Intelligent Automation

AI Tool Execution Flow

Action Types That Can Be AI Tools

AI Livechat - Automated Customer Service

Setting Up AI Livechat

AI Livechat Features

Predictive Lead Scoring (PLS) - Statistical AI

How It Differs from ChatGPT-Style AI

AI Documents - Intelligent Filing

Document Sorting Flow

Example: Invoice Auto-Filing

Other AI Features

AI Website Page Generation

AI Knowledge Article Drafting

VoIP Call Transcription

Web Studio AI Fields

Practical AI Use Cases - Try These!

Use Case 1: Smart Lead Qualification

Use Case 2: Product Description Generator

Use Case 3: Customer FAQ Chatbot

Configuration & Best Practices

API Key Setup

Cost Considerations

Security Considerations

Complete AI Module Reference