Skip to content

Building an AI Knowledge Base for Odoo: Teaching Agents Your Business

DeployMonkey Team · March 23, 2026 11 min read

Why AI Agents Need a Knowledge Base

A generic AI model knows about Odoo in general, but it knows nothing about your specific Odoo instance. It does not know that your company requires manager approval for purchases over $5,000, that product category "A" has special tax rules, that customer "VIP-Corp" always gets net-60 payment terms, or that your warehouse team follows a specific pick-pack-ship sequence. Without this context, the AI agent makes generic recommendations that do not fit your business.

A knowledge base bridges this gap — it is a structured collection of your business rules, processes, and conventions that the AI agent consults before taking any action.

What Goes in the Knowledge Base

1. Business Rules

# Document explicit business rules:
"Business Rules — Sales

  Pricing:
    - Standard discount limit: 10% (sales rep)
    - Extended discount limit: 20% (sales manager)
    - Above 20%: requires VP approval
    - Government contracts: use GSA price list
    - Distributor pricing: cost + 15% markup minimum

  Payment terms:
    - New customers: prepayment or net-15
    - Established customers (6+ months): net-30
    - VIP customers (>$100K/year): net-60
    - Government: net-45 (standard government terms)

  Order processing:
    - Orders over $50K: require credit check
    - International orders: require export compliance review
    - Custom products: 4-week lead time (non-negotiable)
    - Rush orders: +25% surcharge, needs ops manager approval"

2. Process Documentation

# Document how things actually work:
"Process: Customer Onboarding

  Step 1: CRM lead marked as Won
    - Sales rep creates customer record (res.partner)
    - Required fields: email, phone, billing address,
      tax ID, payment terms
    - Assign to appropriate sales team

  Step 2: Credit check (orders > $10K)
    - Finance reviews credit report
    - Sets credit limit on partner record
    - Approves or rejects, notifies sales rep

  Step 3: First order
    - Sales rep creates quote, applies standard pricing
    - Manager approves if discount > 10%
    - Customer receives PDF quote via email
    - Auto-follow-up after 3 days if no response

  Step 4: Post-first-order
    - Account manager introduces themselves
    - Schedule 30-day check-in
    - Add to appropriate email lists"

3. Module-Specific Configuration

# Document your specific Odoo setup:
"Configuration Notes — Inventory

  Warehouse structure:
    Main Warehouse (WH) — Chicago
      Location: WH/Stock (main inventory)
      Location: WH/Input (receiving)
      Location: WH/Output (shipping staging)
      Location: WH/Quality Control (inspection hold)
    
    East Hub (EAST) — New York
      Ships to: East Coast customers only
      Replenished from: Main Warehouse weekly

  Routes:
    Standard: Receive → Quality Check → Stock
    Express: Receive → Stock (skip QC for trusted suppliers)
    Drop-ship: Vendor ships direct to customer

  Reorder rules:
    - Updated monthly by procurement team
    - Safety stock = 2 weeks average demand
    - Lead time includes supplier + transit + inspection"

Knowledge Base Structure

SectionContentsUpdate Frequency
Business rulesPricing, approval limits, policiesQuarterly
ProcessesStep-by-step workflowsWhen process changes
ConfigurationModule-specific setup notesAfter any config change
Team conventionsNaming standards, tagging rulesAs needed
Decision patternsHow specific decisions are madeMonthly
Exception handlingWhat to do in edge casesAs discovered

Team Conventions

# Document naming and tagging standards:
"Team Conventions:

  Product naming:
    Format: [Category]-[Model]-[Variant]
    Example: SENSOR-TS400-HT (high temperature variant)
    Never use spaces in internal reference

  Customer tags:
    VIP — annual spend over $100K
    GOVERNMENT — public sector (special pricing)
    DISTRIBUTOR — reseller (distributor pricing)
    AT-RISK — flagged by sales for churn risk

  Sales order naming:
    Prefix by team: SO/EAST/, SO/WEST/, SO/GOV/
    Custom sequences per team configured in Settings

  Notes format:
    Internal notes: plain text, start with [DATE - INITIALS]
    Customer-facing notes: professional tone, no jargon"

Decision Patterns

Some business decisions follow patterns that are not written in any policy document. The purchasing manager always orders 10% extra for products with high defect rates. The sales team never offers discounts during Q4 because demand is high. The warehouse team expedites orders from certain customers regardless of priority flag. These unwritten rules should be captured in the knowledge base so the AI agent can make decisions that align with how your team actually operates.

Keeping the Knowledge Base Current

A stale knowledge base is worse than no knowledge base — the AI agent will confidently apply outdated rules. Build maintenance into your workflow: when a business rule changes, update the knowledge base as part of the change process. When someone discovers that the AI made a wrong decision because of missing context, add that context to the knowledge base immediately. Over time, the knowledge base becomes a living document that makes the AI agent increasingly accurate.

DeployMonkey AI Knowledge Base

DeployMonkey's AI agent includes a structured knowledge base that you populate with your business rules, processes, and conventions. The agent consults this context before making any decision, ensuring recommendations and actions align with how your specific business operates — not generic Odoo defaults.