Skip to Content
AI
Create a good knowledge for your agents
George Agent User Reference

By: OdooBot

How to Make Your AI Agent Smarter: A Practical Guide for Odoo's Knowledge Module

Think of your AI agent as a highly dedicated assistant with selective memory. Give it well-organized, clear information, and it will help you brilliantly. But hand it a chaotic mess of poorly structured documents... well, even the world's best assistant would get confused. This guide will help you organize your information so your AI agent can truly shine.

Why does organization matter?

When the agent searches for information to answer your questions, it doesn't read everything from start to finish like a human would. Instead, it breaks content into chunks and searches for the most relevant ones for your query. If these chunks are poorly organized or confusing, the answers you get will be equally confusing. It's like trying to find a specific recipe in a cookbook where someone tore out all the pages and glued them back in random order.

💡 Fun fact: Your AI agent uses a technique called RAG (Retrieval Augmented Generation) that combines intelligent search with response generation. If you're curious about how RAG works, you'll find it fascinating how AI can "remember" relevant information right when it needs it.

1. Use descriptive and specific titles

The title of each document or section should say exactly what the content is about. Think of titles like labels on filing cabinet drawers: if the label says "miscellaneous stuff," it'll be hard to find what you're looking for. But if it says "Supplier Invoices - January 2024," you'll know exactly what's inside.

✅ Recommended approach ❌ Approach to avoid 💭 Why it works better
"Product return procedure for items under warranty" "Returns" The specific title allows the agent to identify exactly when to apply this procedure
"2024 Vacation Policy - Sales Department" "Vacation info" Includes year and department, avoiding confusion with old policies or those from other departments
"Inventory module setup manual - Version 16.0" "Odoo manual" Specifies module and version, crucial when you have multiple manuals

2. Write complete and contextualized information

Each section of your document should be understandable on its own, without needing to read other parts. This is crucial because the agent might retrieve only that specific chunk when searching for information. It's like writing each page thinking it might be the only one someone reads, so it must contain all the necessary information to be useful.

✅ Recommended approach ❌ Approach to avoid 💭 Why it works better
"The invoicing process in Odoo begins when the user confirms a sales order. To do this, go to the Sales module, select the corresponding order, and click 'Confirm'. This will automatically generate a draft invoice that can be found in Accounting > Customers > Invoices." "To invoice, confirm the order then go to invoices." The complete text provides context, exact location, and process flow, allowing the agent to give precise instructions
💡 Pro tip: The technique of dividing information into self-contained chunks is called "chunking" in the AI world. Chunking experts recommend each chunk be between 200-500 words for optimal balance between context and precision.

3. Maintain consistency in format and terminology

If you call your buyers "customers" in one document and "clients" or "purchasers" in another, the agent might not understand you're referring to the same thing. Similarly, if you use different formats for the same information, it will be harder for the agent to find patterns and give coherent responses. Consistency is key to making everything work like clockwork.

✅ Recommended approach ❌ Approach to avoid 💭 Why it works better
Always use:
• "Customer" for buyers
• "Product" for items on sale
• Date format: MM/DD/YYYY
• Amounts: $1,234.56
Mix terms:
• Customer/client/buyer
• Product/article/item
• 01-01-24 / January 1, 2024
• $1234.56 / USD 1,234.56
Consistent terminology avoids ambiguity and allows the agent to understand that different documents are talking about the same thing

4. Organize information hierarchically

A good hierarchical structure is like a map that guides the agent to the right information. Start with general categories and delve into specific details. This allows the agent to understand the relationship between different pieces of information and efficiently navigate to what it needs.

5. Update and remove obsolete information

Having outdated information is worse than having no information at all. If you keep the 2019 vacation policy alongside the 2024 one, the agent might get confused and provide incorrect information. It's essential to keep your knowledge base as fresh as today's bread, clearly removing or archiving what no longer applies.

✅ Good practice ❌ Problematic practice 💭 Impact on agent
Clearly mark obsolete documents:
"[OBSOLETE - See 2024 version] 2023 Vacation Policy"
Keep multiple versions without indicating which is current The agent can clearly identify which information to use and which to ignore
Delete or archive procedures no longer in use Accumulate years of unreviewed procedures Reduces the possibility of the agent retrieving incorrect or confusing information

6. Include examples and practical cases

Concrete examples are like lighthouses in the dark for the AI agent. When you include real cases of how to apply a policy or procedure, the agent can better understand the context and provide more precise and useful answers. Don't just explain what to do; show how to do it with real situations your users might encounter.

Content type Recommended example Benefit
Discount policy "Example: If a customer buys 100 units of product A, apply a 10% discount. In Odoo, go to the sales order, add the product, adjust the quantity to 100, and the system will automatically apply the configured discount." The agent can guide step-by-step in specific situations
Refund process "Case study: Mary returned a defective laptop on 03/15/2024. Process followed: 1) Created ticket in Support, 2) Attached invoice #INV-2024-1234, 3) Refund processed within 48 hours via bank transfer." Provides a clear model of how to handle similar situations
💡 For the tech-savvy: Using examples significantly improves what's known as "semantic search" in RAG systems. Studies on semantic chunking show that chunks with concrete examples have up to 40% better retrieval than purely theoretical ones.

7. Avoid excessive fragmentation

Dividing information into pieces that are too small is like trying to read a book where each page is in a different room. While it's important that each section is understandable on its own, you also need to maintain enough context for the agent to understand the complete picture. Balance is key: neither War and Peace on a single page, nor one haiku per document.

✅ Balanced fragmentation ❌ Too fragmented 💭 Reason
One document per complete process:
"Complete employee onboarding process" (3-5 pages with all steps)
One document per step:
- "Step 1: Send email"
- "Step 2: Prepare desk"
- "Step 3: Create user"
Maintains the context of the complete process, allowing the agent to understand the sequence and relationships

8. Test and continuously improve

The best way to know if your organization works is by testing it. Ask the agent frequent questions and observe if the answers are accurate and useful. If you notice the agent gets confused with certain topics or gives incomplete answers, it's time to review how that information is organized. Remember: your knowledge base is a living organism that must evolve with your business.

🔍 Debugging tip: If the agent gives confusing answers, check if you have contradictory information or ambiguous titles. RAG optimization experts suggest keeping a log of failing questions to identify improvement patterns.

9. Create reusable templates

Templates are your best allies for maintaining consistency. When everyone on your team uses the same format to document similar processes, the agent learns to recognize patterns and can give better answers. Additionally, templates save time and ensure important information isn't forgotten.

Document type Suggested template structure Benefit
Operating procedure 1. Procedure objective
2. Scope and application
3. Prerequisites
4. Detailed steps
5. Verification and control
6. Common troubleshooting
Ensures complete and organized information
Business policy 1. Policy purpose
2. Who it applies to
3. Key definitions
4. Rules and guidelines
5. Exceptions
6. Practical examples
7. Effective date
Clear structure that facilitates searching for specific information

Final thoughts

Optimizing your knowledge base for the AI agent isn't black magic nor does it require a PhD in computer science. It's simply about applying common sense and good organizational practices. Always think from the agent's perspective: can it understand this information without additional context? Is it clear what each term refers to? Are there ambiguities that could cause confusion?

Remember that the time you invest today in organizing your information well will translate into more accurate responses, more satisfied users, and less time wasted on clarifications. Your AI agent is only as good as the information you provide it, so give it the best tools to shine.

🚀 Next step: Start with a small area of your knowledge base. Apply these principles to a few documents and observe the improvement in the agent's responses. Once you see the results, it will be easier to motivate your team to apply these practices throughout the organization.

Have questions or suggestions? Odoo's Knowledge module and your AI agent are here to grow with you.
Happy organizing! 📚✨

Accounting with AI

Open source accounting software, powered by AI

A full general ledger with AI invoice capture, automatic bank reconciliation, and real-time reports — in one system. This is open source accounting software built on Odoo, connected to your finance ERP, where an AI assistant does the data entry and you review.

Start free Talk to an expert

AI-powered accounting · Open source · Free plan to start

See how AI automates your accounting

Open source accounting software dashboard with AI invoice capture

Accounting that runs your whole finance ecosystem

Designed for finance teams that want a real ledger with automation built in — not generic ERP promises, and not a closed system you can't control.

Bank reconciliation screen matching transactions with smart rules

Automatic bank reconciliation

Sync bank statements and match transactions with smart rules.

Close your books faster with fewer errors. The system suggests matches based on your history and learns your patterns, and you can match payments during reconciliation without leaving accounting.

Start free

Real-time financial reporting

Profit and loss, balance sheet, and cash flow — always up to date, no spreadsheets needed. Every posted invoice, expense and payment updates your reports the moment it happens.

Start free

Real-time financial report showing payments made and pending
Currency list with exchange rates and latest updates for multi-currency accounting

Multi-company and multi-currency

Manage multiple entities and currencies from one accounting system, with automatic exchange-rate updates and consolidated reporting included — no separate tools, no manual consolidation.

Start free

How Emma automates your accounting in Odoo — in 3 steps

Watch Emma set up your accounting, capture invoice data, and help reconcile your bank — all inside Odoo.

1

Ask Emma

She sets up your accounting and chart of accounts.

2

Upload a bill

AI captures the vendor, lines, taxes and amounts automatically.

3

Review and approve

You confirm, and Emma learns your patterns for next time.

See the full accounting tutorial with Emma:

Video tutorial cover: Emma automating accounting in Odoo

AI invoice & bill capture, no manual data entry

Upload or email a PDF and AI reads it for you — then you review instead of retype.

OCR + AI extraction

Vendor, dates, taxes, totals and line items are detected and a draft bill is created automatically.

Upload or email it in

Drag a file in, or forward supplier invoices to a journal email alias and they become draft bills.

It learns as you go

Your corrections train the model, so recognition improves and reconciliation gets faster over time.

AI document capture uses prepaid credits, so high invoice volumes have a small per-scan cost.

Let AI keep your books, start free

Free plan, no credit card. And Emma helps you set it up from the first click.

See plans & pricing Talk to an expert

Built for the way your business does accounting

Whether you keep the books yourself or run a finance team, the work is the same — minus the typing.

Bookkeepers

Let AI handle data entry and reconciliation so you spend time on review, not retyping bills.

Controllers & CFOs

Real-time P&L, balance sheet and cash flow across companies, without waiting for month-end.

Growing businesses

Start on a free plan and scale into a full ERP as you add invoicing, expenses and payments.

How Binhex compares to QuickBooks, Xero and Sage

The mainstream tools are polished and proven. The difference is openness, AI included in the platform, and one connected ERP instead of add-ons.

What matters

QuickBooks / Xero / Sage

Binhex Accounting

Open source & data ownership Proprietary Open source, your stack
AI invoice capture included Often add-on / higher tier Built in (credit-based)
Part of a full ERP Standalone + integrations One connected system
Multi-company consolidation Higher plans Included
Mature, polished UX Yes Yes
Pricing model Per company / tiered Per user, free plan to start

Odoo vs QuickBooks Start free

Everything included in your accounting workspace

 General ledger and chart of accounts

 Customer and vendor payments

 Bank integrations and imports

 Payment providers (Stripe, PayPal)

 Audit trail and traceability

 Configurable taxes and withholdings

 Electronic invoicing (by country)

 Customizable real-time reports

Finance professional reviewing real-time accounting on a phone

Connect accounting with the rest of your finance apps

Accounting is the financial backbone of your ERP. Every transaction flows automatically.

Invoicing app icon

Invoicing

Invoices post straight to accounting as revenue entries.

Expenses app icon

Expenses

Employee spending is booked to the ledger automatically.

Payment app icon

Payment

Customer and vendor payments matched during reconciliation.

Frequently Asked Questions

Open source accounting software FAQ

Open source accounting software is built on a codebase you can inspect, host and extend, rather than a closed proprietary system. Binhex accounting runs on Odoo, so you keep control of your data and your stack while getting a full general ledger, reporting and AI automation inside one ERP.

Both. You can run it in the cloud managed by Binhex, or self-host it because it is open source. Either way you work in the same web interface from any device, and your books stay connected to the rest of your ERP.

Odoo uses OCR and AI to read scanned or PDF vendor bills and invoices and create draft accounting entries with the vendor, amounts, taxes and lines already filled in. It also suggests bank-reconciliation matches and learns from your corrections, so the more you use it the less manual entry you do. You review and approve rather than type.

Yes. You can import bank statements or connect bank feeds, and the system matches transactions to your records with smart reconciliation rules. Recurring patterns like bank fees or internal transfers can be recognized and posted automatically. Available bank connections depend on your country.

Yes. Manage multiple entities and currencies from one accounting system, with automatic exchange-rate updates and consolidated reporting across companies — no separate tools or manual consolidation spreadsheets.

Yes. Taxes are configurable per jurisdiction so they calculate automatically, and electronic invoicing is supported in many countries. Specific tax and compliance obligations vary by country, so we help you configure it for yours during setup. This is general information, not legal advice.

Yes. Accounting is the financial backbone of the ERP: invoices post straight to the ledger, employee expenses are booked automatically, and customer and vendor payments are matched during reconciliation. Everything posts to the same books in real time.

You pay per user on the plan you choose, and there is a free plan to start. The AI document digitization uses prepaid credits, so very high invoice volumes have a small per-scan cost. The free trial lets you set it up and test it with your team before committing. Start free to try it in your own workspace.

Start free