AI Features in Odoo 17 & 18: what Bangladesh SMEs should actually use.

When you upload a PDF or image of a vendor bill in Odoo Accounting, the AI OCR service reads the document and pre-fills the vendor name, bill date, due date, currency, and line items (product, quantity, price). You review and confirm instead of typing from scratch.

How it actually works: The document is sent over the internet to Odoo's OCR service (hosted by Odoo S.A.). The AI reads it and returns structured data to your Odoo instance. This requires a live internet connection for each document. The extraction usually completes in 5–20 seconds depending on document complexity and connection speed.

Performance on Bangladesh documents: Good on printed English-language invoices (import suppliers, large local companies). Moderate on mixed-language documents. Poor on handwritten Bengali invoices or non-standard local formats — you will spend as much time correcting as you save typing. Train your accounts team to know which document types work well and which to enter manually.

ROI calculation for Bangladesh: At 100 vendor bills/month, OCR saves approximately 4–6 hours of data entry time per month. At BDT 150/hour labor cost, that is BDT 600–900/month in direct savings — before accounting for error reduction. The Enterprise app cost is recoverable within 2–4 months for a company at this volume.

In Odoo 17 and 18, the email composer (chatter) in Enterprise has an AI-assisted "Generate" button. Click it to generate a draft email based on the context of the current record (customer, sale order, support ticket) and a short prompt you provide.

Where it is useful in Bangladesh: Customer support teams writing repetitive responses to tracking queries, quotation follow-ups, or payment reminders. If your staff sends the same type of email 20+ times per week and struggles with English composition, this feature has genuine value.

The friction: Requires an OpenAI API key and a live internet connection for each email generation. Monthly OpenAI cost for a team of 5–10 people generating emails regularly: approximately $10–30/month (charged in USD, requiring international payment).

My honest take: This feature is useful for teams that produce a high volume of external English emails. For internal operations, it adds complexity without proportional benefit. A well-structured email template in Odoo's existing template system often delivers 80% of the value with zero AI complexity.

Odoo's predictive reordering uses machine learning to automatically adjust reorder points (minimum and maximum stock levels) based on historical consumption patterns. Instead of manually setting "reorder when stock falls below 50 units," the system learns from 90 days of sales/manufacturing orders and adjusts the threshold automatically.

Why this is valuable for Bangladesh manufacturing: Seasonal demand swings are pronounced in Bangladesh — RMG production peaks for export seasons, Eid creates demand spikes for consumer goods, cyclone season disrupts supply chains. Manually adjusting reorder points for 500–2000 product variants across seasons is a full-time job. Predictive reordering does this automatically.

Requirements: Minimum 3 months of historical transaction data. The ML model trains on your server — no internet needed after training. Enterprise only (Inventory app advanced features).

Performance in practice: Works well for products with regular demand patterns. Less accurate for highly seasonal products in the first training cycle — it improves as it accumulates more seasonal data. Do not rely on it as the sole reorder signal for critical raw materials in the first 6 months.

Odoo CRM Enterprise can assign a probability score to each lead/opportunity based on historical conversion patterns — which industries convert, which deal sizes close, which sources generate genuine buyers. Sales reps see a percentage score on each opportunity and can prioritize accordingly.

The Bangladesh reality: This feature requires a significant volume of historical CRM data to be meaningful. If you have fewer than 200 closed opportunities in your CRM history, the model does not have enough signal to produce reliable scores. Most Bangladesh SMEs running a B2B business with a sales team of 3–8 people do not meet this threshold in their first 12–18 months on Odoo.

When to use it: Enable if you have an active outbound sales team, have been on Odoo CRM for at least 12 months, and have logged 300+ opportunities in that period. Otherwise, the scores will be statistically meaningless and may actively mislead your sales team.

Odoo 17 and 18 Enterprise include a general-purpose AI assistant (powered by OpenAI) that can help with: generating product descriptions, drafting terms and conditions, writing notes in the chatter, summarizing long email threads, and building simple custom reports with natural language queries.

The honest assessment: The AI Copilot is impressive in demos. In the day-to-day Bangladesh SME context, it solves problems that are not actually your biggest bottleneck. Your accounts payable backlog is not caused by difficulty writing product descriptions — it is caused by paper-based processes, approval delays, and lack of structured data. The Copilot does not address those root causes.

Where it genuinely helps:

• Marketing teams writing product copy for Odoo eCommerce
• Consultants (like me) drafting knowledge base articles inside Odoo
• Managers creating custom Odoo Discuss bot responses for internal knowledge queries

The practical barriers for Bangladesh: Requires a working OpenAI API key with a credit card that accepts international charges. Monthly cost varies ($5–50 depending on usage). The feature requires a live, fast internet connection — generating a GPT-4o response takes 3–8 seconds on a good connection, 20–45 seconds on a slower mobile connection.

Odoo's Maintenance module combined with IoT Box sensors can theoretically enable predictive maintenance — detecting early signs of machine failure from sensor data before breakdown occurs. This is genuinely valuable technology in mature industrial settings.

Why Bangladesh SMEs should skip this for now: The prerequisite investment is significant — Odoo IoT Box (approximately BDT 15,000–25,000 per machine), compatible sensors calibrated for your specific machinery, and IT staff capable of managing the sensor-to-Odoo data pipeline. Most Bangladesh factories are running machines from multiple decades and manufacturers; sensor compatibility is not standardized.

What to do instead: Implement conventional preventive maintenance scheduling in Odoo's Maintenance module first (see the Odoo Preventive Maintenance guide). Scheduled maintenance with work orders and completion tracking delivers 60–70% of the reliability benefit of predictive AI at a fraction of the cost. Add IoT when your conventional maintenance processes are mature and you have the budget.

Odoo's Live Chat module includes a chatbot builder that can handle predefined conversation flows: qualify leads, answer FAQs, route support tickets, and collect contact information. In Odoo 18, this can be connected to the AI Copilot for more dynamic responses.

When it is worth setting up: If your Odoo website or eCommerce receives more than 20 chat inquiries per day, a chatbot that handles the first-line qualification (Are you looking for X? What is your budget? Your name and email?) saves your sales team significant time. For Bangladesh B2B exporters receiving international buyer inquiries during off-hours (different time zones), a chatbot ensures no inquiry is missed.

Without AI: The rule-based chatbot (no OpenAI key required) handles scripted conversations well. Build the rule-based version first and measure impact before adding AI-powered dynamic responses.