Odoo 20 AI Features: What They Actually Do — and Where They Help

What changed with AI in Odoo 20?

Odoo 19 introduced AI agents that could answer questions from your documents and query your own records, along with AI-filled fields, natural-language search and AI drafting. Odoo 20 builds on that foundation. According to Odoo’s Odoo 20 release notes, agents now do more of the work themselves:

The shift matters because an agent that can change records is no longer a writing assistant. It is part of a business process — so it needs the same things any process step needs: a clear job, reliable inputs, defined permissions and someone accountable for the result.

Where AI fits in day-to-day work

Odoo’s release notes describe AI mostly as a platform capability, not as a separate feature in every app. That is worth knowing before you go looking for “AI in purchasing” or “AI in manufacturing”: in several areas, the most useful Odoo 20 improvements are data-driven suggestions and workflow changes rather than AI.

Sales and CRM

There is no CRM-specific AI feature in the Odoo 20 release notes. The practical AI use is an agent working with sales records — creating or updating them from a written request, or building views filtered by period. For many sales teams, Odoo 20’s non-AI changes may save more time: editing the margin on an order line to recompute the price, reusable quotation sections and templates, and lead auto-assignment rotation per salesperson.

Purchasing and inventory

The most useful “smart” change for buyers isn’t labelled AI. Odoo 20 suggests minimum and maximum levels for reordering rules based on demand history, the days of coverage you want and how often you order. Agents can help with repetitive record entry — worth testing on high-volume, low-judgment work, with a person reviewing what was created.

Manufacturing

Odoo 20’s release notes list no manufacturing-specific AI. The changes that matter on the shop floor are execution changes — continuous production, splitting ongoing manufacturing orders and a drag-and-drop work order Gantt. We cover them in Odoo 20 for Manufacturing.

Accounting and finance

Finance gets the most explicit AI feature outside the agents themselves: you can ask the AI assistant for insights into the company’s finances, based solely on the accounting data in the database. Alongside it, Odoo 20 adds automation that reduces manual entry without being called AI — bill lines suggested from bill history, configurable matching rules for reconciliation and automatic invoice reminders.

Service, documents and knowledge work

Helpdesk can detect similar tickets with AI and suggest answers from them. Documents can rename uploaded files with AI. And because agents can answer questions about a file you are previewing, reading long documents — contracts, specifications, instructions — becomes a natural place to start.

Website, eCommerce and marketing

This is where Odoo 20’s AI is most visible: an AI Website Assistant that can create pages, forms and menu changes, AI-assisted product editing in eCommerce (images, names, prices and descriptions), and AI help with marketing automation campaigns and social posts. Useful for speed — but everything it publishes still needs an owner who checks it.

Practical things to know before you turn it on

• Credits: every AI feature now consumes credits bought through IAP. Treat AI like any usage-based cost — decide where it earns its keep.
• Retention: conversations with agents are kept for up to 30 days after the last message. Agree what people should and shouldn’t paste into a chat.
• Permissions: agents ask for permission and confirmation, and an agent that reaches its tool-call limit asks before continuing. Design where those approval points sit — don’t leave it to defaults.
• Models: you choose an AI provider and Odoo selects the model for the task, so the controls that matter are on your side — skills, data and permissions.

What AI doesn’t fix

AI in Odoo works on the records and processes you already have. That makes it a multiplier — including of the things that don’t work yet.

• It doesn’t fix bad master data. An agent that creates purchase records from inconsistent product data creates inconsistent purchase records, faster.
• It doesn’t decide who owns a broken workflow. If nobody owns the hand-off between sales and production today, an agent in the middle doesn’t create an owner.
• It doesn’t make an inconsistent process consistent. If three people handle the same exception three different ways, there is no rule for an agent to follow.
• It doesn’t replace configuration. Reordering rules, routes, approval steps and access rights still have to be set up correctly — AI uses them; it doesn’t design them.

AI is most valuable where the transaction, the process and the data underneath it are already sound.

Where should you start?

Start small and specific. The best first AI use in Odoo is usually a narrow, repetitive task that people already do the same way every time — not the most impressive demo you have seen.

1. Pick one task that scores well on every row above.
2. Write down the rule the agent should follow — if you can’t, the process isn’t ready.
3. Decide what the agent may do alone and what needs a person’s confirmation.
4. Run it inside an automated action with a narrow skill, and read the automation chats for the first weeks.
5. Only then widen it — or move on to the next task.

Want the full picture of the release?Odoo 20: what’s new and what it means for your business→