What changed
From Odoo's own release notes:
- Agents inside automations. Agents can be triggered by automated and scheduled actions, and you can review the automation chats to see how an agent handled a task.
- Agents that act. They can create and update records, answer questions about a file and work from voice dictation.
- MCP. You can connect to your database via MCP, the Model Context Protocol: the open standard AI assistants use to read data from other systems.
- Accounting. You can ask the AI assistant for insights based on your accounting data, pay bills from Odoo through a payment initiation interface, and get better duplicate detection and bill line prediction.
The catch: credits
The release notes are blunt: “All AI features now need credits, which can be bought via IAP.” They don't publish prices. If agents start running on a schedule, AI becomes a running cost that grows with use, so it needs a budget and someone who watches it, like any other supplier.
What it means for a team of 15 to 250
Jobs that used to need a custom module or a chain of outside tools, like chasing an order, sorting a ticket or flagging a duplicate bill, can increasingly be set up inside Odoo itself. That's good news: fewer moving parts.
Two things don't change. Your own business rules still need building, and an agent that writes to your records needs the same guardrails as a new employee: limited access rights, a log of what it did, and a person who signs off on anything that moves money or stock.
Before you upgrade
- List the custom and App Store modules you run. A database with custom modules can't move to a new major version until those modules are ported.
- Test the upgrade in a copy of your database, never in production.
- Pick one or two agent jobs to start with, each with an owner and a monthly credit budget.
- Decide where a person must approve before an agent's action counts.
Planning a move to Odoo 20, or an agent job? We build and port custom Odoo modules at a fixed price, in your repository.
Custom Odoo