Every ERP vendor is currently telling the market that AI is going to run your business for you. As an Odoo implementation partner running the platform ourselves, we wanted to separate what's actually shipping today from what's still roadmap — both in Odoo and across the wider ERP market.
What Odoo actually ships today
Odoo 19 introduced a proper AI application, not a bolt-on chatbot. Ask AI, reachable via Ctrl+K or the AI button in the top corner, handles natural-language queries, summarizes chatter threads, translates messages and drafts text — but it's read-only by default; it can't create a lead or edit a record on its own. AI Fields auto-populate specific fields from a prompt, a quiet but genuinely useful time-saver on things like product descriptions. AI Agents are the more powerful layer: you configure a Topic (what the agent is allowed to do), attach Sources (PDFs, Knowledge articles, weblinks) and write a system prompt, and only then can the agent take real actions like creating CRM leads. Without an assigned Topic, an agent can talk but it can't touch your database — a deliberate guardrail, not a bug. OCR bill digitization has quietly existed for years and tends to get folded into the "AI" story more than it probably deserves.
We've built several of these agents ourselves for internal use, running on a mix of GPT and Gemini models depending on the task — Odoo doesn't train its own model, it routes to OpenAI or Google Gemini, and you choose per agent. On Odoo Online you can let Odoo cover the model cost or bring your own API key; on Odoo.sh or self-hosted, you supply your own key and pay OpenAI or Google directly. Worth flagging for anyone budgeting: the external API that most serious AI integrations depend on is only available on Odoo Online Custom pricing tier, not the standard plans, so check your plan before scoping any integration work.
How the rest of the ERP market compares
The three big alternatives have taken noticeably different paths. SAP's Joule has gone furthest on paper, expanding to more than 40 specialized agents across finance, supply chain, procurement and HR — including agents that autonomously release standard production orders and escalate only exceptions to a human. It's bundled into RISE with SAP at no extra license cost, but it's built for organizations already committed to SAP's scale, and truly autonomous agents (versus ones that just generate a recommendation) are still a minority of the 40. Microsoft's Dynamics 365 takes the most sprawling approach, embedding pre-built agents per app — sales qualification, case management, account reconciliation, time entry and more — all supervised through an "agent feed." It's deeply tied to Microsoft 365 Copilot and Power Platform, but a large share of it is still marked "preview," depends on Azure OpenAI regional availability, and offers no bring-your-own-model option. NetSuite is the most conservative of the three today: its native AI (Text Enhance, Bill Capture, predicted fields) is narrow by its own consultants' admission, and the bigger agentic layer it's promising — NetSuite Next, with Ask Oracle and SuiteAgents — is still rolling out on roughly a twelve-month horizon and hasn't been proven at scale in customized environments.
Against that backdrop, Odoo's AI is smaller in scope right now — there's no 40-agent autonomous production line out of the box. What it offers instead is transparency and control: you know exactly which model is doing the work, you're not locked into one vendor's foundation model, agents are genuinely scoped rather than given broad standing access, and the cost of experimenting is low. For most businesses in Odoo's actual customer base — SMEs and mid-market — that's a more realistic starting point than an enterprise suite built around thousands of skills.
Getting the most out of Odoo's AI
The businesses getting real value aren't switching on every agent at once. They pick one repetitive, well-defined task — drafting follow-up emails, summarizing support tickets, translating customer messages — and prove it out before expanding. They're deliberate about Topics and system prompts, since a tightly scoped agent is easier to trust than a broad one; that scoping is your main safety control, not a hurdle. They decide up front whether they're on Odoo Online with the API included in their plan, or self-hosting with their own OpenAI or Gemini key, because that changes both cost and what's technically possible. They feed agents good source material — clean documentation, well-structured Knowledge articles — because an agent is only as good as what it's given to work from. And they keep a human in the loop for anything touching money, stock levels or customer-facing communication, treating AI output as a strong first draft rather than a final decision.
Odoo isn't going to autonomously run your supply chain the way SAP's marketing suggests Joule might, and it isn't wrapping every app in Copilot the way Microsoft is. What it is doing is putting genuinely useful, cost-controllable AI directly into the hands of businesses that don't have an enterprise IT budget — which is exactly where it should be focusing.
If you're evaluating ERP platforms and want AI capability factored into the decision, we're happy to walk you through what's real versus what's roadmap.