In recent months, we've seen many people talking about integrating artificial intelligence technologies into real-world projects. Beyond the hype, at Dixmit we're mostly interested in how AI can actually solve specific problems within business workflows. And as always, we approach it from a free software perspective, with special attention to the work of the OCA community.
What is the OCA doing about it?
As we mentioned, a lot of people have been discussing Odoo and OCA in the context of AI, but overall, we haven’t seen much that's truly integrated within the system itself. And above all, nothing within the Open Source community. That’s why at Dixmit, we’ve decided to do our part and stir things up in the community by creating the first OCA repository for AI and a general-purpose module.
Thanks to this, we’re already seeing some people starting to collaborate, and we hope it will keep growing gradually. In the end, it’s a very relevant and interesting topic.
We’ve designed a module that allows integration with external systems, which are the ones that actually host and run the AI agents. This approach is based on several reasons:
- Odoo is a transactional system, not an AI engine. It makes more sense to delegate to specialized systems.
- There are very interesting Open Source external systems that allow for easy integration using webhooks.
- We personally believe this field evolves so quickly that it’s unclear whether the community can keep up with that pace. By outsourcing the AI layer, we can use the right tool for each client's needs.
What kinds of tasks are we automating?
We believe current AI systems are ideal for automating repetitive tasks. Here are a few examples:
- Automatic ticket classification in the Helpdesk module, using NLP models to understand message content and assign it to the correct team or category.
- Smart field filling in CRM or sales forms, based on the original input.
- Data extraction from invoice PDFs, going beyond traditional OCR.
Our experience: AI is useful, but we keep our feet on the ground
While it’s an exciting area, it’s important to stay aware of its implications. That’s why we always keep in mind its limitations and challenges:
- Privacy and legal compliance: Many companies don’t want to send data to external APIs. Local models are a solution, but they require more infrastructure.
- Model updates: How can we retrain models without disrupting business processes? We need to think about maintainable pipelines.
- User interface: Integrating AI isn’t just a backend concern. The user experience must be smooth and transparent.
So, what’s next?
We’d love to see more initiatives within OCA to standardize how AI models are integrated into Odoo. Ideally, we’d have a range of examples with different structures and integration systems. On our side, we’ll keep contributing as much as possible, because we see this as a highly relevant and useful topic for the coming years.