Manufacturing
Asking the ERP in plain language: AI over Odoo data
Managers get answers on sales, receivables and stock by asking questions, with every answer drawn from Odoo under their own access rights.
- Customer
- JSP Aqua Pvt Ltd
- Company size
- 20 employees
- Location
- Hooghly, West Bengal, India
- Odoo
- 18.0 · Odoo Community
- Modules
- Sales, Invoicing, Project
- Services
- Customization, Integration
- Minutes Time to answer a routine question
The business
A services company already running Odoo for sales, projects and invoicing, with managers who rely on the finance team for every report.
The challenge
- Routine questions ("Which customers are over 60 days?") waited days for a report.
- Exported spreadsheets went out of date immediately.
- Management wanted AI, but not at the cost of exposing data.
Why it mattered
Decisions were made on stale numbers, and the finance team spent hours each week on ad-hoc reports.
The solution
An AI assistant connected to Odoo through a controlled tool layer. It answers questions by querying Odoo as the asking user, so record rules and access rights still apply.
Architecture
User question → AI assistant → Tool layer (MCP) → Odoo API as the user → Answer with links to records
Standard Odoo and customization
Standard Odoo: sales, invoicing, projects, access rights.
Customized: a curated set of read-only tools and an audit log of questions and the records used.
Integration
The AI provider never receives database credentials; the tool layer holds them and enforces scope.
How we implemented it
- Use-case selection with management
- Data and access review
- Tool design and permissions
- Pilot with two managers
- Feedback and refinement
- Rollout
Results
| Metric | Before | After |
|---|---|---|
| Time to answer a routine question | Days | Minutes |
| Data access | Exported spreadsheets | Live, under the user's rights |
Screens and photos
Have a similar manufacturing challenge?
Tell us how your business works today. We will suggest a practical Odoo approach — no obligation.