ERP & AI Integration
Connect AI tools to a defined ERP workflow.
Start with a useful task and a clear boundary around what the tool can do. I help assess Odoo AI and MCP integrations using documented ERP workflows, existing access controls and explicit acceptance checks.
Manamatha Sarnaker · Montreal, Canada · Remote consulting for clients worldwide. Scope, availability and time-zone overlap agreed directly.
When this helps
For business and technical teams exploring AI access to Odoo who need practical workflow and permission decisions before development.
An assistant connected to business records needs more than a conversational interface. The team must agree which records it can access, which actions it can perform and how to review the result. Model Context Protocol, or MCP, provides a connection between AI tools and an application's exposed capabilities; the value still depends on the workflow and its controls.
What we can agree to deliver
Task and action scope
A defined user task, the records involved and an explicit set of allowed and blocked actions. We distinguish a read-only reporting need from a workflow that changes ERP records.
Integration and access design
A scoped connection to the agreed Odoo capabilities, with identity and permission requirements checked against the actual version, hosting and chosen tools.
Verification and operating guidance
Checks for expected actions, denied actions and representative user permissions, with documented limitations and a handover appropriate to the agreed workflow.
How the work takes shape
Choose one useful task
Identify the work users need help with and whether AI access offers a useful improvement over an ordinary ERP workflow or report.
Define permission boundaries
Agree who can use the tool, which records are in scope and which actions must remain unavailable. Check identity and existing ERP permissions.
Build the agreed connection
Implement the scoped integration and exercise representative requests. Version and tool compatibility are confirmed during assessment.
Verify and explain limits
Review allowed and blocked behavior with the team. Document the workflow, responsibilities and the boundaries of the delivered integration.
Relevant delivery
Documented MCP use includes task tracking and timesheet entry. Access uses SSO and existing Odoo permissions, with delete operations blocked.
Maison TerritoThe case includes AI-related development work alongside its configurator and designer programme. Its development status is distinct from the delivered TKH workflow.
CasavogueAI work is described as research, separate from the live retail ERP. This establishes an explicit boundary between exploration and production delivery.
This is a workflow integration offer, not a promise of autonomous operation or compatibility with every model or ERP version. Research, development and delivered work are identified separately in the supporting cases.
Before we start
What does MCP add to an Odoo workflow?
It can expose agreed application capabilities to a connected AI tool. The published TKH work covers tasks and timesheets; the appropriate capabilities for your system depend on the user task and access boundaries.
Can an AI tool change or delete records?
That must be decided explicitly. The documented TKH integration blocks delete operations and uses existing permissions. A new scope should define its own allowed actions and verification rather than inherit a blanket permission.
Can we begin with an assessment?
Yes. Describe one user task, your Odoo version and hosting, the chosen AI tools if any, and your access requirements. We can assess feasibility and scope before agreeing implementation.
Start with the problem.
Describe one ERP task, who should be allowed to perform it and which actions must stay unavailable.