Deploy

A reactive agent responds only when a specific event, trigger or prompt occurs. A proactive agent anticipates needs and acts on its own using patterns or predictions, within the permissions you’ve approved. For example, a reactive agent answers a question about this month’s bill. A proactive agent can notice that your spending is trending upwards and warn you about a tight budget before the bill arrives.

Most companies have a first workflow in production within the first month. Timelines depend primarily on how quickly a client provides data access and approvals.

Typically, no. The only time tools are introduced, removed or changed is when it makes your work easier. No changes are made until discussed and approved by you.

Great question. We can still build your AI layer today. If you use tools other than Odoo, we can start there, and if we realise there’s a structural or data issue due to your Odoo implementation, we’ll call it out and scope it early.

No. Your integrator maintains and develops Odoo. Basalt handles the AI layer on top: your workflows across Odoo modules and other software. Basalt works with the integrator where scopes meet.

Basalt accesses the records needed for the workflow you approve, such as orders, stock movements or invoices, within the permissions you grant. When an agent uses AI, the relevant task data is shared with the AI services we identify before the agent runs. Basalt also keeps the account details, workspace settings, conversations, reports and activity history needed to operate your workspace.

Odoo remains your system of record. Basalt reads the records needed for a task rather than creating a full copy of your Odoo database. Selected business information can still appear in workspace conversations, reports and logs. Connection credentials are stored encrypted. Before deployment, we confirm the hosting and processing locations and the service providers involved in your setup, including the AI providers.

Your Odoo records remain subject to your existing retention policies. Basalt’s workspace history and operational logs, and any data retained by AI providers, have separate retention periods. Before deployment, we confirm the applicable periods and deletion arrangements for your setup, including what happens when you stop working with us.

Only what you have explicitly authorised. Initially, controls remain tight: the agent can only read and write drafts. As you build trust and confidence, you can expand the agent’s permissions to send, update and delete.

This will become clear and visible in our Measure phase. If this is the case, we’ll fix the workflow, change the priority or stop expanding it. An agent without a result does not become an endless project.