Development

AI applied to your ERP, your documents and your support

Agents that check stock, prices and orders, and create orders with human approval. Invoice and quote extraction with multimodal AI. Multi-provider with fallback, and RAG over your documents.

What usually happens

The same questions every day

Your team keeps answering whether something is in stock, what it costs or where an order is.

Documents typed in by hand

Vendor invoices and quotes that someone copies into the system line by line.

Locked into one AI provider

If the price changes or the service goes down, everything that relies on it stops.

What’s included

Applied AI

01

Agents on top of your ERP

They check stock, prices and order status in Odoo or other systems, with scoped permissions.

02

Actions with approval

The agent drafts the order, and a person reviews and confirms it before it is created.

03

Document extraction

Invoices and quotes in PDF or image form turned into structured data.

04

Multi-provider with fallback

Anthropic, OpenAI-compatible, Groq, OpenRouter or Ollama, switching automatically if one fails.

05

RAG with pgvector

Answers grounded in your manuals, catalogs and policies, stored in PostgreSQL.

06

Traces and usage

A log of every request, the tools used and usage per provider.

Tools, not just chat

Our agents call real functions in your system. We have built an AI connector for Odoo and analyzed an MCP server to expose Odoo to assistants.

  • Stock and price lookups
  • Order tracking
  • Orders with human approval
  • Reading invoices and quotes
FAQ

FAQ

Is my data used to train models?

We pick providers whose API terms exclude training on your data, or local models with Ollama if you want nothing to leave your server.

Can the agent make mistakes?

Yes. That is why actions that change data go through human approval, and the agent only reaches the tools and data we define.

Which AI provider do you use?

Whichever fits best on cost, quality and privacy for each task. Because it is multi-provider, you can switch without rewriting the integration.

How long does it take?

A proof of concept on one specific case usually takes 2 to 4 weeks. Taking it to production with permissions, tracing and approvals, another 4 to 10 weeks.

Contact

Is your Odoo or store not quite fitting how you work?

Write to us. We reply within one business day with next steps and, if needed, a discovery call.