Integration, data and security: bringing AI into an existing IT stack.
AI enters systems that already exist, with their data, rules and constraints. The question is therefore not only what a model can do, but how an agent connects to an ERP, an inbox or an in-house system. This section explores possible architectures, APIs and SFTP files, security, product data and the limits of native ERP AI. We focus on one simple question: how can you apply AI without rebuilding your own information system?
Articles
AI, ERP & IT · 22 October 2026
Building with AI in house has become a necessity, but order entry is a hard first project. When a single customer always sends you the same file, a home-made flow is enough. When dozens of customers each order their own way and an error ends up in a delivery, buy the reading engine and keep your business rules. The cost is not the prototype, it is running it.
Customer stories
“The only question worth asking is: what am I risking? Nothing. The scope I am starting from is narrow.”
Jean-François Le RudulierGroup CIO · Groupe Hedis“Before, you had to fit into the mould of the big, rigid software. With Catalog, we ask, and a few weeks later it is done.”
Mathieu RognonSupply Chain Director · CF GroupAgents
Extracts order lines from emails and records them into the ERP, with full traceability. 70% time saved on data entry.
See the agentChecks every supplier acknowledgment against its purchase order and shows your team only what differs.
See the agentThe diagnostic
48 hours. One PST file. No ERP access. A report with figures on the orders that arrive by email, the quotes left without follow-up and the requests left unanswered over your last 90 days.