Alpino AI

Manufacturing · South Tyrol

AI platform for a manufacturer: knowledge, sales and planning on one data basis

Part of: Data & analytics

Also relevant for: ERP & integration

The starting situation

Technical knowledge sits in data sheets, certificates and manuals, customer knowledge in the ERP and mailboxes, planning often in Excel. Every question about a product, customer or stock level means a separate search.

How we solve it

Alpino AI builds a platform with its own data basis and three modules: knowledge search with sources for engineering, customer and order analysis for sales, and forecasting and production planning for the supply chain. Connections to the ERP, file server and email run through a shared layer.

What you get

Access to the data that knows roles and permissions and backs every answer with a source. New modules build on the basis without rebuilding connections. The decision stays with people.

Do you recognise a similar process in your company?

If this use case feels close to your daily work, we can assess data availability, effort and a realistic first MVP together.

Why AI projects work →

Starting point

A manufacturer that also sells third-party goods manages many articles. The related knowledge sits in different places: technical data sheets, certificates and manuals on the file server, customers, revenue and orders in the ERP, correspondence in mailboxes, production planning partly in Excel. The ERP itself has been in use for many years.

Anyone who wants to answer a technical customer enquiry, assess a customer before a meeting or plan demand for the coming months has to search somewhere else each time.

The solution

The platform consists of a basis and three modules working on the same data:

  • Basis: a database of its own alongside the ERP, so that the AI does not access production data directly, plus roles and permissions, a log of every query, connections to the ERP, file server and Exchange, and the user interface.
  • Technical AI: knowledge search across data sheets, certificates, manuals and reports, every answer with a source. It supports product selection and technical customer enquiries and is intended for internal use.
  • Sales AI: analysis of customers, revenue and orders, linking emails with the ERP, pointers to opportunities, problems and follow-ups, plus briefings and email drafts.
  • Supply Chain & Production Planning AI: sales forecast, demand calculation, scenarios and bottlenecks. The production plan is first created by fixed rules; mathematical optimisation is a later stage.

Process and pilot

The basis is built first and the modules one after another on top of it. Each module goes through a test and pilot phase with a few key users before it goes live for everyone. Beforehand comes a data check: which data exists, how complete is it, and where are capacities, routings and setup times? For planning this is the most important question, because without such data only a range can be calculated, not a fixed plan. Training for users is part of it.

Forecasts are calculated at product-family level and then broken down to articles. Later, it can be measured how well the forecast matches actual demand and how much search time is saved on technical enquiries.

What stays with people

For each process, the company decides whether the AI only suggests or acts by itself. We recommend starting with approval and automating later. Technical statements come with a source, figures from the ERP. Production plan, purchasing and customer contact are still decided by the responsible people. Which mailboxes are analysed is determined by the company itself.

Connection to existing systems

The ERP is read; writing happens only through the interface provided by the manufacturer. Emails come from Microsoft 365, documents from the file server. An abstraction layer separates the AI from the ERP, so that a later change of system only requires swapping the connection. Language models run in the EU cloud; local models are prepared for sensitive tasks. More on this part on our ERP integration page.

Transferable to other companies

The pattern suits companies with many articles, an ERP and distributed knowledge: manufacturers and traders with their own production. A case on the forecasting part is described in Bread forecasting, with background in the article Forecasting in purchasing. An overview of the service is on our data and analytics page.

If you would like to check which data in your company can support a shared basis, book a 30-minute initial conversation.

This use case is based on a real client project. Sector and region are named, the company itself remains anonymous.