Connecting Your Systems

We connect what you have – we do not replace it. In mid-sized companies that is typically an ERP (SAP Business One, Microsoft Dynamics, Sage, industry-specific systems), a CRM, an online shop, a document management system, DATEV and Microsoft 365. Almost all of them have APIs, some only exports. We can work with both.

The result is one shared state: a customer, an article, an order exists once, and every system sees the same version. What is typed three times today is entered once.

Before an AI can suggest quotes or check invoices, it has to be clear which data is correct. We find duplicates, unify spellings, fill missing fields from other sources and build in the rules that keep it clean. Documents – contracts, delivery notes, emails – are indexed so they can be searched and analysed.

The data source register from the opportunity assessment is the starting point. It says which data the first use case needs – and only that data gets prepared first.

A language model only becomes useful when it may access your data and do something in your systems. We build that connection – using the open standard MCP (Model Context Protocol), through which a model can read data and trigger actions without tying you to one vendor.

An example: an assistant reads quote requests from the sales inbox, pulls the customer data from the CRM and the prices from the ERP, drafts a quote – and a person approves it. The approval is not a concession, it is part of the design: every action that touches money or customers stays in human hands.

The models run in European data centres, or on your own hardware if you prefer.

Interfaces break when a system gets an update. That is why we monitor every connection: errors are detected before anyone in the business notices, and fixed before they corrupt data. Monitoring, logs and updates are part of the monthly flat rate, not an extra.