Frequently asked questions about connecting AI
(01)Can you integrate AI into our existing SAP system?
Yes. We connect AI via RFC modules, OData services or the SAP Business Technology Platform. Typical use cases are automated ordering processes, intelligent supplier selection and AI-assisted inventory forecasts. The solution runs alongside your SAP operations in its own layer, so no changes to your customising are needed and updates to your SAP system remain unaffected.
(02)Do we have to replace our existing systems?
No, that is exactly the core idea. We add AI capabilities to your existing systems instead of replacing them. Your employees keep working in their familiar tools; the AI works in the background or appears there as an additional function. This saves migrations, retraining and licence changes and significantly lowers the risk of the entire project.
(03)What is the Model Context Protocol (MCP) and why do you use it?
The Model Context Protocol is an open standard from Anthropic for communication between AI models and external data sources or tools. It lets models access approved company data in a structured way and carry out defined actions. We use MCP because it simplifies permissions, logging and switching models, which keeps integrations maintainable in the long term.
(04)How long does a typical integration take?
The goal is a standard integration with one system, such as a CRM plus an AI assistant, in about 4–6 weeks. More complex projects with several systems and two-way data flows usually take 6–10 weeks. We aim for a working proof of concept after about two weeks, so you can check early with real data whether the approach holds up.
(05)How do you prevent the integration from disrupting ongoing operations?
Through strict separation: the AI runs in its own layer and talks to your systems only through defined APIs. Feature flags allow gradual activation, and rollback mechanisms restore the previous state at any time. Before go-live we test thoroughly in a staging environment with realistic data and agreed test cases.
(06)Which data flows between our systems and the AI?
You define this together with us in the integration design. Typical data flows are documents and texts for analysis, customer data for personalisation, transaction data for forecasts and system events for automated reactions. Every data flow is documented, implemented in a GDPR-compliant way and secured by access controls; sensitive fields can be masked before processing.
(07)What happens if the AI API is temporarily unavailable?
Our integrations include robust error handling: a circuit breaker prevents failures of an AI API from affecting your core systems. For critical workflows we use fallback logic, queues and retries, and on request a second model as a backup. This keeps your business processes running even during temporary outages.
(08)Can you also integrate legacy systems without modern APIs?
Yes. Even legacy systems that only communicate via file interfaces, SOAP services or database views can be connected. We build adapters that translate old protocols into modern API formats and keep data in sync in both directions. This way older applications also benefit from AI functions without you having to replace them first.
(09)What does it cost to connect AI to existing systems?
Costs depend on the number and type of systems, data quality and security requirements; ongoing API costs depend on the model and volume. Reusable connectors noticeably reduce the effort. We quote per project: fixed price after scoping, proposal within 48 hours. For an initial assessment, a fixed-price workshop is a good starting point.