AI Integration Into Existing IT Landscapes - SAP, DATEV, Workday
How do you integrate AI agents into existing enterprise systems? Integration Layer, API decoupling, booking logic separated from the export layer. No parallel system.
The Principle: Extend, Not Replace
AI agents do not replace existing systems. They extend them. SAP remains SAP. DATEV remains DATEV. Workday remains Workday. The agent sits in between: It reads data, makes decisions, and writes results back.
This principle is critical for adoption in enterprise environments. CIOs do not invest in a parallel world. They invest in a layer that makes their existing IT landscape smarter.
The Integration Layer
The Integration Layer in the Gosign reference architecture decouples agent logic from the target system. This has a concrete advantage: When the target system changes, only the export layer changes, not the agent.
SAP FI/CO and S/4HANA: Booking proposals in SAP format. Cost center and profit center logic. Integration via RFC, REST API, or SAP BTP.
DATEV: Booking proposals in DATEV format. SKR03 and SKR04. Tenant-specific charts of accounts. Export via DATEV interface.
Workday: HR data, payroll information, organizational structure. Integration via Workday REST API.
SuccessFactors: Employee data, performance data, compensation. Integration via SAP SuccessFactors API.
Microsoft Graph: SharePoint, Teams, Outlook. For Knowledge Agents that access company documents.
Decoupling: Booking Logic vs. Export Layer
The separation of booking logic and export layer is a central architectural principle. The booking logic, which account, which cost center, which tax rate, is implemented in the Decision Layer. The export layer, how the result is passed to DATEV or SAP, is in the Integration Layer.
When a client migrates from DATEV to SAP, the export layer changes. The booking logic, the rule sets, the agent, everything else stays the same. This significantly reduces migration effort.
No Parallel System
Agents become part of the existing IT governance, not a new parallel world. This means:
- Agents use existing authentication (Azure Entra ID, LDAP)
- Agents run in the existing cloud environment (Azure, GCP)
- Agents are subject to existing security policies
- Agents are integrated into existing monitoring systems
For the CIO, an AI agent is another component in the existing architecture, not a separate system with its own rules.
More on this: Reference Architecture
Book a consultation - We will show you the integration into your existing IT landscape.