SAP LeanIX Assistant in Copilot

SAP LeanIX Assistant in Copilot

| Michael Sukachev |

LeanIX Copilot

YouTube

Anchor

Spotify

Apple

Episode #310

Introduction

In episode 310 of our SAP on Azure video podcast we talk about SAP LeanIX with Microsoft Teams.

Enterprise Architecture repositories are incredibly valuable — at least in theory. But in practice, many of them become hard-to-access systems that only a few specialists use, and even fewer people keep up to date. Over time, the data gets stale, the trust disappears, and the repository becomes more of a museum than a working tool.

Today we want to look at a different approach. Together with Michael Sukachev, we will talk about how SAP LeanIX can become much more accessible through Microsoft Teams and AI agents.

Michael, hello and welcome.

In episode 310 of our SAP on Azure video podcast we talk about SAP LeanIX with Microsoft Teams.

Find all the links mentioned here: https://www.saponazurepodcast.de/episode310

Reach out to us for any feedback / questions:

#Microsoft #SAP #Azure #SAPonAzure #SAPLeanIX #EnterpriseArchitecture #MCP #CopilotStudio #MicrosoftTeams

Summary created by AI

  • LeanIX Assistant Democratization:
  • Michael explained to Holger how a Microsoft Teams-based LeanIX Assistant makes enterprise-architecture information accessible through natural-language questions for architects, IT stakeholders, and non-technical users.
    • Accessibility Goal: The project is intended to move enterprise-architecture information from a specialist repository into a conversational experience that users can access where they already work. Users can ask questions in natural language instead of navigating LeanIX, documentation, or multiple data sources.
    • Initial Implementation: The first version was a declarative Microsoft 365 agent built in Copilot Studio and delivered through Microsoft Teams. It used internal enterprise-architecture practice documents, LeanIX user and modeling guides, vendor best practices, and related governance material as knowledge sources.
    • User Benefits: Michael said the assistant can help people understand their roles and processes, while Holger highlighted use cases such as looking up information while creating diagrams or preparing reports. The assistant is not intended to eliminate enterprise architects; professionals remain responsible for maintaining the models and ensuring data quality.
  • LeanIX MCP Integration And Governance:
  • Michael described the progression from document-based assistance to live SAP LeanIX data through an MCP server and then to a governed Azure API Management gateway that can support multiple agents.
    • Live Data Connection: The second stage connected the Copilot Studio assistant to the SAP LeanIX MCP server, allowing responses to use current enterprise-architecture data in addition to static documentation. Michael and Holger identified SAP LeanIX’s MCP server as the enabler for this direct connection.
    • Governed Gateway: To avoid limiting the integration to one agent and to apply enterprise controls, Michael introduced an MCP gateway through Azure API Management. The gateway provides a managed and reusable connection with the usual API governance benefits.
    • API Center Integration: New MCP servers are registered through Azure API Center and authenticated with Microsoft Entra ID applications. Michael explained that this allows the related information to flow back into LeanIX’s inventory, reducing the need to recreate entries manually whenever a new server is deployed.
    • Future Automation: Michael is exploring Azure AI Foundry and autonomous agents to automate updates from vendor information and internal systems. He described this as a future stage that could change how the enterprise-architecture practice keeps information current.
  • Teams Agent Configuration:
  • Michael demonstrated how the LeanIX Assistant is configured in Copilot Studio and published to Microsoft Teams, including its instructions, knowledge sources, tools, and model selection.
    • Agent Structure: The agent uses defined instructions and prompts, a knowledge base containing multiple sources, and tools connected through Power Automate and MCP connections. Michael showed that the MCP connection appears as a usable tool within the agent configuration.
    • Teams Distribution: Microsoft Teams is available as a default Microsoft channel for the agent because the setup uses Microsoft authentication. Other channels require different deployment arrangements, while Teams provides a direct distribution path to the organization’s users.
    • Model Selection: Michael explained that the agent configuration allows different large language models to be selected. The appropriate model can vary according to the agent’s purpose, balancing response quality and efficiency.
    • Access Scope: The demonstrated assistant was published for organizational users, allowing them to interact with the agent through Teams and access the content made available to it. The configuration therefore requires appropriate governance of knowledge sources and sharing settings.
  • Natural-Language LeanIX Queries:
  • Michael demonstrated that the assistant can retrieve LeanIX information and return structured answers about APIs, enterprise-architecture processes, application portfolios, and technology lifecycle risks.
    • Portfolio Queries: The assistant can answer questions about the application portfolio and provide lists or summaries of applications and related objects stored in LeanIX. Michael showed that the results can be presented in a readable layout rather than as raw repository data.
    • Process Information: A demonstrated query about Teranet’s enterprise-architecture processes returned information on governance activities such as the Architecture Review Board and technology or software approval. The assistant combines repository data with the organization’s documented practices.
    • Data Quality Findings: The displayed results included observations such as missing interface categories and protocols and gaps in quality seals for API applications. These findings can be used to identify data-quality campaigns or follow-up work.
    • Lifecycle And Technical Debt: Michael used technology end-of-life and enterprise technical-debt questions as examples. The assistant identifies affected technologies, summarizes their status, and can produce recommended priorities, such as planning upgrades or replacements for aging infrastructure and software.
  • Knowledge Quality And Maintenance:
  • Michael and Holger clarified that the assistant’s usefulness depends on accurate LeanIX models and authoritative internal definitions, while future write capabilities could reduce the effort required to keep information synchronized.
    • Authoritative Sources: The assistant combines live LeanIX data with internal definitions, agreements, governance standards, and documentation supplied through its knowledge base. Michael explained that this combination makes answers more specific and aligned with the organization’s practices.
    • Human Oversight: Michael emphasized that enterprise architects and the architectural community still need to maintain the underlying model and protect data quality. The assistant improves access to information but does not replace professional stewardship of the repository.
    • Read-Only Limitation: The current LeanIX MCP tool set is read-only and does not provide write functions. As a result, users can query and analyze information but cannot yet update LeanIX directly from the Teams conversation.
    • Potential Write Workflow: Michael said future MCP capabilities or a custom API-based implementation could support write operations. Holger described a possible workflow in which a user reports an update or new functionality in Teams, after which the LeanIX data would be updated and subsequent reports would reflect the change.
    • Synchronization Challenge: Michael noted that architecture information is often maintained manually in multiple unintegrated locations. The intended future workflow is to aggregate and maintain the information more easily so that reports remain current with less duplicated effort.

Video index

Based on and using Hugo Cookbook Theme, Font Awesome Icons and Bulma CSS. | Thanks to Recipes4AJ to get me started. | Contact me via | Imprint & Privacy | Cookie settings