Michael Tenner

Copilot, ChatGPT, Claude and others can create code for you, even if you have never written a script before or are only familiar with one programming language. You can use them and your knowledge to apply it to a new field. MCP Server-based agents are the next leap for engineers and data analysts. The deployment of measures, row-level security, or the creation of user-defined DAX functions can be done with their help. Analysts can also enrich their data and introduce new combinations for ad hoc analysis. In this session, you will learn how to get started with Power BI and MCP. You will learn about basic operations and the new capabilities. This shows a new path for analysts. You will also learn why the combination of TMDL+MCP+UDF is so impressive. How you can use Data from a Semantic Model + Opendata to create a Python Plot .

Near real-time process analyses with Power BI and Snowflake are often implemented using DirectQuery to make operational data available as quickly as possible. In practice, however, it quickly becomes apparent that this architectural decision has far-reaching implications for performance, costs, maintainability, governance and end-user acceptance. Using a real-world process analysis scenario, this session shows how Power BI DirectQuery works on Snowflake, when this approach scales well, and under what conditions performance or stability issues arise. It explains how query behaviour, model flexibility and large time series interact and why near real-time is not the same as true event streaming. Another focus is on the visualisation of large time series: It will be shown how Deneb can be used to display a large number of data points in a high-performance manner and what technical and organisational requirements this entails. In addition, hybrid approaches and strategic options such as mirroring or fabric will be classified. The session provides practical decision-making logic for consciously weighing up architectural options and finding solutions that are both technically feasible and acceptable to end users.