Our teams insights on AI agents, governance and Power BI evolution from the Dufrain team.
Fabric February in Oslo brought together practitioners exploring how Microsoft Fabric continues to evolve, particularly around AI agents, governance and platform intelligence.
Our team Giorgia Tibaldero and Bharani Selvaraju won their place for the event through an internal Fabric Innovation Drive.
Rather than publishing a traditional recap, we asked Megan Livadas Head of BI – Microsoft, who presented at the event, to lead a conversation with the team on what stood out and why it matters.
Here is what they loved.
Fabric’s Direction and the Role of Intelligence Layers
Q – Megan
You were immersed in sessions across the week. What surprised you most about the direction Fabric is heading?
A – Giorgia
One of the biggest surprises for me was Fabric IQ and the role ontologies play in grounding agents in the business semantic layer.
I initially thought it was more of a marketing concept, but after attending the session I realised how important the creation of a shared context layer will become as organisations integrate AI agents into their data platforms. Fabric IQ provides a semantic modelling and cataloguing interface connected directly to the organisation’s data estate. This means applications can both understand the meaning of data and retrieve it.
Because agents are asked questions that span domains and business entities, we need a machine-readable representation of those interactions. This allows agents to navigate the semantic layer with the correct context when extracting insights. Fabric Graph models and ontology items support this, and ontology items can already ground Fabric Data Agents and Operational Agents currently in preview.
Intelligence Embedded Across the Platform
Q – Megan
Bharani, what stood out from your perspective in how intelligence is being embedded across the platform?
A – Bharani
What really stood out was how the keynote positioned Fabric as an intelligent system rather than just a data platform. Concepts like Fabric IQ, Work IQ and Foundry IQ reinforce that direction.
The demonstrations around OneLake shortcuts for OneDrive and SharePoint highlighted the zero-copy principle where data remains in place but is accessible across workloads. There was also emphasis on shortcut transformations, showing both their capability and the need to be mindful of compute consumption.
More broadly, the messaging focused on intelligence layers understanding operational context, usage patterns and business intent to drive smarter outcomes across analytics and operations.
Power BI Workflow Evolution
Q – Megan
Giorgia, you explored MCP within Power BI workflows. What did you love about its potential?
A – Giorgia
MCP modelling tools were designed to make developer workflows more efficient and standardised. I found their ability to automate repetitive tasks particularly compelling.
They support bulk refactoring, renaming based on conventions, table and measure creation, defining relationships and documentation generation. This includes preparing semantic models for AI usage and building data dictionaries that support handovers.
There are still limitations to be aware of. MCP currently handles semantic model refactoring rather than report visuals and works best with simpler DAX syntax. Output quality also depends on the model used, but overall, it reinforces the idea of learning once, use everywhere when approaching development workflows.
Development Practices and Agent Driven Workflows
Q – Megan
How do you see development practices evolving as agent driven workflows become more common?
A – Giorgia
Sessions on Git best practices highlighted how BI development is aligning more closely with engineering disciplines.
Workspace architecture, source control and deployment pipelines will become essential skills as teams scale. If agent driven development becomes part of everyday workflows, managing those changes requires structured lifecycle control similar to software engineering.
That means local development, merging through source control and alignment with engineering teams. It also introduces a learning curve for BI developers unfamiliar with tools like VS Code. While there will be upskilling and maintenance considerations, this approach enables better governance over changes introduced through agent driven processes.
AI Agents and Embedded Intelligence
Q – Megan
Beyond individual tools, what broader AI themes did you both love seeing reinforced across sessions?
A – Bharani
The rise of AI agents was a consistent theme. Operational Agents were positioned as always on digital teammates configured with goals and contextual data to monitor activity, detect anomalies and recommend actions rather than simply raising alerts.
Integration with Power Automate and Microsoft Teams was notable because recommendations surface directly in collaboration workflows where users can review or act without leaving their environment.
There were also demonstrations of Copilot generating pipelines, creating notebooks and extracting insights using natural language prompts. Built in functions for sentiment analysis and sensitive information detection showed how AI capabilities are becoming embedded into everyday workflows.
Security and Cost Governance
Q – Megan
Security and cost governance always matter to clients. What did you love seeing here?
A – Bharani
Security maturity within OneLake stood out. Access management at the storage layer ensures rules are consistently enforced across compute engines including Spark, SQL and Power BI. The ability to apply column and row level restrictions is particularly important for organisations managing regionally sensitive datasets.
Cost optimisation was another key theme. Guidance included starting small with capacity sizing, optimising lakehouses regularly, isolating Copilot workloads where possible and maintaining visibility through monitoring tools. The focus was on proactive management rather than reactive cost control.
We loved Fabric February
Fabric February reinforced many of the themes explored in our recent Microsoft trends blog, particularly the shift toward embedded intelligence, evolving team structures and governance discipline as platforms mature.
What stood out from this conversation was how these shifts are already shaping delivery thinking. The direction of travel is not just about expanding capability. It is about context, lifecycle discipline and organisational readiness to work differently alongside intelligent systems.
We will continue sharing insights as the platform evolves and as our teams apply these learnings in practice.
Read more of our insights here.
And if you are interested in maximising your Power BI suite – take a look at our offers.