Blog

Maximilian Kenfenheuer

Manager · Microsoft 365 & AI Transformation @ BearingPoint

Featured

#Microsoft365Copilot #MicrosoftAI #Copilot #AITransformation #FutureOfWork

Measuring Completed Work: Baselines, Assisted Hours and Business Outcomes

Thursday, September 10, 2026 · Maximilian Kenfenheuer

A team can generate thousands of prompts without improving a single core process. Another uses AI rarely but saves hours in contract review. Microsoft increasingly measures Copilot value by completed work and assisted hours instead of activity. Its internal guidance recommends baselines for painful processes before the rollout. In this article, I describe a three-layer measurement model of adoption signals, time and effort indicators and business outcomes, and explain why role-based deployment makes value measurable.

Power BI Grounding: Copilot Reasons Over Semantic Models

Wednesday, September 9, 2026 · Maximilian Kenfenheuer

Since the August 2026 update, Microsoft 365 Copilot reasons over Power BI reports and semantic models in Chat and Cowork. Copilot thus works with governed business definitions: how revenue is defined, how margin is calculated, which KPIs leadership uses and who may see which numbers. At the same time, a poorly defined KPI now spreads faster. In this article, I explain why semantic models become the foundation for trustworthy AI answers and how organizations can prepare their analytics layer.

From Draft to Deliverable: Agentic Copilot in the Core Apps

Saturday, September 5, 2026 · Maximilian Kenfenheuer

On April 22, Microsoft made agentic capabilities in Word, Excel and PowerPoint generally available and brought business apps into Copilot via agents. AI moves to where proposals are written, financial models built and board decks assembled. At the same time, human judgment gains weight: a polished deck is not automatically a sound strategy, and a well-structured spreadsheet is not automatically a correct model. In this article, I explain what changes when AI participates in the work and which five readiness questions organizations should answer.

Latest posts

Multi-Agent Orchestration with Fabric, the Agents SDK and A2A

Tuesday, September 1, 2026 · Maximilian Kenfenheuer

#MicrosoftAI #Microsoft365Copilot #CopilotStudio #AI #EnterpriseAI

Enterprise processes rarely fit into a single agent. One retrieves information, another analyzes data, a third drafts content in Microsoft 365, a fourth triggers an action. Copilot Studio now supports multi-agent orchestration with Microsoft Fabric, the Microsoft 365 Agents SDK and Agent-to-Agent (A2A) communication via open protocols. In this article, I explain why specialized agents scale better than one monolithic agent and which governance questions arise when agents delegate work to each other.

Agent Evaluation with Custom Graders in Copilot Studio

Monday, August 31, 2026 · Maximilian Kenfenheuer

#MicrosoftAI #Microsoft365Copilot #CopilotStudio #AIGovernance #AgenticAI

"The demo worked" and "users liked it" are not enough once agents support customer interactions, approvals or system actions. Copilot Studio now offers custom graders for agent evaluations. Organizations assess agents against their own criteria: approved terminology, correct internal sources, escalation at low confidence and compliance with process boundaries. In this article, I explain why evaluation and governance belong together and how organizations can define and measure agent quality in five steps.

Ten Generic Tools: The Work IQ MCP Server and Agent Architecture

Sunday, August 30, 2026 · Maximilian Kenfenheuer

#Microsoft365Copilot #MicrosoftAI #CopilotStudio #AI #EnterpriseAI

The Work IQ API endpoints reach general availability on June 16. Work IQ combines Chat, Context, Tools and Workspaces and turns intelligence into a reusable service layer for agents. One design choice stands out: the Work IQ MCP server reduces hundreds of operations to 10 generic tools, and getSchema lets agents discover data structures dynamically. Usage is independent of Copilot licensing and billed on consumption. In this article, I explain what this architecture means for consistency, speed and governance of custom agents.

Evaluating Models Instead of Benchmarks: Model Choice in Copilot

Saturday, August 29, 2026 · Maximilian Kenfenheuer

#Microsoft365Copilot #MicrosoftAI #AI #Copilot #EnterpriseAI

Microsoft 365 Copilot now offers more models, including Anthropic models and new frontier reasoning options for agentic work. A sales summary, a policy review and a multi-step agent workflow place different demands on a model. Enterprises do not want endless model sprawl. They want controlled choice inside a governed platform. In this article, I explain why structured evaluation matters more than benchmark narratives and how organizations can group use cases for their model strategy.

Domain Exclusion in Practice: Ownership, Maintenance and Limits

Saturday, August 29, 2026 · Maximilian Kenfenheuer

#Microsoft365Copilot #MicrosoftAI #Copilot #AIGovernance #ResponsibleAI

AI governance used to ask whether employees may use AI and which model is approved. Today, organizations need finer controls, including which external sources may influence Copilot answers. Domain exclusion lets admins exclude up to 1,000 websites from web grounding in Microsoft 365 Copilot. The scope is limited to web page results, and the list is maintained via PowerShell and CSV. In this article, I explain why source controls shape answer quality and who in the organization should own the exclusion list.