Article content
In briefShow moreShow lessPower BI could open a Copilot summary from a Teams message and improved data questions.
- Power BI could open a Copilot summary from a Teams message and improved data questions.
- Microsoft 365 Copilot added richer Pages and Copilot during Teams calls without recording.
- Project teams should separate fixed workflows from autonomous agent tasks and measure quality as well as use.
From message to inspectable report
The 31 March Power BI update added a route from a Teams preview to a Copilot summary in Power BI. It also improved language understanding, ad hoc calculations and DAX queries. This could accelerate initial analysis, but a summary still needed traceability to report, filter, timestamp and model version before becoming decision evidence.
More surfaces for collaboration
Microsoft 365 Copilot gained richer Pages and support during Teams calls without recording or transcription. Project teams could collect drafts, questions and follow-up closer to the meeting. The absence of a recording also meant less material for later verification. Material decisions still had to be written down with assumptions, owner and deadline.
The study showed variation among users
A six-month study in Australia’s CSIRO research organization described useful summarization and drafting alongside variation in benefit, trust and skill. That challenged a rollout measured only through activated licenses. Teams should choose concrete tasks, provide examples and test whether delivery became faster without more factual errors or editing.
Workflow and agent were different designs
Anthropic distinguished predefined workflows from agents that selected their own actions. A controlled chain often suited status reporting: retrieve approved measures, flag missing data, draft and obtain project-manager approval. More autonomous search could surface possible deviations, but its findings should not change a plan or notify a customer without review.
Control before scaling
Access and cost belonged in the same experiment. An agent with broad read access could find more, but could also retrieve irrelevant or sensitive material. Usage-based billing might make a simple status routine expensive when it queried many sources or retried repeatedly. The team should log source count, runs, latency and cost per approved report together with denied access attempts. This allowed comparison between an agent and a simpler automated workflow, selecting the least complex design that met the need.
A measurable monthly routine
Begin with one status report. Log sources, edits after the AI draft, time from data cutoff to publication and questions still requiring manual analysis. Connect every conclusion to a task or decision in the project workspace. After four cycles, assess whether automation reduced coordination, improved risk visibility and made the report clearer for its audience.
Sources
Microsoft Power BI: Power BI March 2025 Feature Summary, 2025-03-31
Microsoft 365 Copilot: What’s new in Microsoft 365 Copilot | March 2025, 2025-03-31
Bano et al. (arXiv): A qualitative study of user perception of M365 AI Copilot, 2025-03-22
Anthropic: Building effective agents, 2024-12-19
Join the discussion
Which experience from your own organisation is most relevant to this story? Share what you learned.









