A shared space for drafts and decisions

On 16 September, Microsoft introduced Copilot Pages as a dynamic workspace where Copilot responses could be added, edited and shared. That made the output less transient than a personal chat: a team could combine material from several prompts, add its own assessment and use the page as a working document. Teams meeting recaps also combined the transcript with meeting chat, allowing decisions and unresolved points to be considered together.

For project work, this was a practical change. A summary could become the first version of a plan or decision paper, but Microsoft’s launch did not turn the content into an approved fact. Teams still had to decide who checked the assumptions, maintained the page and made the decision. The value was faster synthesis and collaboration when the workspace had a clear owner.

Analysis moved closer to everyday work

Copilot in Excel became generally available, while Copilot in Excel with Python entered public preview for analysis expressed in natural language. The Power BI update on 25 September also added dark mode and developed Copilot report creation further. Its dialogue gathered more context before creating a report page and then showed an outline of the fields used. That was a concrete aid to traceability: users had a better chance to check whether the report actually answered the question.

For organisations, the combination suggested a connected analytical process: explore data in Excel, publish management information in Power BI, and develop findings together in Pages. The benefit still depended on defined metrics and governed datasets. AI could shorten the route to an analysis, but it could not decide whether a customer concept, project target or financial definition was correct.

Management therefore had to distinguish a demonstration from production use. A demonstration could show Copilot creating a page or proposing a report. Production also required access control, known data sources, named metric owners and a way to compare the result with the current process. Without these elements, it was difficult to know whether faster production also improved decisions.

Agents made automation more operational

The Copilot Studio announcement on the same day described agents as specialised assistants that could use organisational knowledge and take actions for users. SharePoint agents focused on content in sites and libraries, while Copilot Studio allowed organisations to build their own agents and connect them to processes.

A sensible first experiment was therefore a bounded, low-consequence chain: retrieve documents from a defined location, prepare a proposal, and route it to a named person for review. Before wider rollout, the organisation needed to determine which sources the agent could use, which actions it could perform, how errors would be detected and who would follow up. September’s release made the technology more accessible and made process ownership and data quality more visible as conditions for value.

A peer-reviewed study published on 18 September likewise found that human-centred design needed to support targeted work and learning. This reinforced the need to assess task design and the substance of the work, alongside technical execution.

Sources

Microsoft 365 Blog: Microsoft 365 Copilot Wave 2: Pages, Python in Excel and agents, 16 September 2024.

Microsoft Copilot Blog: Copilot agents built with Microsoft Copilot Studio, 16 September 2024.

Microsoft Power BI Blog: Power BI September 2024 Feature Summary, 25 September 2024.

Microsoft WorkLab: 2024 Work Trend Index Annual Report, 8 May 2024.

Springer / Zeitschrift für Arbeitswissenschaft: Effects of AI-based technologies on employees’ work engagement, 18 September 2024.

For discussion

Which bounded process has both a clear data foundation and an accountable owner, making it suitable for an agent pilot without obscuring responsibility?