The knowledge base could be bounded

Microsoft’s November roundup made SharePoint agents available for selected sites and documents. A project owner could create a specialist assistant around a bounded content area instead of asking Copilot to search everything a user could reach. Word also supported up to ten references and longer source material, allowing drafts to draw on several documents, meetings and emails.

The boundary was valuable only if the document set had been maintained. Outdated procedures, parallel versions and unclear access would appear in the agent’s answers. Work before publication therefore included naming a content owner, removing obsolete versions and deciding where new decisions would be stored. This was ordinary knowledge management made more visible by AI.

A bounded agent also exposed gaps in its sources. If an answer could not be found in the selected documents, the team had to decide whether the knowledge was missing, stored elsewhere or should remain with a specialist. That failure list was useful project information and could be prioritised with the content owner.

Usage and effect became different measures

Copilot Analytics was introduced with usage reports and a preview report that could relate Copilot use to an organisation’s performance indicators. This shifted attention from counting licences and prompts towards examining relationships with sales, service, finance or marketing outcomes. Microsoft also made clear that several new reports would roll out later.

A project therefore needed to define its measure before the data arrived. Active-user counts described adoption but did not show quality, time saved or better decisions. A suitable outcome might be less time to a reviewed decision paper, fewer searches for the current procedure or less rework. It needed comparison with a clear baseline.

OneLake made analytical assets easier to find

The Power BI update on 12 November introduced the OneLake catalogue as a shared place to find and govern data items. Copilot also reached the Power BI mobile apps, while a new map path layer could show connections between locations. The analytical catalogue was becoming more accessible to operational users and project teams needing insight while work was under way.

A Nature meta-analysis published in late October found that human–AI combinations did not automatically outperform the better party working alone. For November’s features, that suggested a sober test: compare the agent and analytical flow with today’s best process. If a curated SharePoint agent and OneLake catalogue did not reduce searching or misuse, the team should improve its sources before expanding the rollout.

Mobile access also widened the audience for reports. Authors needed to check whether key measures were readable on a small screen, whether maps used the correct locations, and whether an operational user could move from insight to the right task. A catalogue made content discoverable; it did not replace designing for the situation in which data was used.

Sources

Microsoft Community Hub: What's new in Microsoft 365 Copilot | November 2024, 29 November 2024.

Microsoft Power BI Blog: Power BI November 2024 Feature Summary, 12 November 2024.

Nature Human Behaviour: When combinations of humans and AI are useful, 28 October 2024.

Gartner: The Customer Side of AI-Driven Productivity, 1 November 2024.

For discussion

Which project area has both a bounded document set and a defined performance measure that can be tested together?