SharePoint Copilot Now Connects Outside

Key insights

• Project summary: Built a business intelligence engine inside SharePoint that pulls live external data into a SharePoint library and makes it usable without touching raw files.
  It combines Copilot in SharePoint, new plugins, and an external scraper to automate monitoring tasks.
• Data ingestion: Used the Apify plugin inside Copilot in SharePoint to scrape a public site and save results as markdown files—one for current state and one for run history.
  The first run produced a baseline of 292 companies and recorded authentication through the Copilot coworking flow.
• Tracking and history: The solution records first seen, last seen, and last checked for every company to enable clear change tracking.
  The run history keeps a full audit so teams can see when entries appeared, disappeared, or changed.
• Skills and agents: Built a grounded business intelligence agent that reads only the markdown folder and a site skill (using SharePoint’s Create Skill pattern) so non-technical users can run workflows safely.
  The skill stacks refresh, data update, and an HTML report behind a single prompt: “Refresh current attendees.”
• Operational workflow: The skill runs on demand when invoked; scheduled runs were not covered in the demo.
  This pattern replaces manual checks and can be reused for competitor pricing, tenders, regulatory updates, or finance data monitoring.
• Platform context: Microsoft is extending Copilot in SharePoint to reach external data with Microsoft 365 Copilot connectors and a SharePoint Server connector for on-prem content.
  General availability starts rolling out on September 30, 2026, and skills respect existing user permissions to keep access governed.

Summary of the video

In a recent YouTube walkthrough, Daniel Anderson [MVP] demonstrates how to build a simple business intelligence engine inside SharePoint - Lists by combining Copilot in SharePoint, the Apify plugin, and SharePoint skills and agents. The video centers on a practical scenario: monitoring which companies attend industry conferences, replacing manual tracking with an automated monitoring engine that records history and reports changes. Anderson shows a full proof-of-concept that pulls live data from a public website into a SharePoint library, stores that data as markdown files, and exposes the results through a site skill so non-technical users can run it without touching the raw files.

How the solution is built

First, Anderson uses the Apify plugin inside Copilot in SharePoint to scrape a public attendee list and save it into SharePoint. Then he stores two kinds of markdown files: one representing the current state and another capturing run history, which allows the system to track changes over time. Next, he layers a skill and an agent on top of that folder to provide a single prompt — "Refresh current attendees" — which refreshes data, updates history, and produces an HTML report in one go.

Importantly, Anderson also tracks meta-data such as first seen, last seen, and last checked for every company, which provides useful signals for downstream analysis. He authenticates Apify through Copilot Cowork and notes that the initial run captured a baseline of 292 companies while using Apify’s free credits. The skill he builds is an on-demand tool in the video; while scheduling is possible, the walkthrough focuses on the manual invocation and testing workflow.

Technical choices and grounding

Anderson emphasizes grounding the business intelligence agent only in the specific SharePoint folder that contains the markdown files, which reduces scope and helps the agent reason from a known data set. Then he uses SharePoint’s Create Skill capability to promote a local markdown skill into a site skill, enabling others to run the same process without accessing raw data. This pattern separates the data layer from the user interface and keeps the run history auditable for compliance and review purposes.

Furthermore, the approach demonstrates a modular pattern: the same architecture can ingest competitor pricing, tenders, regulatory updates, or finance metrics, as Anderson references similar work with aged receivables. Therefore, teams can reuse the pipeline architecture for diverse monitoring needs while keeping the customization primarily in the scraping and markdown transformation step.

Tradeoffs and operational challenges

However, the video also raises important tradeoffs between convenience and governance. On one hand, connectors and plugins let Copilot reach external sources, which reduces manual work and speeds insight discovery; on the other hand, organizations must maintain strict permissions and vet external scraping to avoid compliance or terms-of-service issues. Microsoft 365 documentation suggests that Copilot in SharePoint respects user permissions and that skills should not expand access, yet teams still need to balance automation with auditability and access controls.

In addition, scraping public websites introduces reliability risks: target pages can change, blocking measures can appear, and data quality can vary, so maintenance becomes an operational cost. Meanwhile, choosing between real-time federated connections and indexed connectors involves a further tradeoff of freshness versus search performance, and the video shows an on-demand refresh model that may need adaptation to support scheduled or near-real-time requirements.

Practical implications for teams

Practically speaking, this demonstration shows how non-developers can gain access to near-custom BI inside SharePoint with limited engineering overhead, provided teams manage the supporting scripts and credits for third-party services. As Anderson demonstrates, the approach reduces manual tracking, creates a clear run history, and presents results as readable HTML reports within SharePoint, which improves collaboration and transparency. At the same time, organizations must plan for monitoring, error handling, credit/billing for third-party tools, and documentation so business users trust the outputs.

Finally, the wider context matters: Microsoft’s broader Copilot ecosystem is rolling new capabilities that extend Copilot’s reach to external data sources and on-prem content, which increases potential use cases but also requires thoughtful governance. Consequently, teams that combine these features gain powerful, integrated workflows, but they should also invest in operational practices and security reviews before deploying production monitoring systems based on this pattern.

Conclusion

Daniel Anderson’s video offers a concise, practical example of how Copilot in SharePoint can "talk to the outside world" by combining plugins, markdown-based data stores, and SharePoint skills to create an auditable monitoring engine. While the approach opens useful automation and insight opportunities, it also requires balancing data freshness, maintenance effort, permissions, and legal considerations when scraping external sites. Overall, the walkthrough provides a reusable pattern for teams that want to bring external signals into SharePoint while maintaining control and clarity over how those signals are processed and presented.

Keywords

SharePoint Copilot external access, Copilot SharePoint connectors, SharePoint Copilot web integration, Copilot external data SharePoint, Microsoft Copilot SharePoint update, SharePoint AI assistant external sources, Copilot for SharePoint public web access, SharePoint Copilot website integration