SharePoint CRM Gets User-Tested Fixes in Copilot Code

Key insights

• The CRM's first version used SharePoint lists and libraries as its back end.
• Anderson tested contacts, companies, deals, calls, tasks and documents in the app.
• He found that calls could not be linked to contacts or companies.
• He asked Copilot Code to add activity links and drag-and-drop deal stages; both worked in his test.
• Document metadata set in the app appeared on the SharePoint library.

Daniel Anderson [MVP] recently published the second part of a working session testing a CRM he built with Copilot Code. The prototype began with requirements for a HubSpot-style CRM and used SharePoint lists and libraries as its back end. In this follow-up, Anderson clicks through the app, spots issues, asks Copilot Code to make changes, and tests the results.

What's new

The video is a practical test of the first version rather than an announcement of a Microsoft product release. Anderson tries common CRM activities: adding a contact, associating it with a company, creating a deal, logging a call, and uploading documents. He also checks how the app handles tasks and files linked to a company.

Two missing or desired behaviors shape the session. Anderson finds that he cannot link a call to a contact or company, and he wants to move deals between stages by dragging them. He describes both changes in the chat interface and continues working while Copilot Code makes them.

Afterward, he tests the changes and reports that both are working: activities can be linked, and deals can be moved between stages with drag and drop. He also demonstrates that document metadata entered in the app appears on the SharePoint library. The video includes a brief detour in which he asks for a Space Invaders game; the description says that this used 267 credits.

Why it matters

For makers evaluating AI-assisted app building, the useful detail is the iterate-and-test cycle. The example starts with a generated application, then uses ordinary interaction to expose requirements that were not met in the first pass. The maker can describe the gaps in chat and check whether the revised app behaves as intended.

The SharePoint connection is also concrete: the original CRM uses lists and libraries for its back end, and metadata from the app shows up in a library. That gives SharePoint practitioners a view of how an app-building session can interact with SharePoint content and data. The video does not establish that the prototype is ready for business use or that its design suits other organizations.

For administrators, the demonstration is best read as an example of a prototype, not as guidance on deployment, governance, or support. It does not discuss permissions, security, data protection, licensing requirements, or a rollout stage. Makers and end users can see the interactions being tested, but the video does not provide enough information to judge the app's performance or suitability for production.

How it works

Anderson's description and demonstration show a straightforward sequence. Copilot Code first produced an initial version from requirements for a HubSpot-style CRM, with SharePoint lists and libraries behind it. In part two, Anderson uses the app like a user rather than presenting a scripted walkthrough.

1. He adds a contact and links it to a company, then creates a deal.
2. He logs a call and discovers that it cannot be associated with a contact or company.
3. He asks Copilot Code in chat to add the missing activity links and requests drag-and-drop movement between deal stages.
4. He continues testing while the requested changes are made, then checks that the two behaviors work.
5. He uploads documents and verifies that metadata entered in the app appears on the SharePoint library.

The video also shows where to find the code Copilot Code writes, though the supplied description does not provide the location or explain how to deploy or maintain it. The game-building aside is separate from the CRM workflow. Its reported 267-credit cost is specific to that moment in the session, not a general measure of what building a CRM costs.

Limitations and open questions

This is one creator's hands-on test, and the supplied material does not describe formal testing, a production deployment, or use by a team. It does not identify the precise Copilot Code plan, licensing, availability, or rollout status. Readers should not infer from the demonstration that the same features or results are available in every environment.

Several operational questions remain unanswered: how permissions are configured, how data relationships are maintained, what happens when an update fails, and how the app would be governed or supported. The material also does not explain the CRM's full feature set or offer a comparison with a commercial CRM. It shows two fixes working in the session, but does not establish reliability beyond that test.

For a maker considering a similar prototype, the video is most useful as an example of what to inspect: everyday actions, links between records, deal-stage interactions, and the connection between app metadata and a SharePoint library. The next step for any real deployment would require evidence beyond this demonstration, especially on security, access, and operational requirements. Those details are not covered in the video description or transcript provided here.

Sources

• Daniel Anderson's Copilot Code CRM test, part two

Frequently asked questions

What did the CRM use for its back end?

The first version used SharePoint lists and libraries, according to the video description.

What did Copilot Code change during the session?

Anderson asked it to add links between calls and contacts or companies, and to support dragging deals between stages. He reports that both worked when tested.

Does the video show whether this CRM is ready for production?

No. It shows a user-testing session but does not discuss production readiness, governance, security, or support.

What happened to the document metadata?

Anderson demonstrates that metadata set in the app appears on the SharePoint library.