Dataverse Security via coding agent for IT admins

Author: Microsoft Power Platform · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Dataverse Security via coding agent for IT admins · Channel: Microsoft Power Platform · 09:58 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Software engineering with AI, Privacy and data protection

Summary

Paul Liew demonstrates how Dataverse security skills in a coding agent can help administrators assess security posture, identify access risks, and manage security workflows using natural language. The agent performs a read-only review that surfaces issues such as excessive administrator assignments, weak business-unit segregation, and disabled identities retaining roles, and can help remove specified role assignments. The demonstration also shows how administrators can design and test column-level security for sensitive contact information. The video concludes that coding agents can make Dataverse security management faster and more proactive, and outlines how to install the Dataverse skills.

Key points

Dataverse security skills support security posture assessment, access-risk assessment, issue monitoring and mitigation, and security lifecycle management.
The coding agent can install the Dataverse skills and connect to a Dataverse environment.
Administrators can use natural language prompts to request a read-only inventory of security configuration and potential overbroad assignments.
The assessment identified 26 direct system administrator assignments and weak business-unit segregation, with 247 of 252 enabled users in one business unit.
The agent identified 49 disabled identities that retained roles; the administrator then requested removal of their role assignments.
A further assessment flagged inactive or unlicensed administrators, stale privilege access, overlapping roles, and applications that can impersonate users.
Administrators can describe security requirements in natural language and use the agent to design and implement a security model.
The demonstration enabled column-level security for sensitive contact information and created an app to test access.
Nancy was assigned a security profile that allowed her to read the mobile phone number field, and the app was used to validate that access.
The video says Dataverse skills can be added to Claude Code or GitHub Copilot by adding the marketplace and installing the plugin.

Chapters

Start	Chapter	Summary
00:00	Introducing Dataverse security skills	Paul Liew introduces the coding agent and explains how Dataverse security skills can help administrators address manual security work.
02:05	Assessing security posture and mitigating risks	The agent performs a read-only security inventory, surfaces privilege and governance concerns, and helps remove roles from disabled identities.
05:18	Monitoring issues and designing security models	The demonstration reviews high-privilege access and shows how administrators can describe and implement security requirements.
07:21	Testing column-level security and installation	The agent configures and tests access to sensitive contact fields, then the video summarizes the benefits and installation steps for Dataverse skills.

Quotes

In this session, I’ll show how to use Dataverse skills in Coding Agent to help administrators strengthen security using natural language. — 00:00

Do not change anything because you just want to do an assessment to get an inventory. — 03:06

It came back with a read-only assessment reviewing concentrated privilege access, several governance gaps. — 05:49

It helps administrators understand their security posture, assess and mitigate risks, troubleshoot access issues, and design security solutions faster and more confidently. — 09:25

Concepts and entities

Dataverse, Coding Agent, Dataverse security skills, Copilot, Visual Studio Code, Claude Code, GitHub Copilot, Agent 365, Paul Liew

Full transcript (original language)

00:00 [ Music ] PAUL LIEW: Hi, my name is Paul Liew, and I’m a product manager on the Agent 365 Dataverse team. In this session, I’ll show how to use Dataverse skills in Coding Agent to help administrators strengthen security using natural language.

00:32 Here’s the problem. Access reviews, role and privilege design, and compliance checks are still largely manual. And they’re slow, error-prone, and hard to keep up with as your environment grows. The Dataverse security skills with coding agent supports administrators across four security-connected areas. You can use it to assess the current security posture, assess user and data access risk, monitor and mitigate security issues,

01:03 and manage the security lifecycle. Now, let’s take a look at how an administrator can use the Coding Agent directly from Visual Studio Code to complete security analysis and management workflows. First, we begin by equipping the Coding Agent with Dataverse security skills. From Visual Studio Code, the administrator asks Copilot to install the Dataverse skills. The Coding Agent autonomously validates the installation

01:33 command, checks available plugin sources, discovers the required marketplace repository, registers it, and completes the installation. Notice how the agent reasons through each step, identifies missing dependencies, performs web lookup when needed, and successfully installs the Dataverse skills without requiring the administrator to manually troubleshoot the process. This demonstrates one of the core values of the security Coding Agent, reducing operational complexity

02:05 by automating tasks that traditionally require deep technical knowledge. The installed Dataverse skills include the ability to easily discover and connect to your Dataverse environment. Once connected to a Dataverse environment, the Coding Agent rapidly analyzes the overall security posture. Administrators can ask natural language questions such as show me users with elevated privileges;

02:35 identify tables containing sensitive business information or summarize my environment security configuration. Here, I want to discover and document the current security posture of my environment. So here’s the prompt. Analyze this environment’s security configuration, show business units, teams, security roles, column security profile, application users, agent users,

03:06 and assignments that appear broader than necessary. And here’s the critical part. Do not change anything because you just want to do an assessment to get an inventory. The agent completed a read-only security review without making changes. It surfaced some highest risk findings. For example, it found that an intra-security group team hosts

03:36 system administrator access while allowing members and guests with external membership not visible from Dataverse. It also found 26 direct system administrator assignments, including application identities and test style accounts. It also has weak business units segregation with 247 out of the 252 enabled users in one business unit. It shows the number of business units and the number of teams.

04:10 It tallies the number of application and agent users and identifies the application with system administrator security role assignment. And it also shows the column level security profile that there are 24 profiles, 14 managed and 10 unmanaged. It provides an inventory of security profile and shows recommendation and summarizes them in a table. Together, these findings give the administrator a prioritized

04:42 view of where privilege access, identity governance, and environment structure requires immediate review. Here, it identifies disabled identities retaining roles. There are 49 of them. The next step, we can do some mitigation. And the admin decides to take an action. The admin asks to provide the list of identities that have disabled status and remove the role assignment. Now with the copilot Dataverse skill, those role assignments

05:18 from disabled identities were being removed. So the next example is how the admin can monitor and mitigate security issues. The admin creates a prompt to assess high privilege roles and assignments, checklist privilege for each user, team, application, and agent,

05:49 and identify significant activities and inactivity. It came back with a read-only assessment reviewing concentrated privilege access, several governance gaps. It found 26 direct system administrator assignments across licensed users, unlicensed administrative accounts, application identities, and only 3

06:21 out of the 9 licensed administrators showed recent activity, while 5 unlicensed administrators retained access with little or no usage evidence. The agent also flagged an intragroup whose effective memberships must be verified externally, an active agent with no authentication configured, stale privilege access, overlapping roles, and applications that can impersonate users.

06:51 These findings help administrator prioritize, identify, validate, and remove unnecessary privilege and strengthen agent and application controls. Recommendations. The admin might want to take an action for this risk, review and remove all dominant administrator. Beyond analysis and risk assessment, the coding agent can assist with designing

07:21 and implementing security models. Administrator can describe desired business requirements in natural language, such as create a security design where regional sales manager can view all opportunities within their region while sellers can only access their own records or enabling column-level security for high-sensitive data and create security profile. Design, implement, and test a new security model.

07:52 You can design and create column-level security for sensitive columns for the contract table, example, email and mobile phone number. And update the security profile for these two columns and create an app to test out different persona who can have access to this secure column. The email and mobile phone number columns on the contact table will identify a sensitive column and the column-level security properties were enabled.

08:23 And a visualizer app or portal was created to allow testing and validation. Two tables were shown on the app. The contact table showed two column-level security columns. Admin can assign user team to the updated column-level security profile. Here, Nancy is assigned to be able

08:54 to read the mobile phone number column-level secure field. Next, you can test Nancy’s access with the updated security profile using the app. Select Nancy and scroll down to the contact table, refresh it, and now you can see that Nancy has access to the secure mobile phone number. Using database skills

09:25 with coding agent transforms how administrator manage security by combining AI-powered reasoning with deep Dataverse security expertise. It helps administrators understand their security posture, assess and mitigate risks, troubleshoot access issues, and design security solutions faster and more confidently. Thank you for watching this demonstration of the Dataverse security coding agent. We look forward to helping administrators make Dataverse

09:58 security management more intelligent, proactive, and efficient. Previewing Dataverse skills takes just two commands. Add the Microsoft Dataverse skill marketplace to your coding agent, whether that is Claude Code or GitHub Copilot. Then install the Dataverse plugin and your agent is ready to start working with Dataverse security in natural language. Thank you.