LinkedIn – June 16, 2026

Author: Kevin Kieller

Meet Claude, Your New Senior UC Engineer

Your next senior UC engineer is likely to be named Claude, or Gemini, or Copilot, or ChatGPT (but probably not Grok).

IT and UC administrators have tough jobs. There are always more users. More locations. More policies. More complexity. More compliance requirements. More security settings. More Microsoft Teams features, arriving at a pace that makes even experienced administrators struggle to keep up.

AI has offered to help, but this help has not always been accepted or useful.

Initially there was mostly avoidance. Many UC administrators did not use AI at all, or they treated it as an interesting side tool that was not ready for production administration. Truthfully, Microsoft Teams administration is not a place for vague suggestions. A small mistake in policy assignment, emergency calling configuration, number management, or user provisioning can quickly become a real business or security problem.

Quickly however, admins started asking tools such as ChatGPT, Claude, Copilot, or Gemini to help create PowerShell scripts. An admin could ask AI for a script to complete a specific task: identify users without assigned phone numbers, bulk update Teams policies, or on-board a new batch of employees. The AI did not administer the environment directly, but AI-generated scripts helped the admin move faster. AI generated scripts are immensely helpful for the more complex, far-reaching changes that previously took a significant amount of time to research. AI-generated scripts also provide significant value to less experienced admins, although there is a risk that the human admin could lack the ability to understand and adequately double-check the scripts.

With AI-generated scripts, the admin still had to know whether the script was correct, whether the right module was being used, whether the permissions were appropriate, whether the script handled exceptions properly, and whether it could be safely run at scale. AI helped write the instructions, but the admin still had to be the engineer, the reviewer, the operator, and the safety net.

Increasingly we have AI inside the admin interface. Microsoft and other platform vendors are adding AI capabilities to their portals and admin tools. Copilot shows up in the Admin Center, The Teams Admin Center (TAC) and a variety of other admin center portals (arguably Microsoft has too many admin portals!) This can help administrators find settings, understand recommendations, and perform common tasks more efficiently. That is valuable, especially when the AI is grounded in the product experience and constrained by the admin interface.

But the bigger shift may be what comes next.

With Model Context Protocol (MCP), AI tools can connect to external systems through a more standardized interface. Instead of asking Claude, ChatGPT, or another model to simply explain how to perform a task, the AI can interact with tools that expose real operational capabilities. In the UC world, this means an AI assistant could potentially analyze, recommend, and initiate Microsoft Teams administration actions. The AI assistant evolves to being the Virtual Engineer.

The AI assistant is no longer just a script writer. It becomes a natural language interface to an operational platform. A business user, help desk analyst, regional administrator, or UC engineer could ask a question or request an outcome in plain language. For example:

• “Show me all users in the Toronto office who are licensed for Teams Phone but do not have a number assigned.”
• “Move these 25 users to the new calling policy for the Vancouver branch.”
• “Prepare the changes required to onboard the new sales office, but do not apply them until I approve.”
• “Why did call quality drop for users in this location last week?”
• “Compare call quality across all of the North American locations I manage, creating a graph that ranks the total number of poor calls by office.”

In these examples, AI has direct access to Microsoft Teams and potentially other production systems. This is powerful but also potentially problematic!

A raw AI connection to Microsoft Teams administration would be risky. Giving a general-purpose model direct access to execute administrative actions is not the same as giving an experienced UC engineer access to a production environment. The model may misunderstand intent. The user may ask for the wrong thing. The request may be incomplete. The action may affect users outside the intended scope. The change may need to happen across several systems, not just inside Teams.

That is why the MCP server interface is only part of the story. The more important part is what sits behind it.

I recently was given a demo of using Claude along with the VOSS platform and left impressed.

A platform such as VOSS provides an operational control plane that can be safely leveraged by AI. An admin platform can enforce permissions, so users only see and do what they are authorized to do. It can apply role-based administration, so a regional admin can manage users in their region without being granted global rights. It can maintain audit trails, so every request, recommendation, approval, and change can be reviewed. It can use workflows, so actions are performed consistently rather than improvised one prompt at a time.

Importantly, it can treat an administrative process as a transaction.

That matters in UC. Many tasks are not single step actions. Assigning a user for enterprise voice may involve licensing, number inventory, policies, emergency location, routing, caller ID, voicemail, and reporting. Moving a group of users may require validation before the change, orchestration during the change, and verification after the change. If one step fails, the system should know what happened, what completed, what did not complete, and whether the change can be safely retried or rolled back.

This is the difference between AI-assisted administration and AI-driven operations.

AI-assisted administration helps the admin think, write, search, and troubleshoot. AI-driven operations can help initiate and complete work, but only if they have the guardrails, permissions, workflow, auditability, and rollback capabilities required for production environments.

AI becomes a new interface to UC engineering expertise. It gives more people access to the right operational processes without giving everyone unrestricted administrative power. It helps senior engineers scale their knowledge by encoding it into workflows, policies, templates, and approved actions. It lets junior staff and business users request outcomes while the platform ensures those outcomes are executed safely.

Claude is not the senior UC engineer by itself. ChatGPT is not the Teams admin. Copilot is not the change manager. The LLM is the conversational front end. MCP is the connection mechanism. A platform like VOSS is the governed execution layer.

The value comes from combining all three.

This is where UC and IT administration are headed. Not just better scripts. Not just smarter portals with AI added. The next step is natural language access to governed, auditable, transactional administration.

For current Microsoft Teams administrators (and IT admins in general), this will be a major shift. The future of UC operations will not be a human clicking through five admin portals or copying AI-generated PowerShell into a terminal. The future will be an authorized user asking for an outcome, an AI assistant interpreting the request, and a trusted platform doing the heavy lifting safely.

Go deeper

• Attend my session in the UK on June 24, 2026 at Commsverse – Copilot is Redefining Your Job (And You Have 12 Months to Adapt) where I talk about how AI is changing the role of the IT Administrator along with my co-presenter @Tristan Gayford from EY.
• Sign up to attend a LinkedIn Live session on June 30th where I explore AI-powered administration in more detail with Bill Dellara from VOSS Solutions, including some real demos.
• Schedule your own 1:1 demo of the future of IT and UC Administration with a specialist from VOSS.
• Connect with me on LinkedIn and feel free to ask any questions you may have. I respond to all comments and questions.