Microsoft is creating a formal way for employees to ask for access to AI agents that their organization has deliberately blocked, a new feature that lands in Microsoft 365 in July 2026.

The change, listed under Roadmap ID 494809, introduces a worldwide, web-based approval flow for Microsoft-built Frontier agents that are currently unavailable to users due to organizational policy. It marks a significant step in Microsoft’s balancing act between empowering employees with AI tools and giving IT administrators control over which agents run inside the company.

Microsoft shared the news in an update to its Microsoft 365 Roadmap, a public-facing site that details features in development. The roadmap entry doesn’t yet include full documentation, but the preview text explains the core mechanics: when a Frontier agent is blocked, employees will see an option to request access. That request triggers an approval workflow, and if granted, the agent becomes available through the Agent Store in Microsoft 365.

What’s actually changing

Starting in July 2026, Microsoft 365 will offer a built-in request mechanism for Frontier agents. These are Microsoft-created AI agents that extend Copilot and other Microsoft 365 services. Examples include agents for summarising meetings, drafting documents, or analysing data.

Right now, IT administrators can block specific agents from appearing in the Agent Store, typically to enforce data governance rules, manage costs, or limit exposure to experimental features. But until this feature arrives, if an agent is blocked, employees have no structured way to ask for it. The only option is an informal email or chat to the help desk, which creates friction and inconsistent responses.

The new workflow changes that. Here’s how it works, based on the roadmap description:

  • The block remains in place – An administrator still decides which Frontier agents are unavailable.
  • Employees see a request button – Inside the Agent Store or perhaps within the Copilot interface, users will be able to initiate a request for a blocked agent.
  • The approval flow fires – A web-based form, likely integrated with Power Platform or Teams Approvals, sends the request to the relevant decision-maker.
  • Admin review and action – IT staff or business approvers can approve or deny the request. Approved agents then appear in the requesting user’s Agent Store.

Microsoft has not yet published screenshots or detailed configuration options. The roadmap summary cuts off after “admini,” but the intent is clear: give end users a voice without sacrificing administrative guardrails.

What it means for you

For everyday Microsoft 365 users: This is about removing barriers. If you’re a project manager who wants to use a meeting-summarisation agent that your IT team hasn’t rolled out, you’ll soon be able to ask for it directly inside the app. No more filing tickets or chasing approvals through email. The process should be quick and transparent, with notifications when a decision is made.

For IT administrators and governance leads: Prepare for a potential wave of requests. The feature may reduce the shadow-IT drive—employees won’t need to find workarounds like installing unmanaged browser extensions—but it will also force a clearer internal process for evaluating agents. Microsoft’s approach ties the approval flow into the existing Agent Store infrastructure, so you’ll likely be able to set rules on who can approve requests, set expiration dates, and apply conditional access policies.

For developers and ISVs: This feature only applies to Microsoft-built Frontier agents, not third-party or custom agents. So if you’re building an agent using Copilot Studio, this doesn’t directly change anything. But it signals that Microsoft is investing in user-driven adoption patterns that could eventually extend to custom agents as well.

For business decision-makers: The change aligns with a broader trend of shifting procurement from top-down mandates to employee-driven requests. It could accelerate the adoption of useful AI tools that might otherwise get overlooked. Make sure your IT governance board is ready to evaluate these requests quickly, or you risk annoying employees who see colleagues in other companies using the latest agents.

How we got here: The agent explosion and the governance gap

Microsoft first introduced AI agents as part of Copilot in early 2024, starting with simple plug-ins. Over the last year, the agent ecosystem has grown rapidly. At Ignite 2024, Microsoft announced a dedicated Agent Store, place for both Microsoft-made and partner-built agents. At the same time, IT administrators gained controls to allow or block specific agents through the Microsoft 365 admin center.

But the controls were binary: block or don’t block. There was no middle ground for “blocked, but employees can ask for an exception.” That created a dilemma: block too many agents and employees complain; block too few and data risks multiply.

Roadmap ID 494809 is Microsoft’s answer. It follows several other governance features, like the “Copilot Control System” launched in early 2025, which lets admins set agent usage policies. The request flow adds a layer of human-in-the-loop approval on top of automated blocks. It’s reminiscent of how Microsoft has handled similar request patterns in Azure—for example, the ability for developers to request access to restricted Azure services via the Azure Portal.

The July 2026 timeline fits Microsoft’s pattern of announcing long-lead roadmap items. It suggests the feature is still in early development, possibly entering private preview later this year. The worldwide tag means all Microsoft 365 tenants will get it, not just those in specific regions.

What to do now

For IT administrators:
- Audit your blocked agent list. Go to the Microsoft 365 admin center and check which Frontier agents you’ve blocked. Decide which ones you’d be willing to approve under certain conditions.
- Designate approvers. The feature will likely let you assign approval responsibility to specific users or groups. Start identifying those people—perhaps a mix of security, compliance, and business-line leads.
- Document criteria for approval. What evidence would you need before allowing a blocked agent? Data classification, business justification, budget sign-off? Having a clear rubric will speed up requests.
- Prepare end-user communication. Before July 2026, tell employees this feature is coming and how it works. Clarify that requesting an agent doesn’t guarantee approval.

For end users:
- Familiarise yourself with the Agent Store. See what’s already available. If you know of a Frontier agent that would boost your productivity, take note of its name and intended use.
- Build a business case. When the feature goes live, your request will likely require a justification. Think about how the agent saves time, reduces errors, or ties to business goals.
- Don’t expect instant approvals. The process still involves human review; plan accordingly.

For Microsoft 365 Champions and adoption specialists:
- Include the request flow in your training materials. Employees might not know they can ask for blocked agents unless you tell them.
- Collect feedback early. If your organisation gets early access through a public preview, report any quirks in the approval experience to Microsoft and to your internal IT team.

Outlook: A smarter balance between enablement and control

Feature 494809 points to a Microsoft 365 future where governance isn’t just about saying “no.” It’s about creating a channel for “not yet, but maybe.” For organisations struggling to keep up with the pace of AI innovation, that’s a welcome maturing of the platform.

The key watchpoint is how deeply customisable the approval workflow will be. Will admins be able to tag requests by sensitivity label, set automated conditional approvals, or integrate with third-party GRC tools via Graph API? Microsoft hasn’t shared those details, but given the long development runway, we expect a robust set of controls.

Once the feature lands, it could become a model for other blocked resources in Microsoft 365, from SharePoint apps to Viva Engage communities. The idea of democratising access requests while keeping a tight rein on what’s actually deployed is a trend we’ll see more often. For now, July 2026 is a dot on the horizon—plenty of time for IT teams to prepare.