Microsoft’s Copilot call delegation for Teams Phone became available in April 2026 for Frontier-enrolled organizations with Microsoft 365 Copilot licensing, letting an AI voice agent answer incoming calls. The feature signals a strategic shift in how enterprises handle voice communication, blending generative AI with telephony in a way that promises efficiency but demands rigorous governance. For IT admins and business leaders, this is not just another feature update—it is a fundamental rethinking of the receptionist role.

Copilot call delegation builds on the existing Teams Phone capabilities by introducing an agentic AI layer. Instead of simply forwarding calls to a human delegate, the AI can engage callers in natural conversation, determine intent, and take action—scheduling a callback, transferring to a specific department, or even answering FAQs. Microsoft positions it as a way to reduce missed calls and offload routine reception tasks, but the autonomy it grants to an AI raises critical questions about oversight, accuracy, and data governance.

How Copilot Call Delegation Works

The feature is tied to Microsoft 365 Copilot and requires enrollment in the Frontier preview channel, which gives early adopters access to cutting-edge features. Once enabled, administrators can configure call delegation policies for specific users or Teams Phone numbers. When a call arrives, the AI agent picks up, using a natural-sounding voice to interact with the caller. It leverages the same large language models that power Copilot across Microsoft 365, trained to understand context and generate relevant responses.

Behind the scenes, the AI works with Teams Phone’s call routing and compliance recording systems. Callers might not even realize they are talking to an AI, as the responses are designed to be conversational and adaptive. The agent can be programmed with specific business rules—for example, if the caller mentions “sales,” it connects to the sales queue; if they ask for office hours, it reads the company’s schedule. It can also take messages and transcribe them into Teams chats or emails.

But the autonomy goes further. The AI can make decisions about whether to interrupt the recipient, offer a callback slot based on calendar availability, or even answer simple questions by pulling data from connected SharePoint sites or CRM systems—provided those integrations are set up. This agentic behavior is what sets it apart from traditional auto-attendants and IVR trees.

The Agentic Receptionist: Hype vs. Reality

“Agentic reception” is Microsoft’s term for an AI that doesn’t just respond but acts. In practice, that means the Copilot call delegate can initiate tasks on behalf of the caller, such as creating a support ticket in Dynamics 365 or finding the nearest store location. For small businesses, this could replace a live receptionist entirely for after-hours calls. For large enterprises, it adds an intelligent layer on top of contact centers.

Yet the reality is more nuanced. Voice AI is tricky—accents, background noise, and unexpected questions still trip up even the best models. Microsoft has baked in guardrails; the AI will gracefully hand off to a human when it detects confusion or sensitive topics like healthcare or legal emergencies. But those guardrails are only as good as their configuration, and that falls on the organization’s IT and compliance teams.

Early feedback from the Frontier community indicates that while the voice interaction feels polished, the setup is complex. Admins need to craft detailed prompts, test extensively, and continuously monitor logs to catch misrouted calls or inappropriate responses. One tester noted that the AI once misinterpreted a caller’s request for “billing” as “Billing”—a person’s name—and tried to transfer to a non-existent extension, highlighting the importance of robust testing.

Governance and Licensing Requirements

Copilot call delegation is not a free add-on. It requires an active Microsoft 365 Copilot license for each user who will use the feature as a delegate, plus the standard Teams Phone license. For organizations not already on Copilot, this adds significant cost. Licensing is per-user, so every employee who wants an AI receptionist for their line must be covered. Volume licensing and enterprise agreements may offer discounts, but Microsoft has not announced any standalone SKU.

Governance extends beyond cost. Since the AI processes call audio and may access sensitive data, compliance with regulations like GDPR, HIPAA, and PCI DSS becomes paramount. Call recordings and transcriptions are stored in Microsoft’s cloud, so data residency requirements must be reviewed. Microsoft says data is encrypted in transit and at rest, and that the AI does not use call data to train its models, but customers must still perform their own due diligence.

Administrators can enforce policies via the Teams admin center and Microsoft Purview. They can restrict which data sources the AI can query, define escalation paths, and audit every interaction. Role-based access control ensures that only approved personnel can modify delegation settings. However, the onus is on the organization to configure these controls—out of the box, the defaults may be too permissive for regulated industries.

The Risk Landscape

Every autonomous system introduces risk, and an AI answering your business calls is no exception. The most obvious risk is hallucinations—the AI generating false information with convincing delivery. If a caller asks about a return policy and the AI fabricates a deadline, the company could be on the hook for honoring it. Reputation damage is another concern; a tone-deaf AI response in a sensitive situation could go viral.

Security is a shared responsibility. The AI telephone agent becomes a new attack surface. Malicious callers could try prompt injection—phrasing requests in ways that trick the AI into revealing confidential data or performing unauthorized actions. Microsoft has implemented input filtering and output monitoring, but sophisticated attacks will always be a cat-and-mouse game.

Privacy risks are amplified when the AI handles personal information. A call about medical results or financial details requires the AI to recognize the sensitivity and route accordingly, but misclassification is possible. Moreover, retained transcripts become discoverable in legal cases, so data retention policies must be carefully aligned with legal holds and minimization principles.

Availability is another vector. When AI call delegation goes down due to an outage or misconfiguration, callers hit a dead end. Unlike a traditional IVR that simply fails open, the AI might simply hang up or loop indefinitely. Redundancy and failover mechanisms are essential—yet they add complexity and cost.

Preparing Your Organization for AI Call Delegation

Adopting Copilot call delegation should start with a pilot limited to non-critical internal lines. Test with a broad spectrum of caller personas, accents, and scenarios. Collect feedback from actual callers, not just IT staff. Use the telemetry from Teams to identify patterns and refine prompts. Microsoft provides a set of predefined prompts as a starting point, but customization is where the value—and the bugs—emerge.

Governance teams need to be involved from day one. Draft acceptable use policies that clarify who can configure the AI, what tasks it can perform, and how data is handled. Consider a human-in-the-loop approach for high-risk calls: the AI can handle the initial interaction but quietly alert a human supervisor if confidence is low. Unfortunately, Microsoft’s current implementation does not natively support a seamless agent desktop for such supervision, third-party integration may be needed.

Cost-benefit analysis must account for more than just reduced receptionist headcount. Factor in Copilot licensing, setup time, ongoing monitoring, and the potential for direct and indirect losses from AI errors. For some organizations, the ROI will be immediate—law firms with high call volumes may recover billable hours. For others, the risk may outweigh the benefit, especially until the technology matures.

The Competitive Landscape and What Comes Next

Microsoft is not alone. Competitors like Google with Vertex AI Conversation and Amazon Connect already offer AI-powered voice agents. However, Microsoft’s integration with Teams and the Microsoft 365 ecosystem gives it a distinct advantage for organizations already invested in Office apps. The ability for the AI to access Outlook calendars, Teams chats, and SharePoint documents creates a unified experience that standalone solutions struggle to match.

Looking ahead, expect tighter coupling with Copilot for Sales and Copilot for Service. A customer calling a sales number might be automatically routed to a pre-created deal room in Dynamics 365, with the AI having already sent the caller relevant brochures via email during the call. Microsoft is also likely to expand the feature to outbound calling, where the AI handles appointment reminders or preliminary qualification before a human takes over.

On the governance side, Microsoft will need to provide more granular controls. Today, many policies are binary—the AI is either on or off. Future releases should allow conditional access based on caller identity, time of day, or content sensitivity. Without such controls, adoption in regulated industries will stall.

The Bottom Line

Copilot call delegation is not just another Teams feature—it’s a test case for agentic AI in the enterprise. The April 2026 rollout marks the beginning of a long journey toward truly autonomous corporate communication. For early adopters, the promises of efficiency and 24/7 availability are compelling, but they come with a hefty governance bill. The organizations that succeed will be those that treat the AI not as a set-and-forget tool, but as a new employee requiring training, supervision, and continuous improvement. As the Frontier community has already discovered, the path to agentic reception is paved more with careful planning than technological wizardry.