Dynamics 365 Contact Center: Automate Overflow When Queues Close or Agents Sign Out

Microsoft is addressing this exact operational blind spot in Dynamics 365 Contact Center under AI at Work roadmap ID 576759, titled "Dynamics 365 Contact Center: Unified routing - Handle waiting conversations when queues close or become unstaffed." Rather than evaluating operating hours and staffing levels only at the moment a conversation first enters a queue, this capability continuously monitors conversations that are already waiting and triggers automated fallback actions the moment a queue closes or loses all signed-in representatives.

Here is a full technical breakdown of how the capability works, how it fits into Microsoft's conversation orchestration architecture, and what administrators need to know before putting natural-language routing playbooks into production.

Release Status and Timeline: What Admins Need to Know​

Before diving into configuration menus, enterprise administrators should note the exact deployment state of this feature. According to Microsoft's AI at Work roadmap (ID 576759, published October 16, 2026) and the archived Dynamics 365 2026 release wave 1 plan on Microsoft Learn, the feature is listed with a Preview availability date of June 2026 (specifically June 20, 2026 in the release plan) and a General Availability target of October 2026 for Worldwide (Standard Multi-Tenant) cloud instances on the Web platform.

However, Microsoft's roadmap record still carries an In development status as of mid-October 2026, and Microsoft Learn's release-plan documentation explicitly notes that delivery timelines are projections subject to change. Furthermore, Microsoft's official technical documentation for Conversation Orchestration in the Copilot Service admin center currently carries a Preview label. Administrators should verify whether the specific templates have rolled out to their tenant ring before building production dependencies around them.

Two Core Scenarios: Closing Time and Unstaffed Queues​

Standard unified routing in Dynamics 365 Contact Center operates in two main phases: classification (where rules and machine learning models enrich incoming work items with skill requirements, urgency, and customer context) and assignment (where prioritized work items are matched to representatives based on skills, capacity, and presence). Historically, if a work item made it into a queue when representatives were at capacity, it waited until capacity freed up or an elapsed wait-time overflow timer expired.

Roadmap item 576759 adds state-aware lifecycle monitoring around that waiting period for two specific conditions:

1. Queue Transitions to Out-of-Operating Hours​

When a customer enters a queue before closing time, the system continuously monitors the active waiting conversation against the queue's schedule. If the clock ticks past the configured operating hours while the conversation is still unassigned and waiting, the system automatically applies admin-defined overflow actions. Instead of leaving the caller in limbo or requiring supervisors to hold shifts late to manually clear the backlog, the conversation can be guided to:

• Leave a voicemail
• Request a callback
• Transfer to another available queue (such as an after-hours or follow-the-sun regional support desk)

2. All Service Representatives Are Signed Out​

Even during normal business hours, a queue can suddenly become unstaffed. An unexpected network outage, a building evacuation, an unscheduled all-hands meeting, or a local holiday that an admin forgot to add to the operating-hours calendar can result in every assigned representative logging out.

Instead of forcing the customer to sit through a 10- or 15-minute in-queue wait-time threshold before overflow kicks in, the system evaluates real-time representative sign-in status across the queue. If all assigned representatives are signed out—including when the last representative logs off while conversations are already waiting—the fallback action triggers immediately.

• Section Summary: Rather than relying solely on entry-time checks or long wait-time timers, Dynamics 365 Contact Center monitors already-waiting conversations in real time, triggering immediate fallback routing when a queue crosses outside its operating hours or when the last assigned representative signs out.

Under the Hood: Natural Language Playbooks in Conversation Orchestration​

How do administrators actually build these rules? Rather than wiring together deeply nested conditional trees in legacy rulesets, Microsoft routes these capabilities through Conversation Orchestration using natural-language playbooks.

According to Microsoft Learn's conversation orchestration documentation, an administrator defines a playbook by selecting a guided prompt template and specifying a trigger event, conditions, and actions. When the administrator publishes the playbook, a language model converts the natural-language instructions into a structured runtime format that evaluates active conversations in real time.

Prerequisites for Deployment​

Before you can configure these playbooks in your environment, Microsoft Learn specifies several mandatory prerequisites:

1. Licensing and Billing: Specific Dynamics 365 Contact Center licensing rules apply. If your organization uses the Power Platform Pay-as-you-go plan, you must link an Azure subscription with consumption-based billing configured, as the system charges the subscription when the orchestration agent runs.
2. Permissions: You must hold the System administrator or Omnichannel administrator security role.
3. Routing Foundation: Unified routing must be enabled in the environment, with at least one voice or messaging channel, workstream, and queue already configured. (Note: Unified routing requires representatives to use the Copilot Service workspace app and is not supported in the legacy Unified Service Desk.)

Step-by-Step: Configuring Overflow Playbooks​

If your tenant has Conversation Orchestration enabled, you can configure these two queue-handling scenarios through the Copilot Service admin center:

1. Navigate to Conversation Orchestration: Sign in to the Copilot Service admin center and go to Customer support > Conversation Orchestration (Preview).
2. Open the Prompt Gallery: Select Prompt gallery to view guided templates across scenarios.
3. Select the Overflow Template: Under the Overflow handling category, choose one of the two templates matching your goal:
   • Configure overflow when queue goes out of operating hours
   • Configure overflow when all users are logged out
4. Define the Queue Scope: In the Apply to field, choose whether the playbook governs All queues (for the selected channel), a specific List of queues, or All queues except specific exclusions, then click Save.
5. Add Context Variables (Optional): If you want different fallback treatments for different callers—for instance, routing Gold-tier customers to a 24/7 escalation queue while offering standard callers a callback—select Add customer attribute under Add variables and choose attributes such as CustomerTier, Is VIP Customer, Preferred Language, or Customer Region.
6. Configure Conditions and Actions: Refine the natural-language conditions and specify the exact fallback action (such as transferring to a backup queue, offering a callback, or routing to voicemail).
7. Save and Validate: Click Save to store the playbook in Draft status. Review any validation warnings, and click Publish to switch the playbook to Active.

• Section Summary: Administrators configure both features inside the Copilot Service admin center using guided natural-language playbook templates that apply to voice and messaging queues, with optional context variables for granular customer treatment.

Critical Operational Cautions for IT Admins​

While natural-language configuration sounds effortless on a slide deck, veteran systems administrators know that AI-translated business logic requires strict guardrails. Microsoft's own documentation highlights four crucial operational realities you must plan around:

• Language Model Conversion Risks: Because a language model translates your natural-language playbook into the structured runtime format upon publishing, Microsoft explicitly warns that the converted output might not always capture your exact intended logic—and deviations only become apparent during live runtime. Always test playbooks thoroughly in a non-production sandbox with test conversations before publishing to live customer queues.
• Cross-Playbook Publishing Conflicts: You can run multiple active playbooks on the same queue if they cover different scenarios (for example, pairing a "Dynamic Prioritization - Wait Time" playbook with an "Agent Availability Based Overflow" playbook). However, you cannot publish two active playbooks for the same queue and the same scenario. Importantly, the system only checks for cross-playbook conflicts when you hit Publish, not when you save a Draft.
• Immediate Stop on Deletion: Editing an active playbook requires clicking Save & publish (the plain Save button is disabled to prevent accidental live changes). If you delete an active playbook, conversation orchestration for that queue stops immediately.
• ALM Solution Migration Dependencies: Playbooks can be exported and imported across environments using Power Platform Solutions. However, a playbook references underlying tables like queues and context variables. If those exact queues and context variables do not already exist in the destination environment prior to import, the solution import will fail.

Finally, keep compliance in mind: Microsoft notes that conversation orchestration is designed for operational routing and customer satisfaction, not for automated employment, compensation, or disciplinary decisions regarding service representatives, and organizations must adhere to applicable call-monitoring and user-consent laws.

By pairing continuous queue state monitoring with structured fallback actions, Dynamics 365 Contact Center gives IT and support leaders a practical way to eliminate end-of-day queue abandonment and staffing black holes—provided administrators test their natural-language playbooks rigorously before letting them run the floor.