Last updated: November 17, 2026
By ZapAI Team
Unified routing decides which queue and which person gets every case, email, chat and call in Dynamics 365 Customer Service. Our Dynamics 365 Customer Service consultants design and configure it for mid-market US support teams, and this guide covers workstreams, rules, assignment methods, capacity, skills, overflow, the new AI playbooks, licensing and migration, checked against Microsoft Learn on November 17, 2026.
Dynamics 365 unified routing sends cases, emails, chats and calls to the best-suited queue and representative. Workstreams take the work in, classification rules add skills and priority, route-to-queue rules pick a queue, and the queue’s assignment method picks the person.
Here’s a hypothetical you might recognize. A support manager turns on unified routing on a Friday. By Monday, priority cases sit in a fallback queue nobody watches. Two senior reps hold nine items each. The new hire has none.
Nothing was broken. Every rule did exactly what it was told. Our Dynamics 365 Customer Service guide covers the product at a high level, including plans and AI agents, while this post goes one layer down, into routing design and setup, because that’s where the rollouts we’ve reviewed tend to lose their first month.
What Is Unified Routing in Dynamics 365 Customer Service?
It’s one routing service for every channel. Microsoft’s unified routing overview describes two main stages, classification and assignment, and says the service weighs what a representative is already handling on other channels before it hands them new work.
That cross-channel view is the whole point. A rep on a phone call shouldn’t get a chat. A rep with three open cases shouldn’t get a fourth while a colleague sits idle.
| Building block | What it does | Where you set it |
|---|---|---|
| Workstream | Takes in one type of work (record, messaging or voice) and holds its routing rules | Copilot Service admin center |
| Intake rule | Picks the workstream for a routed record, such as a case | Record routing settings |
| Classification ruleset | Adds skills, priority, capacity profiles or custom attributes to the work item | Workstream |
| Route-to-queue ruleset | Picks the queue, with optional percentage splits | Workstream |
| Queue | Holds work and the reps who serve it, plus assignment method, operating hours and overflow | Queue settings |
| Assignment method | Chooses the rep inside the queue | Queue |
| Capacity profile | Caps how much work each rep can hold | User management |
One prerequisite surprises teams. Microsoft says reps must use the Copilot Service workspace app to receive work through unified routing. Unified Service Desk isn’t supported.
How Does a Work Item Move Through the Routing Pipeline?
In five steps. Order matters at every one.
- Intake. For records, intake rules choose a workstream. Chat and voice channels already belong to one.
- Classification. Rulesets run in order and enrich the item with skills, priority, a capacity profile or custom attributes.
- Route to queue. A single ruleset picks the queue. No match sends the item to the workstream’s fallback queue.
- Prioritization. Inside the queue, work is first in, first out unless you add a prioritization ruleset.
- Assignment. The queue’s method finds an eligible rep with the right skills, presence and capacity.
Two route-to-queue behaviors are worth memorizing. Per Microsoft’s route-to-queue guide, the default hit-all policy evaluates every rule plus each queue’s overflow state, then picks the first matching queue that isn’t overflowing. Hit-first stops at the first match. Overflow doesn’t count.
Percentage-based routing lives in the same place. You can split one rule across up to five queues with allocations that add up to 100, which Microsoft positions for spreading work across contact center vendors. Expect small deviations at low volume, since Microsoft says the gap narrows as item counts grow.
Keep conditions lean. Microsoft recommends fewer than 15 link-entity elements in routing rule conditions to avoid time-outs, and its best practices page suggests checking static values like VIP status once during classification, then reusing that attribute in later rules.
How Is Record Routing Different From Chat and Voice Routing?
Records move slower and fail quieter. Microsoft documents a periodic five-minute assignment trigger for record work items, on top of event triggers such as a new arrival or a rep’s presence change, so a case can sit a few minutes before anyone sees it.
| Factor | Record workstream | Messaging workstream | Voice workstream |
|---|---|---|---|
| Work arrives through | Intake rules after a record is created or routed | Chat widget, SMS, social and Teams channels | Support phone numbers |
| Distribution mode | Push or pick | Push or pick | Push only |
| Least active method | Not available | Available | Available |
| Pre-queue overflow conditions | Out of operating hours only | Operating hours and work item limit | Operating hours, work item limit and average wait time |
| In-queue wait threshold range | 3 minutes to 2 days | 1 minute to 2 days | 1 to 60 minutes |
| Capacity released when | A case resolves or is canceled. Emails need manual release | The conversation ends and the session closes | Same as messaging, subject to wrap-up blocking |
| AI agent on the workstream | Microsoft recommends against it | Push workstreams only | Push workstreams only |
Capacity release is the record-routing trap. Microsoft’s record routing setup guide covers intake rules and work distribution, and the capacity docs add the catch. Email activities routed as records don’t release capacity by themselves, so a rep fills up quietly unless someone deactivates the queue item or clears the Worked By field.
Supervisors get one useful toggle. Turn on Consume capacity on supervisor case assignment and a manual Assign moves capacity from the old rep to the new one. It’s off by default.
Plan the cutover too. After you enable unified routing, Microsoft says the active basic routing rule won’t route records until you configure intake rules. Chat and voice channel setup belongs to Contact Center, which our sister post on Dynamics 365 Contact Center covers.

Which Assignment Method Should Each Queue Use?
Start with a built-in method. Microsoft’s assignment methods guide, updated October 30, 2026, says to prefer the defaults because they’re validated for most use cases.
| Method | How it picks a rep | Use it when |
|---|---|---|
| Highest capacity | Most available capacity, with ties broken by round robin | Case queues with mixed complexity where the least-loaded rep should go next |
| Advanced round robin | Rotates by last assignment time within each queue, still checking skills, presence and capacity | Reps have similar skills and even counts matter to the team |
| Least active | Longest time since capacity was last released across voice and messaging. Voice and messaging queues only, and the default there | Phone and chat teams where idle time matters more than open counts |
| No auto assignment | Work waits until a rep picks it or a supervisor assigns it | Specialist or back-office queues where people choose their own work |
| Custom | Your prioritization and assignment rulesets, with order-by options like proficiency or capacity | Defaults can’t express a rule, such as matching a rep’s region to the customer’s |
| Percentage split | Not an assignment method. A route-to-queue option across up to five queues | Outsourcer or vendor splits |
Custom methods carry three caveats from the same page. Workstream defaults for presence, capacity and skill matching aren’t applied, so you write those conditions yourself. Default methods ignore rep operating hours, and you need the is working operator on calendar schedules to respect shifts. And each queue gets exactly one prioritization ruleset.
Cross-queue priority is simple once you see it. Higher-priority queues assign first. When priorities tie, queues on default methods go ahead of queues with custom prioritization rules.
How Do Capacity Profiles and Presence Decide Who Is Eligible?
A rep is eligible only if their presence is on the workstream’s allowed list and they have capacity left. Skills come after that.
Microsoft’s capacity profiles guide recommends using either capacity profiles or unit-based capacity, not both. Profiles suit mixed work. Each sets a work item limit, a reset frequency of immediate or end of day, and an assignment blocking option that stops all automatic assignment for that rep once the limit is hit.
Stack profiles for layered caps. Microsoft’s example sets a total profile of five cases a day on the workstream, then uses classification to add a high-priority profile of two or a normal profile of four. The rep needs room in both profiles to get the item.
Two settings get missed. Reset frequency can’t be changed later without recreating the profile, and custom per-rep limits set through enhanced user management can take up to 15 minutes to sync before routing honors them.
Presence is the other half. Leave Inactive and Do not disturb off the allowed list if reps who miss notifications shouldn’t keep getting work, and set a default presence at sign-in so every shift starts in a known state. Microsoft’s best practices page also asks reps and supervisors not to flip presence by hand, so assignment cycles run without interruption.
Wrap-up blocking costs real capacity. Always block holds it until the conversation closes. Don’t block frees it the moment wrap-up starts. Custom time sits between, at 30 seconds to 60 minutes.
How Do Skill-Based Routing and the Intelligent Skill Finder Work?
Classification attaches skills. Assignment matches them. You build a rating model, skill types and skills, give them to reps, then choose exact or closest match on the workstream.
Exact match keeps only reps whose proficiency meets or beats the requirement. Nobody qualifies? The item waits. Closest match ranks everyone by how near they are, exact first, then overqualified, then underqualified, and Microsoft notes it might assign someone with none of the required skills. Speed versus precision. Choose per workstream.
Microsoft’s best practices give a clean example. A team handling delivery issues and refunds, where most reps know one and some know both, can run one queue with skill matching instead of two separate queues.
Skills can come from rules or from machine learning. The intelligent skill finder trains an AI Builder category classification model on your historical records, needs at least 50 approved training records, and accepts Excel imports if you have no skill history. Three caveats apply. AI Builder preview models must be enabled, AI Builder must be available in your region, and a model trained in a test environment works only in another test environment.
Our practice opinion? Start with deterministic rules. Add the skill finder after six to eight weeks of reps correcting skills at runtime, because Microsoft suggests retraining on exactly those corrections.

When Should You Use Preferred Representative Routing?
For relationship accounts. You map a contact to up to three preferred reps in order, per Microsoft’s preferred representative routing guide, and the system tries them before the normal assignment method.
Read the fine print. Preferred routing checks allowed presence but skips capacity, skills and assignment rules, works only on push workstreams, needs the rep to belong to the queue the item lands in, and maps to contacts only. For records, the contact must sit in the case Customer field or the activity’s Regarding field.
Fallback is a choice. Next best representative suits live chat and voice. Leaving work unassigned in the queue suits asynchronous channels.
A newer route goes further. The Assign to a previous or preferred expert playbook in conversation orchestration adds previously engaged reps with a recency window of up to 30 days, then falls back to the queue’s method.
How Do Overflow and Operating Hours Work?
Overflow fires before an item is queued, after it’s queued, or both. Microsoft’s overflow guide, updated November 16, 2026, lists what each channel supports.
- Records get one pre-queue condition, out of operating hours, with assign anyway or transfer to another queue.
- Messaging adds a work item limit between 1 and 100, with end conversation or transfer.
- Voice adds average wait time, plus callback, voicemail and transfer to an external number.
- Once queued, all three use a wait time threshold. Records can only transfer.
Operating hours have a quirk. Add an operating hours record to a queue and the system creates an Out of operation hours condition with the action Assign to queue anyway. Fallback queues ignore overflow settings entirely, so an after-hours plan that relies on the fallback queue never fires.
VIPs need rule-specific overflow. On a route-to-queue rule, you can override the queue’s overflow behavior so a priority caller isn’t sent to voicemail. Pre-queue checks also run only once per item.
Legacy in-queue overflow can now be upgraded to a conversation orchestration playbook, but upgraded queues stop using the legacy setting immediately and Microsoft says unpublishing the playbook doesn’t bring legacy overflow back. Test in a sandbox first.
Which AI and Agent Routing Features Arrived in 2025 and 2026?
Three matter for routing design. Their status differs.
| Feature | What it does for routing | Status we found in November 2026 | Cost note |
|---|---|---|---|
| Conversation orchestration playbooks | Natural-language playbooks for dynamic priority, overflow, preferred or previous experts and bullseye routing, on voice and messaging only | Documented on Learn without a preview label. Message center item MC1536059 says rollout began late October 2026 | Copilot credits used when you publish, not per conversation |
| Intent-driven routing | Uses intent, intent group and line of business from the Customer Intent Agent in classification and route-to-queue rules, for live chat, voice and case | Production-ready preview | Needs the Customer Intent Agent and pay-as-you-go billing |
| Case Management Agent | Creates cases from conversations and predicts case fields, such as category, that routing rules can read | Setup documented on Learn, with shadow mode still in preview | Charged through pay-as-you-go when the agent runs |
Playbooks are the big one. Per Microsoft’s conversation orchestration guide, a language model converts your plain-English playbook into a runtime format, and Microsoft warns the converted logic might not fully capture what you wrote. Deviations show up only at runtime. Test every playbook with real conversations before you trust it.
Credit use is narrower than people assume. Publishing consumes Copilot credits and a published playbook then handles any number of conversations at no extra credit cost, while upgraded overflow playbooks with one branch, only a wait time condition and no context variables use none at all. Budget the rest with our Copilot Studio pricing guide.
Intent-driven routing carries a production-ready preview label and replaces the older intent-based routing, so intent group to user group mappings no longer apply. Unified routing doesn’t detect intent itself for chat or voice. A Copilot agent on the workstream does that before escalation.
On intake, the Case Management Agent can predict and fill case fields, including on cases created from email by automatic record creation rules. Fields like category then become routing inputs, so test the timing in a sandbox, since we found no Learn page stating whether predictions land before or after intake rules run.
Agents answer better from clean articles, which our sister post on Dynamics 365 knowledge management covers. For a ranked view of which Copilot features pay off, see Copilot in Dynamics 365 use cases.
Which Licenses Do You Need for Unified Routing?
Customer Service Enterprise or Premium for record routing, and Premium or Dynamics 365 Contact Center for chat and voice. That’s our reading of Microsoft’s Customer Service pricing page on November 17, 2026. Learn itself only says specific licensing requirements apply and points to the licensing guide.
| Plan | US list price per user per month | Unified routing | Chat and messaging | Voice |
|---|---|---|---|---|
| Customer Service Professional | $50 | Not included | Not included | Not included |
| Customer Service Enterprise | $105 | Included | Not in the plan | Not in the plan |
| Customer Service Premium | $195 | Included | Included | Included, with Azure Communication Services billed separately |
| Dynamics 365 Contact Center | $110 | Not named, voice routing listed | Included | Included |
| Contact Center Digital or Voice | $95 each | Not named | Digital plan only | Voice plan only |
Contact Center carries a discount when bought with Customer Service Enterprise, per Microsoft’s Contact Center pricing page. Agents and playbooks bill separately through Copilot credits and pay-as-you-go, and conversation diagnostics adds Azure Monitor ingestion costs that grow with every routed item you log. Confirm the exact SKU against the licensing guide before you sign.
Watch the app, too. Microsoft removed the Customer Service Hub app for new organizations with Enterprise licenses in February 2025, so new Enterprise orgs work in Copilot Service workspace. Our Dynamics 365 overview shows where Customer Service sits next to Sales and Field Service.
How Do You Migrate From Basic Routing Rules and Legacy Diagnostics?
Basic routing rulesets still work, but they route cases only, and Microsoft recommends unified routing for automatic record routing. Automatic record creation rules in Unified Interface still work as well. The 2020 deprecation covered the old web client version, per Microsoft’s Customer Service deprecations list.
Diagnostics is the real deadline. That same list says routing diagnostics in unified routing was deprecated on May 9, 2024, with removal set for March 31, 2026, and historical analytics dashboards for unified routing lost support on August 31, 2025. The replacement is conversation diagnostics in Azure Application Insights.
Setting it up takes Azure work. Per Microsoft’s conversation diagnostics guide, you need Contributor access in Azure, a Managed Environment and an Application Insights resource, and data export isn’t available in GCC, GCC High or DoD. The Diagnose command in Copilot Service workspace then shows classification, route-to-queue and assignment events for a conversation or case. Microsoft’s worked example lands near 10 KB per routed item.
Our migration sequence looks like this.
- Inventory active basic routing rules, ARC rules, flows that assign owners and plugins that set queues.
- Rebuild the logic as intake, classification and route-to-queue rules in a sandbox.
- Load users in Microsoft’s recommended order. Enable users, add them to queues, create bookable resources, then skills, capacity profiles and roles.
- Connect Application Insights before go-live.
- Move configuration between environments through your ALM pipeline.
- Cut over one queue at a time. Watch fallback queues daily.
Bulk loads have a pace limit. Microsoft suggests 500 change requests every 15 minutes, warning that faster loads can leave skills or profiles missing. Flows that set owners after creation will fight unified routing, so retire or rewire them, and our Power Automate use cases post shows patterns that coexist well with routing.
Playbooks move through ALM too, but import fails if the queues and context variables they reference don’t exist yet in the target. Our Power Platform ALM guide covers pipelines, and our customization best practices explain why routing logic belongs in configuration rather than plugins.
What Mistakes Break Unified Routing?
In our experience, these nine cause most bad first weeks.
- A fallback queue with no members, or members nobody monitors.
- Capacity profiles and unit-based capacity mixed in one org.
- Allowed presences that include Do not disturb.
- Email records routed with no plan to release capacity.
- Custom assignment methods that forget presence and capacity conditions.
- Assuming default methods respect rep shifts. They don’t.
- After-hours plans built on the fallback queue, which ignores overflow.
- Reps without permission on the routed record. Microsoft says assignment then stops and the conversation closes.
- Testing as System Administrator.
Permissions deserve their own pass. Our Dataverse security roles guide explains why copied roles drift, and our Dynamics 365 implementation checklist shows where routing testing fits in a go-live plan.

What Belongs on a Unified Routing Setup Checklist?
Ten steps, in order.
- Confirm licenses for every rep and channel, plus pay-as-you-go billing if you plan playbooks or agents.
- Enable unified routing with an admin who holds System Administrator on the root business unit.
- Create users as bookable resources, then add queues, skills, capacity profiles and roles.
- Build queues with members, priority, operating hours, assignment method and overflow.
- Enable record types in the records channel and write intake rules.
- Create workstreams with fallback queue, capacity, allowed presences and skill matching. Distribution mode can’t be edited later.
- Add classification rules, then route-to-queue rules.
- Connect conversation diagnostics.
- Test with non-admin rep accounts across shifts and presences.
- Go live one queue at a time and review fallback and overflow volume weekly.
ZapAI Decision Rules for Unified Routing
These are our practice opinions from projects, not Microsoft guidance. Treat them as defaults.
| Situation | Our rule |
|---|---|
| Fewer than about 15 reps working cases only | Highest capacity, exact match and one capacity profile. Skip custom methods |
| Phone and chat teams | Least active, with wrap-up blocking set to a custom time near real wrap-up |
| Two skill sets with some overlap | One queue with skill matching instead of two queues |
| Named account managers | Preferred representative routing with next best fallback on live channels |
| Outsourced overflow partner | Percentage split in route-to-queue, reviewed monthly |
| No skill history yet | Rule-based skills first, skill finder after runtime corrections pile up |
| Leadership wants AI routing now | Playbooks for overflow first, intent-driven routing in a sandbox while it’s in preview |
One rule beats the rest. If a supervisor can’t explain why an item went to a person, the design is too complex.
Wrapping Up
Get eligibility right before cleverness. Capacity, presence and queue membership decide who can get work, rules decide where it goes, and the method picks among the eligible. Connect diagnostics early, staff the fallback queue, and test as a real rep.
ZapAI is a Microsoft Dynamics 365, Power Platform, Fabric and AI consulting firm in Irving, Texas, working with businesses across Dallas-Fort Worth and nationwide. Our Customer Service team designs and configures routing for mid-market support teams. Want a second opinion? Ask for a routing design review, and we’ll walk through your workstreams, queues and rules with you.
Questions Support Leaders Ask About Unified Routing
Is unified routing included in Customer Service Professional?
Not per Microsoft’s pricing table, which lists it for Enterprise and Premium. Professional orgs can still route cases with basic routing rulesets.
Can unified routing route records other than cases?
Yes. Record workstreams handle cases, emails, other activities and tables you enable in the records channel configuration. Basic routing works for cases only.
Why did a case land in the fallback queue?
Either no route-to-queue rule matched or an error occurred. Open Diagnose on the case in Copilot Service workspace, which needs conversation diagnostics connected to Application Insights.
Do default assignment methods respect rep shifts?
No. Microsoft says default strategies don’t consider rep operating hours. Use a custom assignment method with the is working operator on calendar schedules, or route on schedules imported from a workforce management system.
What’s the difference between highest capacity and least active?
Highest capacity picks the rep with the most free capacity. Least active picks whoever has gone longest since their capacity was last released, and it’s only available for voice and messaging queues.
Do conversation orchestration playbooks cost money per conversation?
Not per Microsoft. Credits are consumed when you publish a playbook, and a published playbook handles any number of conversations without extra credits. Consumption-based billing must be set up first.
Is the old routing diagnostics feature still available?
Microsoft deprecated it in May 2024 and set removal for March 31, 2026. Use conversation diagnostics in Azure Application Insights instead.
