A supervisor sees three cases waiting while two experienced agents appear unavailable. Dynamics 365 may seem at fault, but a profile could be full, the work might need two profiles, or a completed record may not have released capacity.
Treat capacity profiles as routing eligibility, not a staffing forecast or agent score. Use them when work types need separate limits or one channel must stop other assignments. Pick one model, test the complete path, and expose releases and overrides.
Decide whether capacity profiles fit the problem
Microsoft says capacity profiles define the types and amount of work a representative can take. They can set concurrent or daily limits, vary by person, and block extra assignments when a limit is reached. Microsoft also recommends using either profiles or unit based capacity in one environment, not both. The full behavior is documented in its guide to creating and managing capacity profiles.
- Choose one modelUse profiles or units in the environment so every assignment follows one capacity logic.
- Define real workBase each profile on work behavior, required attention and the event that proves completion.
- Match every profileConfirm the work item and eligible representatives share all default and conditional profiles.
- Test exceptionsExercise manual picks and forced assignments and record how overload becomes visible.
- Prove releaseClose representative work correctly and verify that capacity returns before launch.
Profiles help when one company wide number is misleading. Voice work may block other assignments, specialist investigations may need a daily ceiling, and learners may need lower limits.
Do not use a profile to invent the limit. First estimate a safe workload from active attention, complexity, response promises, and quality evidence. Our guide to support agent conversation capacity explains that operating decision. The profile is how you enforce part of it in routing.
Map the work before you configure anything
Map real work first. For each type, name the customer outcome, required skill, likely duration, effect on other channels, and completion event. If two categories behave alike, one profile may be enough.
| Operating situation | Better control | Main check before launch |
|---|---|---|
| Similar work with one shared load | One unit model or one general profile | Confirm every item consumes and releases capacity consistently |
| Specialist work needs its own ceiling | Default profile plus a conditional specialist profile | Confirm the item and eligible people carry both profiles |
| One channel must stop other automatic work | Assignment blocking on the relevant profile | Confirm presence changes and waiting work remain visible |
| People need different limits for the same work | A representative specific custom limit | Confirm the change has synchronized before judging routing |
| A rare emergency exceeds the limit | Named supervisor override | Record the reason, owner, customer risk, and recovery step |
Do not create a profile for every queue. Express a workload boundary operators can explain. Queue membership, skill, presence, and capacity answer different questions.
Test the whole assignment chain
A workstream can apply one default capacity profile to all incoming work. Classification rules can append another profile for a specific condition. Dynamics 365 then looks for a representative with available capacity in every required profile. This is powerful, but it also creates a predictable failure mode. If the person or work item is missing one profile, the case can wait even when the dashboard appears to show spare capacity elsewhere.
Before production, test an ordinary case, a specialist case, full and available representatives, and someone without the conditional profile. Check classification, required profiles, assignment, and remaining capacity.
Test custom limits separately. Microsoft notes that they affect new assignments, do not move existing work, and can take time to synchronize. A lower new limit may therefore coexist with current consumption above that limit. That is a transition state, not proof that the setting was ignored.
Keep the test close to the customer workflow. A shared inbox helps an operator see ownership and waiting work, while automation rules can support predictable classification and handoff. Neither replaces the Dynamics configuration test.
Treat manual work as an exception
Manual action can bypass the guardrail. Microsoft's queue guidance says a manual pick does not evaluate schedule, assignment rules, skills, presence, or capacity constraints. It can reduce available capacity and may leave a person with negative capacity. Read the current behavior in Microsoft's queue operations documentation.
Keep that override for contingencies, not daily queue clearing. Require a reason and owner. Record the person's skill and authority, delayed work, and when load returned to normal.
Review overrides as system evidence. Frequent manual picks may mean the profile map is too strict, skills are missing, work is classified badly, or staffing does not match demand. Raising every limit hides the distinction.
Release capacity when work really ends
Profiles only work if consumption is released reliably. Dynamics 365 releases conversation capacity when the representative ends the conversation and closes the session. For cases, resolution, cancellation, or clearing the Worked By field can release it. Other routed records and activities may need the queue item deactivated or the representative removed from Worked By.
This is where apparently idle teams often lose capacity. The customer has an answer, but the operational record remains open. Define one closure event for each work type and audit stale items daily. Do not close customer work merely to improve availability. Confirm the promised outcome, required notes, next owner, and any follow up first.
A reliable support operation keeps queue state, customer state, and capacity state aligned. If a custom integration updates routing records, place it inside a reviewed developer workflow with clear failure handling and reviewed access controls. The safest setup is the one an operator can reconstruct after something goes wrong.
Judge the profile by customer outcomes
Capacity is not the final measure. Track the oldest eligible work, time to an accepted owner, reassignments, overrides, reopened cases, late follow ups, and quality failures by work type and profile.
If specialist work waits while general capacity is free, inspect profile matching before adding people. If agents show negative capacity, inspect manual picks and forced assignments. If capacity never returns, inspect completion and queue item state. If the configuration is correct but customers still wait, the problem may be demand, staffing, skill coverage, or the service promise itself.
Change one control at a time. Keep the previous value, the evidence, the expected effect, and a rollback point. If you need help mapping the routing boundary to a real customer journey, bring one representative queue to a DripTell review.
Frequently Asked Questions
Should we use capacity profiles or capacity units
Use one model in the same environment. Profiles fit separate work types, person specific limits, and cross channel blocking. Units fit a shared numerical workload model. Mixing both makes assignment and troubleshooting harder.
Why is work waiting when an agent has free capacity
The work may require a profile the agent does not have, or it may require capacity in both a default and appended profile. Also check presence, skills, queue membership, classification, and synchronization time.
Why does an agent show negative capacity
A manual pick or forced assignment can exceed the configured boundary. Record it as an override, check the customer risk, and decide whether it exposed a bad rule or a genuine contingency.
When should capacity be released
Release it when the product specific completion event occurs. That differs for conversations, cases, and other routed records. Verify the customer outcome and required record before closing anything.
Want help applying this guide?Ask the DripTell team
From reading to action
Want help applying this to your business?
Share your details. We will read the article you are on and start with a useful answer for your team.
DripTell Editorial
Practical guidance reviewed by the DripTell product and customer workflow team.
See how DripTell checks product claims, uses primary sources and handles corrections.
Editorial and source policy
