Updated 16 Oct 2026. Atlassian is renaming and shipping fast. Service Triage Agent has been renamed Request Router, and Workforce Management is now available in Jira Service Management. We update this page when that changes.
This is for IT and support managers, Jira admins, and engineering leads on teams of 20 to 100, and on Jira Service Management teams of 3 to 20. You have just seen native routing land. Do you still need anything custom?
For a lot of service spaces, no. Rovo picks the team, then Workforce Management assigns the least-loaded available, on-shift agent. Custom still earns its cost when the decision needs data the work item does not hold, or when the work is not in a service space.
Short answer
Four ways to route a request. Where the support pages we cite are silent, the cell says so.
| Workforce Management with Rovo | Rovo Request Router | Custom Rovo agent | External custom layer | |
|---|---|---|---|---|
| What it decides | The team, then an agent on that team | Request type, priority, and escalation suggestions | Whatever you instruct an agent to handle inside Rovo. No field list in the docs cited here | Team, assignee, and the follow-up steps you define |
| Inputs it reads | Work item details, team or service routing context, capacity, schedule, availability | Summary, description, comments, sentiment, SLAs, and similar past items | What you give that agent. Not specified in the pages cited here | The work item, plus systems outside it, such as a CRM, contract, or on-call tool |
| Plan needed | Services, capacity, and schedules are Premium and Enterprise only | Not stated in the pages cited here | Whatever plan includes the Rovo agent you configure. Not stated here | Your own. Compare an upgrade against a fixed build before you choose |
| Confidence control | A default team when Rovo cannot confidently identify a team. Auto-assign off shows recommended agents | Not stated in the pages cited here | Not stated in the pages cited here | You can hold a band for review in the system you own. Confirm the native controls first |
| Audit trail | Not described on the support pages cited here | Not described on the support pages cited here | Not described on the support pages cited here | The log you keep in the system you own |
| Works outside a JSM service space | No. Configured per service space. The assignee has to be an agent in that space | The docs cited here are for service requests. We do not claim Jira Software projects | It runs inside Atlassian. We do not claim it routes work outside a service space | Yes. A layer can span projects that are not service spaces |
| Cost model | Basic AI feature. 10 Rovo credits per billable event. Extra usage billed from 3 December 2026 | Not stated in the pages cited here | Rovo credits, plus the time to configure the agent | Fixed build from €5K, about two weeks, you own the code, plus your model costs. Not a claim that this is cheaper |
What Jira Service Management now does out of the box
Workforce Management "routes work to the right team – and the right agent within that team – using the work item's details and each agent's capacity, shift schedule, and availability." You configure it per service space. On Premium or Enterprise, workload is no longer a reason to commission a build.
Team assignment has three options: auto-assign the team with Rovo, auto-assign a default team, or assign each work item by hand. With Rovo, team routing means you describe what each team works on. Field or service routing maps affected services, an Assets object field, or a custom field to the owning team. Routing context can be typed in, or generated by Rovo from historical work items. A default team catches requests where Rovo cannot confidently identify a team.
An agent is eligible only if they belong to the owning team, are scheduled to be working, and have marked themselves available. Work goes to the eligible member with the lowest capacity utilisation. If nobody is eligible, the item stays unassigned. Auto-assign can be off, and then the Assignee field shows recommended agents. Triggers and the assignment factors are configurable. Manual override is always possible.
Request Router, formerly Service Triage Agent, suggests request type, priority, and escalation. That is classification, not assignment. You can take the suggestion and leave assignment off.
"Some features like services, capacity, and schedules are only available on the Premium and Enterprise plans." If Workforce Management does not appear in space settings, the plan may not include it. "Workforce Management is not HIPAA compliant."
What it needs from you before it routes anything
Native routing is configuration. It will not invent your teams. Before the first request is assigned, this has to exist:
- An Atlassian Team connected to the service space.
- A clear routing context for each team, or a service, Assets, or custom field that already maps to an owner.
- Schedules, availability, and capacity limits, on a plan that includes them.
- A default team for the requests Rovo cannot place.
- A person who owns the unassigned queue, because an item with nobody eligible stays unassigned.
Write each team in a sentence or two. Overlapping descriptions push work onto the default team. If Rovo drafts the context from old tickets, read the draft before you trust it. Eligibility depends on people marking themselves available. If they don't, the item waits. Treat that as part of setup, next to the team descriptions.
When native routing is enough
Stay on Workforce Management when all of these are true.
- The requests live in one Jira Service Management service space per team.
- Each team, or each service, can be described in a sentence or two, or you already have a services or Assets field that maps to an owner.
- You are on Premium or Enterprise if you want capacity, schedules, and services. If you are not, compare the upgrade with a custom build before you assume you need the build.
- The controls in the docs are enough for review: a default team when the team is unclear, recommended agents when auto-assign is off, and a manual override.
When a custom layer still earns its cost
We build what Rovo does not cover. Each case is only the part the docs support.
The work lives outside a Jira Service Management service space. Workforce Management is configured per service space, and the assignee has to be an agent in that space. A Jira Software project is outside that. The custom piece is a routing layer that spans those projects.
The decision depends on data outside the work item. The documented inputs are the work item, the team or service routing context, and capacity, schedule, and availability. A CRM tier, a contract, or an on-call tool is not in that list. We pull that context in first, or a rule sets Team so native assignment can run after.
You want capacity and schedules, and you are not on Premium or Enterprise. Those features are Premium and Enterprise only. A custom layer can assign by workload without the upgrade. Price it against the upgrade.
Regulated data. Workforce Management is not HIPAA compliant. If the work cannot go through that feature, routing has to run in an environment you already treat as compliant.
The step after routing leaves Jira. Automation can run alongside Workforce Management, for example to escalate items left unassigned. A Slack thread, a CRM update, or a weekly report outside Jira is further work. Check which of those steps a rule already covers. Where the step leaves Jira, we build it. See the Slack bot that routes Jira tickets and the Slack intake bot. If the queue is already in a service space, start native and add the bot only for the handoff.
The engagement for that custom layer is a pilot from €5K, about two weeks, and you own the code. The service behind it is Jira optimization: audit the workspace, decide native versus custom, and only then train a classifier on the signal native routing cannot see.
Rules first: what Jira Automation handles without AI
Atlassian suggests Automation alongside Workforce Management. Use that. Rules are the right tool when the mapping is fixed.
- Customer tier to priority.
- SLA timers, and the alerts around them. The practical version is in Jira SLA automation.
- Keyword to label, when the word really does mean one thing.
- Escalate an item that has sat unassigned for too long.
Use AI for the fuzzy part: a vague request whose team is not obvious. Rules plus AI is more reliable than AI alone. The rule catalogue is Jira automation rules. Leaving triage on a person has its own cost, in the cost of manual ticket triage. For the wider bill, see Jira automation cost, and for the service desk around this decision, Jira Service Management automation.
What we chose not to automate
We do not let a model reply to a customer, and we do not let it change status without a person. A wrong team can be moved. A wrong public reply sits in the thread. Keep replies and status changes behind approval, even after assignment has earned auto-apply.
Confidence and human review
The sequence we use is suggest, then review, then auto-apply. Workforce Management already has the first two switches.
Turn auto-assign off and the Assignee field shows recommended agents. A person decides. When the recommendations hold for a week or two, turn auto-assign on for the categories that held. The rest stay as suggestions.
When Rovo cannot confidently identify a team, the default team catches the request. That is the low-confidence path the docs describe. If you want a number you set yourself, confirm what your site offers before you commission one. Measure on your own tickets: route a labelled sample, read it per category, and automate only the categories that hold.
How we built a Jira routing system
We built workload-aware routing before Jira shipped it. The project is the Jira routing case. This is what the system did, and what we would reuse today. It is not an argument that native Jira cannot assign by workload. It can.
The layer sat on top of Jira. It read each new ticket's type, urgency, and content, routed it to the right team, and assigned by current workload. It watched SLA timers and escalated before a breach. It sent a weekly report on queue health, bottlenecks, and utilisation.
The bottleneck was never headcount. It was ticket routing. Escalation was built into the workflow, not the culture. Weekly reports created accountability without meetings.
Today we would start with Workforce Management and not rebuild the workload assignment. We would keep escalation on SLA timers, reports that leave Jira, and context the work item does not contain. Check which escalation steps a rule already covers. Atlassian already points at rules for items left unassigned.
Cost: Rovo credits on your plan vs a fixed-price build
Workforce Management routing is a basic AI feature. It uses Rovo credits: 10 credits per billable event. Allowances are pooled per organisation and depend on the app and the plan. Extra-usage billing takes effect on 3 December 2026. Admins can set spending limits, and there are alerts at 80% and 100% of the included allowance. Check the allowance in Atlassian Administration. The source is Rovo usage limits.
Atlassian counts each AI-feature invocation as a billable event and publishes per-plan allowances, but doesn't say what one invocation is for routing.
The custom shape is a pilot from €5K, about two weeks, you own the code, plus your own model costs. Credits scale with use. A fixed build does not. Which is smaller depends on volume, plan, and whether you needed the layer. If native routing covers the queue, the comparison is the credits, not the pilot.
Decision checklist
- Are your requests in a Jira Service Management service space? Start native.
- Can each team or service be described in a sentence or two? Use team routing or field routing.
- Are you on Premium or Enterprise, where capacity, schedules, and services live? Workload routing is native. If not, compare the upgrade with a custom build.
- Does routing need data outside the work item? Use a custom layer, or a rule that sets Team first.
- Do you need a threshold you set, and per-category accuracy? Confirm the native options first. The cited docs describe a default team when Rovo cannot confidently identify a team.
- Does the work fall under HIPAA? Workforce Management is not HIPAA compliant. Check Atlassian's compliance docs. A custom layer has to run in an environment you already treat as compliant. We are not stating our certification here.
FAQ
Is Rovo good enough for Jira ticket triage now?
For many Jira Service Management teams, yes. Workforce Management lets Rovo pick the team from each request, then assigns the least-loaded available, on-shift agent. Custom still makes sense when routing needs data from outside Jira, or when the work sits outside a service space.
Does Jira Service Management assign tickets by workload now?
Yes, if you use Workforce Management. An agent is eligible when they're on the owning team, scheduled to work and marked available. Work goes to the eligible agent with the lowest capacity utilisation. Capacity, schedules and services are Premium and Enterprise features. If nobody's eligible, the item stays unassigned.
What does Workforce Management routing cost?
Atlassian lists Workforce Management routing as a basic AI feature that uses Rovo credits: 10 credits per billable event. Credits are pooled across your organisation and depend on your plan. Extra usage is billed from December 3, 2026, and admins can set spending limits. Check your allowance in Atlassian Administration.
Should AI auto-assign tickets or only suggest?
Start with suggestions. In Workforce Management, turn auto-assign off and the Assignee field shows recommended agents, so a person still decides. Switch it on once the recommendations hold up for a week or two. Keep anything customer-facing, like replies and status changes, behind human approval.
Jira Automation rules or AI: which should handle SLA routing?
Use rules for anything deterministic: customer tier to priority, SLA timers, keyword to label. Use AI for the fuzzy part, like vague requests whose team isn't obvious. Automation can run alongside Workforce Management, for example to escalate items left unassigned. Rules plus AI is more reliable than AI alone.
How accurate is AI Jira routing?
It depends on how clear your team descriptions and historical tickets are. Measure it on your own data: route a labelled sample of past tickets, check results per category, and only automate the categories that hold up. Keep a person on the categories that don't.
When is a custom AI triage system worth it?
When native routing can't see what you need. Typical cases: a CRM or on-call tool, or work outside a JSM service space. Atlassian also says Workforce Management is not HIPAA compliant. Our pilots start at €5K, about two weeks, and you own the code.
Sources
- What is Workforce Management in Jira Service Management
- Set up Workforce Management
- Route work to the right agent
- Rovo usage limits
- Request Router (Service Triage Agent) docs, for request type, priority, and escalation suggestions
