The Triage & Dispatch Agent handles new tickets like a good service coordinator or dispatcher. It gathers the details your technicians need, classifies and prioritizes the ticket, and recommends or makes the right assignment.
This guide covers three things: the account setup that makes the agent's decisions accurate, the agent configuration itself, and how to run a low-risk pilot.
How the agent works
The agent listens for tickets in New status on the boards you choose. After the delay you set, it works through three phases.
Intake. The agent reads the ticket. If follow-up questions are turned on, it replies to the end user the way a technician would and asks for anything that's missing, such as when the problem started, what error appears, or how many people are affected. It then adds what it learned to the ticket and rewrites the subject into a clear title.
Triage. The agent sets the category and sub-category, calculates the priority, adds tags, and links relevant knowledge base articles.
Dispatch. The agent chooses the best technician or board based on skills, proficiency, past successful tickets, and current workload. In Suggest mode, it recommends the assignment, and a dispatcher makes the final call. In Auto mode, it assigns the ticket itself.
Each phase has its own confidence threshold. When the agent's confidence meets the threshold, it applies its changes. When it doesn't, it shows its recommendation without applying it, so a person stays in the loop on uncertain tickets.
Technicians can follow every step from the ticket page and take control at any time.
Before you start
The agent makes decisions based on the data already in your account, so the quality of that data determines the quality of its triage and dispatch.
1. Business hours and technician working hours
The agent uses your business hours and each technician's working hours to determine who's available and where tickets go outside working hours.
Set your business hours. Go to Control Center → Organization → Company Profile. In the Business hours card, click Edit, choose Weekdays, 24 x 7 Support, or Custom, set the times, and save. Check the Default time zone on the same page.
Check each technician's hours. Go to Control Center → Team & Access → Team members and open a team member. On the Overview tab, under Locale & region, confirm the Timezone and Working hours.
Tip: A technician with the wrong time zone or shift looks unavailable when they're working, or available when they're off. This is one of the most common causes of odd dispatch suggestions, so check it for every technician on your pilot board.
2. Boards
When deciding where a ticket belongs, the agent reads each board's name and description and looks at the skills of the technicians linked to it.
Go to Control Center → Service Desk → Boards and open each board you'll use.
Name and description. Give the board a clear name and a description that says what belongs there, for example: "Desktop and end-user support for Windows and macOS: Outlook, Teams, printers, password resets." A one-line description like "Helpdesk support tickets" gives the agent very little to work with.
Team members. Click Link team members and add the technicians who work this board.
3. Technician skills and proficiency
Dispatch matches each ticket to technicians whose skills fit it. Proficiency levels (Beginner, Intermediate, and Expert) let the agent tell a specialist from a generalist.
You can add skills in two ways.
Option A: role profiles (recommended). Go to Control Center → Team & Access → Roles & Permissions → Role profile. Your account already includes profiles such as Help Desk L1, Help Desk L2, Network Engineer, and M365 / Cloud Admin. Open a profile, adjust its skills and proficiency levels, and add members. Everyone in the profile inherits its skills, which keeps similar technicians consistent.
You can also assign a role profile from a team member's Overview tab, using the Role profile field.
Option B: skills for one technician. Open a team member, go to the Skills tab, and click + Skill. Choose skills by category and set a proficiency level for each one. Skills that come from a role profile are labeled Inherited.
Tip: Be realistic about proficiency. If everyone is marked Expert in everything, the agent loses the skill signal it needs to choose between technicians.
4. Service categories and severity
Service categories are how the agent sorts tickets. Each sub-category's Severity tells the agent how much attention that issue type deserves, which feeds into the priority it calculates.
Go to Control Center → Service Desk → Service categories and expand a category to see its sub-categories and their severity (Critical, High, Medium, or Low).
Your account starts with a standard set that you can edit. Add categories and sub-categories for the services you support, and set each severity.
Tip: Severity is not an SLA. It doesn't set response or resolution targets. It tells the agent which kinds of issues matter most to your business, so an ISP outage can rank above a single faulty network jack.
5. Workflows
If a workflow and the agent both change a new ticket at the same moment, they can overwrite each other. For example, a workflow might move every new ticket to a board while the agent routes it elsewhere.
Go to Control Center → AI & Automation → Workflows and review any workflow that runs when a new ticket is created. If it changes the board, assignee, priority, category, or status, do one of the following:
Add a time delay in the workflow using the Add time delay step so it runs after the agent finishes.
Increase the agent's delay so the workflow finishes first.
Turn the workflow off for your pilot board if the agent now does that job.
Configure the agent
Click AI in the left sidebar. Click Activate your first agent.
Setup has five steps, and your progress is saved as a draft as you go. The agent doesn't start working until you click Activate agent on the last step.
Step 1: Overview
Agent name. The name end users see when the agent replies, so choose one you're happy for customers to see.
Short description. An internal note describing what this agent does.
Channels. Choose where the agent picks up tickets: Email, IT-Connect, API & MCP tickets, or Support desk portal. Select the channels your pilot tickets will arrive through.
Boards to monitor. Select board(s) the agent listens to. Start with a single board in pilot tests.
Agent delay. How long the agent waits after a ticket is created before it starts intake. Pilot setting: 30 seconds.
The delay gives the user a moment to add an attachment or a quick follow-up, and gives other automation time to finish first.
Step 2: Intake
Auto-format summary. Rewrites vague ticket subjects into clear titles.
Link duplicate tickets. Links a new ticket to an existing ticket that looks like the same issue.
Ask follow-up questions. Set the Follow-up question limit (3 to 5), choose a Tone & personality (Friendly or Formal), and decide whether to allow attachments/screenshots in replies.
SLA-aware fast-track. When a ticket is close to breaching its SLA, the agent skips follow-up questions and goes straight to triage and dispatch.
Intake confidence threshold.
Pilot setting: 70%. Leave low-confidence behavior on Suggest only: show recommendations, don't apply. The other option, Skip, leaves the ticket for a dispatcher without a recommendation.
How follow-up questions work in each channel. The agent checks whether a ticket came in by email or through IT-Connect and adjusts its questions to suit the channel.
Email. The agent sends one email that groups a few related questions together, so the user can answer everything in a single reply. If the reply still doesn't give enough information, it sends one more follow-up email at most. An email ticket gets no more than two follow-up emails, even if the follow-up question limit is set to 5. This keeps users from receiving a separate email for every question.
IT-Connect. The agent asks its questions conversationally, one after another in real time, so the user can answer each one straight away.
Step 3: Triage
Auto-categorize (always on). Predicts the category and sub-category.
Link KB articles. Attaches relevant knowledge base articles.
Add tags. Adds searchable keyword tags.
Define priority. Calculates priority (P1 to P4) using DeskDay's priority scoring, which takes your service category severity into account.
Triage confidence threshold. Pilot setting: 70%, with low-confidence behavior on Suggest only.
Step 4: Dispatch
Dispatch mode. Pilot setting: Suggest. The agent recommends a technician, and a dispatcher makes the final assignment. Move to Auto once you trust its recommendations.
After-hours routing. Optional. Choose a board for tickets that arrive after hours, on weekends, or on holidays.
Confidence threshold for dispatch. Pilot setting: 70%, with low-confidence behavior on Suggest only.
When you switch to Auto, more settings appear:
Dispatch to. Route tickets to a Team member or to a Board.
Team members in the dispatch pool. Who the agent can assign to.
Base workload on. How the agent measures a technician's workload, and when it treats them as busy. Recommended: Active tickets (tickets the technician is working on right now), with a Busy at number that fits your team. Other options include Assigned tickets (all open tickets assigned to the technician), Critical & high priority tickets, and Committed hours (logged time plus calendar bookings).
Tip: "Busy" isn't a hard stop. The agent prefers technicians with more room, but it can still choose a busy technician when they're clearly the best fit for the ticket.
Step 5: Guardrails and exclusions
Excluded customers and excluded users. The agent won't act on tickets from these customers or users; for example, a VIP account you want to keep fully manual during the pilot.
Instructions (optional). Plain-language rules for the agent, for example: "Do not process requests involving password resets, access permissions, or sensitive customer data."
When you're ready, click Activate agent.
Run your pilot
Start with one board. Keep the pilot to a single board so you can review every decision the agent makes.
Start in Suggest mode with 70% thresholds.
Send a realistic mix of test tickets. Include well-described tickets, vague one-liners ("VPN issue"), tickets with misleading subjects ("Quick question" for what is really an outage), and at least one genuine outage. Use a test customer or contact if you can.
Give each ticket a couple of minutes. The agent waits for the delay and then works through each phase. Category and priority can update a minute or two after the ticket first appears, so a ticket that briefly shows Low priority may not be finished yet.
Review each decision. On the ticket page, check the AI triage badge at the top, the AI triage note in the conversation, and the Agents tab in the Helena panel. The Agents tab shows each phase, its confidence, and the reasoning behind the dispatch recommendation.
Take control when you need to. A technician can take over a ticket at any time. The agent keeps what it has already applied and skips its remaining steps.
Track how the agent is doing
Open the agent from AI in the left sidebar to see the recent agent activities. Select a ticket to see the details of each action the agent performed and the reason behind each decision.
When to expand
Once your team agrees with most of the agent's recommendations and the override rate stays low across a full week of normal ticket volume, expand one step at a time. Add a second board, or switch the dispatch mode to Auto, but not both at once, so you can see what changed.
Recommended pilot settings
Setting | Where | Pilot value |
Boards to monitor | Step 1: Overview | One board |
Agent delay | Step 1: Overview | 30 seconds |
Ask follow-up questions | Step 2: Intake | On |
Intake confidence threshold | Step 2: Intake | 70% |
Define priority, Link KB articles, add tags | Step 3: Triage | On |
Triage confidence threshold | Step 3: Triage | 70% |
Dispatch mode | Step 4: Dispatch | Suggest |
Dispatch confidence threshold | Step 4: Dispatch | 70% |
Low confidence behavior (all phases) | Steps 2 to 4 | Suggest only |
Base workload on (when you move to Auto) | Step 4: Dispatch | Active tickets |
Troubleshooting
The agent didn't pick up a ticket. Check that the ticket is in New status, on a monitored board, and came in through a selected channel, and that its customer or user isn't excluded. Allow time for the agent delay.
A ticket still shows Low priority right after it's created. Wait a couple of minutes. Priority is set during triage, after intake finishes.
Dispatch suggestions look wrong. Check the technicians' time zones and working hours, the board's description and linked members, and the technicians' skills and proficiency.
Ticket fields change back or flip between values. A workflow is probably acting on the ticket at the same time as the agent. Add a delay to the workflow or to the agent.