Skip to main content

​
Setting up your Triage Agent (user guide)

What your Triage Agent does
Your Triage Agent handles the first touch on every incoming thread. It greets the requester, asks the right questions, gathers the context your techs usually chase down, and either resolves common issues or hands off a ticket that is ready to work. This setup tunes it to your real work and lets you watch it in action before it ever reaches a client, so you can turn it on with confidence. What good looks like: every thread reaches your techs already triaged, with the details they need and less back-and-forth. Your team spends its time resolving, not gathering.
Before you start
  • Your PSA should already be connected and syncing.
  • Select a board you know well. Setup learns from the real threads on it.
  • Nothing is locked in. You can change everything after setup.
The setup steps
Admin → Get Started → Set up Triage Agent
1

Name and personalize your agent

Give the agent a name and select an avatar. This is what your clients see in chat, and a familiar name and face make it feel like part of your team. This step shows whatever name and avatar you already have in Messenger settings, so you can keep or change them here. Changes made during setup overwrite your Messenger settings on publish, and you can update them any time later in Messenger settings.
2

Select a board to learn from

Select a board with strong, representative threads — ideally one where the Triage Agent would clearly thrive. We read the recent real requests on that board, identify the three most common types, and use them to show the agent in action. The better your example board, the more useful the rest of setup. The agent learns from your actual tickets, not generic scripts.
3

Review and coach the conversations

For each of the three topics, the agent runs a full simulated back-and-forth, just as it would with a real requester. Approve what looks good, or give feedback in plain language (for example, “ask for the asset tag before troubleshooting”). The agent adjusts on the spot, and your feedback becomes a Custom Rule you can edit now or later. This is where the agent learns your standards and starts to sound like your best technician.Tell it how to behaveUse this to shape how your agent sounds and behaves, not just what it does. Think tone, greetings, and conversational habits. For example: “Always open by introducing itself by name (eg. Mr Agent) and welcoming the client,” or “Use the contact’s first name in the first message only, not in every reply,” or “Keep replies short and friendly with max 1 emoji, and avoid jargon unless the client uses it first.” Describe the behavior you want and when it applies, and we will save it as a rule you can review and edit in the next step.
4

Review and refine

Before you give any client access, the Review step is where you make the agent yours. You have just watched it work in the Test step, so now you tune the details and confirm it behaves the way your team would.Default rules. Three core behaviors, already on and ready to adjust:
  • Collect specific information when identifying the issue. Select what the agent always tries to pin down before handing off: scope and identifiers (who or what is affected), symptoms or goal, impact and urgency, access and authentication, environment, and recency or change. Select at least two; the more you select, the more complete the ticket your techs receive.
  • Keep responses a certain length. Short asks only the essentials, Medium probes a bit deeper into root cause, Long gathers as much relevant detail as it can. Match this to how much back-and-forth your clients will tolerate.
  • Use a specific tone. Friendly and conversational, Professional and formal, or Neutral and concise, so the agent sounds like your brand.
Custom rules. This is where the coaching you gave during the Test step lives, saved as editable rules. Open any rule to fine-tune it, add new ones, or set whether a rule applies to all clients or only some. Nothing is locked in; everything here can change now or any time after onboarding.Keep testing. The live test panel stays on the right the whole time. After any change, send it a sample issue and watch how it responds. Spend a few minutes here getting comfortable. The more you test and adjust now, the more confident you will be when real client threads start flowing.When the agent is asking the right questions, in the right tone, at the right length, you are ready to give it access to your first clients.
5

Set client access

Select which clients, sources, and boards the Triage Agent works on once it is live. It only acts on threads that match, so you decide exactly where it shows up.
Recommended: start with your internal teams first. Seeing it prove itself on your own people is the lowest-risk way to build confidence before you expand to client-facing threads. You can widen access any time.
Refine who the agent engages. On top of clients, sources, and boards, you can narrow access at the contact level and silence noisy senders:
  • Excluded contacts — pick individual contacts the agent should never engage with. Contact-level exclusions take precedence over company-level access, so an excluded contact stays blocked even if their company is allowed.
  • Contact types — include or exclude requests based on a contact’s type (ConnectWise contact types).
  • Request blocking — stop the agent from replying to out-of-office auto-replies, alerting-tool noise, and no-reply mailers by sender domain, email address, or subject keyword. See Request Blocking and common use cases.
6

Turn on your agent

On the final step, a confirmation popup includes a Triage Agent toggle.
Recommended: leave this toggle on. With it on, the agent starts triaging the threads you just scoped, which is the fastest way to see the value in your own queue. If you leave it off, your setup is saved but the agent stays idle until you turn it on later.
After you publish
  • The agent starts triaging new threads within the access you set, so your team feels the difference on the next ticket.
  • The Custom Rules created during setup appear in your Custom Rules list, where you can refine, add, or remove them as you learn what works.
  • Name and avatar live in Messenger settings; client access lives in Triage Agent settings. Both editable any time.

​
Fine-tune behavior and escalation

The guided setup already sets your agent’s tone, response length, and the information it gathers (see Review and refine above). A few more behaviors live in Agent Settings (Admin → Magic Agents → Triage Agent → Edit Agent Settings), where you shape how the agent troubleshoots, structures its messages, and hands off to a human.

​
Troubleshooting

Choose how the agent handles troubleshooting before it escalates:
  • Ask if they want to troubleshoot or escalate — the agent offers the contact the choice.
  • Suggest a few troubleshooting steps — the agent offers simple, guided steps when appropriate.
  • Do not troubleshoot, only collect information — the agent skips troubleshooting and focuses on gathering details for escalation.

​
Communication style

Choose how the agent structures its questions:
  • Ask each question one by one — the agent asks individually, adjusting to each response.
  • Ask all questions in one message — the agent sends every question at once so the contact can answer in a single reply.

​
Escalation words

Escalation words tell the agent when a message needs to be prioritized or handed to a human technician. When any of these words or phrases is detected, the agent escalates the conversation automatically. Click + New word to add one. Common examples:
  • “talk to technician”
  • “it’s urgent”
  • “need immediate help”
  • “this isn’t working”
  • “can someone call me”
  • “critical issue”
  • “system down”

​
Escalation message

The escalation message is what the contact sees when a thread is handed to a technician. The agent sends this text verbatim — the exact message you enter. Leave it blank and the agent uses the system default (“Ok, I will find the next available team member to assist you. Hang tight—we’ll be with you shortly!”).

​
Custom thread closed message

Shown to the contact when the agent automatically closes a thread. Unlike the escalation message, this field is prompt-based: the agent treats your text as a guiding instruction and writes its own closing message in that spirit (for example, “Respond that the request was resolved successfully.”).
Leave this field blank and the agent uses the system default message when it closes a thread automatically.

​
Custom rules, intents, and settings

The guided setup gets your Triage Agent live. These dedicated guides cover everything else you can configure:

Triage Agent Custom Rules

Target structured rules by client or contact type, test them in the emulator, and publish safely. The coaching you gave during setup lives here as editable rules.

Build your Service Catalog

Define the intents your Triage Agent recognizes and acts on for repeatable requests.

Create your first intents

Add Thread’s recommended intent templates and build custom intents for your own use cases.

Triage Agent Settings

Response timeouts, the timeout message, post-deactivation Magic AI actions, and the status mapping required for resolutions.
FAQ
The agent builds your sample conversations from real, recent requests on that board. A board with clear, representative threads gives you the most useful setup.
No. The agent only acts within the clients, sources, and boards you set in step 4.
It lets you see the agent prove itself with low risk before client-facing threads.
Your setup is saved, but the agent will not process threads until you turn it on.
Yes, in Messenger settings → Design in Admin. What you see during setup comes from there, and changing it during setup overwrites the Messenger values on publish.
They are simulations built from real past requests on the board you choose. Nothing is sent to clients during setup.

Related topics

Configuring the Triage Agent for OOO and Other Noisy RequestsGetting Started with Triage AgentListening Mode for Triage Agents