If you're implementing ServiceAI for your MSP, one of the most critical - yet often overlooked - setup steps is configuring exclusions. While it might seem like a minor administrative task, proper exclusion management can mean the difference between actionable insights and polluted analytics that lead you astray.
- The Problem: Not All Tickets Are Created Equal
- Why Automated Alerts Poison Your AI
- How Different Exclusion Types Work
- Strategic Implementation: Timing Matters
- What Happens After Exclusion
- Common Exclusion Patterns for MSPs
The Problem: Not All Tickets Are Created Equal
Modern MSPs generate thousands of tickets monthly, but not all of them represent genuine client-technician interactions. Your PSA is likely flooded with:
- RMM alerts (disk space warnings, CPU spikes, patch failures)
- Backup notifications (success confirmations, failure alerts)
- Security alerts (antivirus detections, Huntress notifications, EDR warnings)
- Automated monitoring tickets (server down alerts, network disruptions)
These automated alerts serve important operational purposes, but they're fundamentally different from tickets where a client reaches out with a problem and a technician provides a solution through professional communication.
In short, they will be scored poorly by ServiceAI and will skew your service desk scores.
Why Automated Alerts Poison Your AI
When ServiceAI analyzes your tickets, it's looking for patterns in human communication to:
- Train itself on how your team solves problems
- Score technician performance (empathy, responsiveness, de-escalation, and more)
- Measure client satisfaction (user RPS scores)
- Identify knowledge gaps and opportunities for documentation
Automated alerts break this model in several ways:
They Tank Your RPS Scores
Remember that ServiceAI uses three core metrics:
- Ticket RPS - Quality and clarity of ticket notes
- Agent RPS - Technician performance and professionalism
- User RPS - Client behavior and satisfaction
An automated alert typically has:
- A system-generated description (no real "user" communication)
- Technician responses like "closing" or "resolved" with no context
- No conversational exchange showing problem-solving
The AI interprets this as poor service: a technician who didn't respond professionally to a client. Your scores plummet, not because your team performed poorly, but because the AI is analyzing the wrong type of data.
They Obscure Real Performance Issues
When your dashboard shows 40% of tickets have low agent RPS scores, you need to know if that's a training issue or just RMM alerts cluttering your data. Exclusions ensure you're only seeing actionable insights about actual client interactions.
They Waste AI Training Resources
ServiceAI learns from your historical tickets to provide better recommendations. Feeding it thousands of "Backup completed successfully" tickets doesn't teach it anything useful about troubleshooting client problems. Clean data means smarter AI assistance. Garbage in, garbage out.
How Different Exclusion Types Work
ServiceAI's exclusion system operates on three levels:
Subject Line Exclusions (Most Common)
This is your primary tool for filtering automated alerts. The system uses "contains" logic, meaning you only need to identify the repeating keyword in your automated tickets.
Examples:
-
Huntress alert- catches "Huntress alert: Malware detected on CLIENT-PC" -
RMM Alert- catches all variations with that phrase -
Backup Failed- filters backup notifications -
[RESOLVED]- excludes auto-closed tickets
You don't need the full subject line - just the consistent identifier that appears across all similar tickets.
Technician Exclusions
Some MSPs have dedicated automation accounts or service accounts that handle automated workflows. Excluding these "technicians" prevents their non-human interactions from affecting team performance metrics.
Tip: You can find the Agent ID by clicking on Agents on the left-hand side and viewing their ID from there. The Agent ID field validates numeric input — typing an agent name now shows a validation error on save instead of silently accepting it, so be sure to enter the numeric ID rather than the agent's display name.
Company Exclusions
If you have internal infrastructure or test clients that generate tickets you don't want in your analytics, you can exclude entire companies from processing.
Tip: Like Agent ID, you can select Companies on the left-hand menu and find the ID from there. The Company ID field validates numeric input — typing a company name now shows a validation error on save instead of silently accepting it, so be sure to enter the numeric ID rather than the company's display name.
Source Exclusions (Triage-Specific)
In addition to the analytics exclusions described above, ServiceAI's Triage feature offers Excluded Sources — a separate filtering mechanism that prevents Triage from processing tickets based on their PSA source field.
While Subject Line, Technician, and Company exclusions affect analytics and AI scoring across all of ServiceAI, Source Exclusions apply only to the Triage pipeline. Tickets from excluded sources are skipped before any AI analysis occurs, preserving your monthly triage credits.
To configure Source Exclusions, navigate to Settings > Triage and use the Excluded Sources multi-select. See the Triage documentation for full details.
Channel/Source Exclusions
Channel Exclusions let you exclude tickets by their originating channel (CloudRadial terminology) or source (Autotask terminology) directly from the general exclusions page. This is distinct from the Triage-only Excluded Sources feature described above — Channel Exclusions apply broadly to analytics, AI scoring, and training data across all of ServiceAI, not just the Triage pipeline.
This is especially useful for MSPs receiving large volumes of RMM or monitoring alert tickets (Huntress, Datto RMM, SentinelOne, etc.) that all flow in through a single channel. Instead of maintaining dozens of subject-line exclusions to catch every variation of an automated alert, you can exclude the entire channel in a single rule.
To configure a Channel Exclusion, navigate to Settings > Exclusions > Channel Exclusions and select the channel or source you want to exclude. The list is populated from the channels and sources discovered in your connected PSA, so the available options match what your PSA actually delivers.
Channel Exclusions apply retroactively after the next cache update — existing tickets from the excluded channel are reclassified the same way they would be if you had added a Subject Line or Company exclusion. Excluded tickets remain visible in your PSA and ServiceAI but are shown with the red strikethrough and removed from RPS calculations and AI training.
Strategic Implementation: Timing Matters
The Early Advantage
Configure exclusions before or immediately after your initial ticket sync. Here's why:
Small dataset benefits:
- Exclusions can process in minutes, not hours
- You can iterate quickly to test different keywords
- Scores stabilize faster with clean data from day one
Large dataset challenges:
- Thousands of tickets take substantial time to reprocess
- You're waiting potential days to see accurate analytics
- Initial reports are misleading during the processing period
The Identification Process
Step 1: Add the AI Ticket RPS Column
In your ServiceAI ticket view, add (or move up) the AI Ticket RPS column. This immediately highlights outliers - tickets with unusually low scores that often indicate automated alerts rather than genuine performance issues.
Step 2: Look for Patterns
Sort by low RPS scores and look for:
- Repetitive subject lines
- System-generated usernames
- Lack of conversational notes
- Instant close times
Step 3: Create Targeted Exclusions
Start broad with the most common patterns, then refine:
- Begin with your most prolific automated systems
- Add 5-10 exclusions initially
- Monitor results over 24-48 hours
- Add additional exclusions as patterns emerge
What Happens After Exclusion
Once you add an exclusion:
Immediate effect: Matching tickets show a red strikethrough in the interface. Excluded agents now also appear with a red strikethrough in the Agents list — previously only tickets reflected the exclusion visually, but agent rows now provide the same at-a-glance confirmation that the exclusion is active.
Within 24 hours: Ticket RPS values are marked as "not processed" when you click on them.
- They remain visible in your PSA and ServiceAI - but struck through in red.
- They're excluded from RPS calculations
- They don't feed AI training data
- They don't appear in performance analytics
Visual confirmation: Excluded tickets and excluded agents display differently, so you can verify your rules are working as intended
Common Exclusion Patterns for MSPs
Based on typical MSP workflows, consider these starting exclusions for your subjects:
RMM Alert
[AUTOMATED]
Backup Failed
Backup Successful
Disk Space Warning
Huntress
SentinelOne
Windows Update
Patch Failed
[AUTO-CLOSE]
[RESOLVED AUTO]
ConnectWise Automate
Datto Alert
Veeam NotificationCustomize these based on your specific tools and naming conventions.
Comments
0 comments
Article is closed for comments.