If you've been heads-down running your service desk, you may have missed some significant updates we've shipped to ServiceAI over the last seven weeks. From a major overhaul of the Orion Assistant to tighter Triage controls and meaningful quality-of-life improvements across the platform, here's a look at what's changed and why it matters for your team.
Orion Assistant Just Got a Lot More Powerful
This is the biggest update in this batch, and it changes how you interact with ServiceAI in a fundamental way.
Ask Orion is now context-aware from your prompt, not just your screen. Previously, Orion answered questions based on the page you were on, which meant asking about tickets while viewing a company page didn't work well. That limitation is gone. Orion now uses tools to determine what data is relevant based on what you're actually asking, pulling from your own ServiceAI data as well as the web for current general knowledge.
In practice, this means you can now ask questions like "Which problem has occurred most frequently in the last week?" and Orion will look through your ticket data and answer — not guess. Orion has also moved to the left sidebar for easier access from anywhere in the platform, and it's running on a more powerful underlying model.
Orion Custom Slash Commands are now available. You can define reusable prompts with parameters and invoke them with a /command in any Orion conversation. Start a conversation with / to see available commands, and build your own under Settings > Commands. If your team asks Orion the same types of questions repeatedly, this turns those into one-keystroke workflows.
Triage Is Getting More Precise
Triage is one of the highest-leverage features in ServiceAI, and we're continuing to tighten the controls around it.
You can now exclude tickets from triage by source. If you're generating tickets through your Unified Client Portal (UCP) or another specific source that doesn't need AI triage, you can configure those sources to be skipped entirely. This keeps your triage processing focused where it actually adds value and avoids consuming slots on tickets that were already correctly categorized at submission.
Triage Professional now includes 200 tickets per month. Professional plan users get 200 monthly tickets included. If your volume exceeds that, contact your account manager for options.
Spam detection now runs at sync ingestion. Previously, spam detection happened after tickets were already in the queue. Now, AI-powered spam detection flags tickets during ingestion — before they hit your triage pipeline. Flagged tickets get an internal note posted to your PSA and are excluded from further processing automatically. (And if something gets incorrectly flagged, you can undelete spam tickets without the spam note being re-posted.)
Triage field labels now match your PSA's terminology. If your PSA calls it a "Board" instead of a "Queue," that's now what you'll see in ServiceAI. Less translation work, less confusion for your team.
A new "View in Help Desk" button on the Triage page lets you jump directly to the ticket in your PSA from triage results. One click instead of tab-switching and searching.
Triage Diagnostics are now available. A new diagnostics blade lets you verify that PSA webhook registrations are correctly configured — so if triage isn't behaving as expected, you have a real troubleshooting starting point instead of guesswork.
Data Flexibility You've Been Asking For
Excel export is now available across ServiceAI. Any table in the platform — Tickets, Agents, Companies, Articles, Triage, Users — can be exported to Excel using the new download icon in the grid toolbar. Filtered data exports with it, so you're getting exactly the slice you need for reporting, QBRs, or sharing with stakeholders who aren't in the platform.
You can now set a personal default time range. Rather than resetting your time filter every time you open a view, you can set your preferred default and have it applied automatically. This is one of those small things that makes a real daily difference when you're in the platform constantly.
Tickets, agents, and companies can now be manually deleted or restored from within ServiceAI without requiring a full sync, giving administrators more direct control without waiting on a background process to catch up.
Security and Access Controls
Administrators can now enforce SSO at the tenant level. Password-based login can be disabled entirely, requiring all users to authenticate via SSO or email link. For MSPs with strict security posture requirements, this closes a meaningful gap. Configure it under Settings > Access.
Email notifications can now be disabled by role. Administrators can turn off email notifications for Editor and Viewer-level users, reducing inbox noise for team members who don't need alert emails.
Integration Reliability
Across PSA integrations, we've addressed a number of issues that were causing real operational friction:
- Halo sync user-to-company mapping and source URLs are corrected
- Hudu articles now associate properly with companies
- ITGlue export no longer stops permanently on transient API failures — it retries
- Kaseya BMS agent names now display correctly in the Agents grid
- Syncro comment retrieval has been updated ahead of a March API deprecation
- ConnectWise triage assignment, callback pagination, and article sync issues are resolved
- Autotask triage assignment rules no longer leave tickets unassigned due to missing resource role
- IT Glue article import no longer skips articles when the associated company lacks a PSA sync
The Bigger Picture
Looking across these updates, there's a clear pattern: we're building ServiceAI to handle more of the operational complexity your service desk deals with every day. Orion is no longer just a sidebar assistant — it's becoming a true data interface for your service operations. Triage is getting smarter and more controllable. And we're reducing friction in the places where small annoyances compound over time.
We ship on a regular cadence because the MSPs using ServiceAI are running real businesses, and standing still isn't an option. Keep an eye on the changelog — there's more coming.
CloudRadial
