Three buckets, not one
The mistake in most help desk automation projects is treating every ticket the same. A sensible split looks like this.
| Bucket | Examples | What automation does |
|---|---|---|
| Safe to automate | A stopped service, a jammed print queue, a full temp folder, a stale DNS cache, a pending-reboot alert that is no longer true | Runs the fix, checks it, writes the note |
| Prepare, then a person decides | Restarting a server with users on it, installing Windows updates, changing a driver, anything on a domain controller | Investigates, proposes the exact step, waits |
| Leave to people | Failing hardware, a vendor escalation, a policy or budget decision, deleting an account | Gathers evidence, then hands over |
A good automation tells you which bucket a ticket landed in, and why.
Finding out which tickets are worth automating
Most MSPs cannot answer “what do we fix over and over?” from memory. Automation Opportunities reads your closed Autotask tickets for a period you choose, groups them by underlying problem rather than by identical title, and ranks them by how often they occur and how automatable the evidence suggests they are.
It drafts an Autopilot from the real resolutions your technicians wrote, flags the assumptions it made for you to check, and saves it switched off. Nothing runs until someone reads it and turns it on.
Where it sits next to your PSA and RMM
- PSA: Autotask. Tickets in, internal notes back. Other PSAs are not supported today and this page will not claim them.
- RMM: ConnectWise RMM and Automate, used to see your devices and push the agent to them. It does not replace them.
- Microsoft 365: user, group and licence tasks through Microsoft Graph, after a one-time admin consent per client.
To be plain about the boundary: this is not an RMM. There is no remote desktop session, no third-party software patching and no backup management. It does one job - investigating and fixing endpoint problems - and sits beside the tools that do the others.
Measuring whether it helped
Counts of tickets “deflected” are easy to inflate, so the platform keeps the awkward ones visible: how many sessions needed a technician to steer, how many waited on an approval, how many ended as “insufficient evidence”, and how many came back after being marked resolved. Use the ROI calculator with your own numbers, then check them against a pilot.
Frequently asked questions
What is help desk automation for MSPs?
It is using software to handle part of the support queue: classifying tickets, running diagnostics and applying known fixes. Good implementations separate the tickets that are safe to automate from those that should only be prepared for a person.
Which tickets should an MSP not automate?
Anything involving failing hardware, vendor escalations, policy or budget decisions, and account deletion. For those the useful automation is gathering evidence and handing over, not acting.
Does help desk automation replace my PSA or RMM?
No. It sits beside them. Tickets come from the PSA and notes go back to it, and an RMM is used to find devices and push the agent, while the remote investigation and fixing is the part it adds.
How do I know which recurring tickets to automate first?
Look at your closed tickets grouped by underlying problem, not by title, and rank them by how often they recur and how repeatable the fix was. A tool that does this for you should show its evidence and leave the result switched off until a person reviews it.