Blog ·
Where to start with helpdesk automation: sort by margin, not volume
Most MSPs automate the wrong tickets first. The high-volume categories aren't the same as the high-margin ones. Here's the sort to run before you build anything.
Most MSPs automate the wrong tickets first.
If you're planning your first AI L1 deployment (or just deciding which categories to script first), the obvious move is to start with the high-volume work. Password resets, license adds, anything that hits the queue 50 times a week. Easy to identify, easy to test.
It's the wrong sort.
What you should sort by
The thing eating your margin isn't the ticket count. It's the total tech-minutes per category. And the two numbers don't track the way you'd think.
Real-shape example from an MSP I worked at:
- Password resets: 2 min average × 200 tickets/mo = 400 tech-minutes/mo
- M365 user lifecycle (onboarding/offboarding): 45 min average × 30 tickets/mo = 1,350 tech-minutes/mo
The M365 category is 3.4× the margin opportunity of password resets, even though it's a sixth of the volume. Automating password resets first feels productive. It moves a small number.
How to run the sort
Pull 90–180 days of resolved tickets from your PSA. For each ticket, you want two numbers: category (or issue type) and total time logged on the ticket.
Group by category. Two sorts:
| Sort by ticket count (the wrong one) | Sort by total tech-minutes (the right one) |
|---|---|
| Password resets | M365 user lifecycle |
| License adds/removes | Long-tail break/fix |
| DL membership changes | Backup/restore tickets |
| New device setup | Vendor coordination |
| Office365 sync issues | Email migration cleanup |
Your actual top-5-by-margin list will look different. That's the point. Run the sort against your data.
What this changes about the build order
The top of the by-margin list usually has one of two shapes:
A small number of expensive multi-step tickets. M365 onboarding/offboarding fits here. Each ticket is 45 minutes of clicking between Syncro, M365 admin center, AD, distribution lists, license management, and an HRIS. High variance, lots of judgment, but the workflow is consistent enough to encode. This is what an AI L1 tech is genuinely good at.
Long-tail break/fix with high time variance. These are harder to automate because the work is genuinely novel each time. AI helps with the drafting and lookup, but the human stays in the loop on resolution. Don't try to fully automate this; do build agent-assisted tools for it.
The high-volume low-margin work (password resets, simple license adds) goes on the build list, but later. They're easy wins for momentum, not margin.
The two-week test
Before you commit to a build plan, do this:
- Export 90 days of tickets, group by category, sum the tech-time.
- Show the top-10-by-margin list to your senior tech.
- Ask: "Which of these categories has a workflow stable enough to write down?"
Anything they can describe in a one-page runbook is automation-eligible. Anything they can't is judgment work: agent-assisted at best, not automated.
That two-week diligence is worth more than three months of building the wrong thing.
What category at your shop has more total tech-time than you'd have guessed?