A One-Person MSP Tests AI on the Proactive Work He Never Gets To

Jay runs a one-person MSP with clients all over Australia. By his own count that's about 1,290 endpoints, one RMM, one simple ticketing tool and fewer than 20 tickets a day. The reactive work gets done. The proactive work doesn't: "If I'm looking after 1200 endpoints by myself, I'm not proactive."

He signed up for the OpenFrame beta after an ad came up on Instagram. His current tools have no AI built in, so he wanted to see where AI is at. At the interview in April 2026 he'd had about a week of hands-on time. In that week Mingo, OpenFrame's AI agent for technicians, cleaned up disks across a group of his clients, told him why a slow machine was slow, and set up a printer. Slowly, but it got there.

Jay's ticket queue is light for a fleet his size. What he never gets to is the maintenance that keeps machines healthy.

Disk space is the big one. A lot of his clients' computers were running out of space. The options were a road trip around Australia upgrading SSDs, or swapping machines out, which plenty of clients would rather do. Neither is quick when you're the only technician.

Slow computers are the other constant. People run three screens off a three- or four-year-old laptop with "200 Chrome tabs open", then wonder why it crawls. Today that means remoting in and showing them what's going on. His RMM has the data, but, as he puts it, "you've got to go looking for it."

None of his current platforms have AI built in. He wanted to know what AI could take off his plate.

He started small. He put a few clients on OpenFrame and set Mingo a challenge: this is how much space my RMM's scripts can clean up, show me what you can do. Mingo wrote and ran its own cleanup scripts, and it cleaned up "so much more than what I thought was possible."

That was enough to go bigger. He had a group of clients that all needed cleanups, so he asked Mingo to clean every computer except the servers. It worked through them and asked him a lot of questions about scripts along the way. "It would have taken all day," he said, "but it did get the job done." His separate security monitoring tool was watching the whole time as a safety net.

For slow machines he just asked Mingo what was going on. It came back with the processes using the most system resources. No digging through menus.

He also played the end user. With a portable printer next to him and a laptop that had never seen it, he asked Mingo to set it up. It could, "but it took a long time."

At the interview in April 2026 Jay had used OpenFrame for about a week before two back-to-back projects took him away.

On disk space, Mingo's own scripts freed more than his RMM's scripts on his first test clients. One request then cleaned every non-server computer across a group of clients. It was slow and needed a lot of approvals, but it worked. His goal was to buy a bit of time before a road trip of SSD swaps, and "it did a really good job."

On slow computers, one question to Mingo listed what was eating the machine's resources. The printer install worked, slowly.

Where it fits him: "Where this tool would be handy for me is just doing proactive maintenance on computers, cleaning them up."

What he still wants: remote control that takes keyboard and mouse input after Ctrl+Alt+Del, steadier approve and reject buttons when several Mingo actions queue up, tags to mark servers, an easy view of all devices for one client, labels for Mingo chats, an FAQ, clearer onboarding, and a PSA his clients can reach from a tray icon. Next on his list for Mingo: installing software on request and helping set up new devices, "just to take a small bit of workload off."