What we learned building AI agents for MSP operations
Six months into building real autonomous tooling on top of our platform. What worked, what didn’t, what comes next.
Lacy MooreJune 3, 20261 min read
We started with a narrow target: a ticket remediation agent that could resolve common issues end-to-end without a human. Six months later we have something working in production on a curated set of ticket types, and a clear sense of what the next year looks like.
The big lesson is unglamorous: tools and data shape outcomes more than model choice. A capable model with the wrong tools and stale documentation produces confidently wrong actions. A modest model with good tools, fresh data, and tight verification produces useful work.
We are not bullish on agentic MSP operations because it is trendy. We are bullish because it is the mechanism by which we can keep raising the bar on service quality without burning out the team.
Related service