Blog
Field notes from the L1 queue.
Operator writing on AI L1 techs, MSP automation, Syncro scripting, and the field-tested practice of building help-desk infrastructure that doesn't break.
-
An honest postmortem of an AI feature with no war stories
Most AI postmortems are war stories. This one isn't. Notes on a small Spanish-translation tech that ran for months without drama, and the design decisions that kept it that way.
Read post → -
Least privilege for non-IT staff doing M365 admin
If you've given HR Global Admin to do password resets, you've made a security decision that looked like a convenience one. Here's what scoped delegation actually requires.
Read post → -
Six questions to ask an AI helpdesk vendor that they'll struggle to answer
Vendor pitches in this category blur together. Here are six questions that surface the actual difference between a polished demo and a system you can operate for two years.
Read post → -
Why dispatchers beat monolithic chatbots for L1 work
One big chatbot with access to everything is the design pattern that doesn't survive contact with real tickets. Here's why the dispatcher pattern wins, and what it costs.
Read post → -
What "multi-tenant by design" actually requires
Adding a tenant_id column isn't multi-tenancy. Here's what real multi-tenant infrastructure for MSPs actually requires, and the one-question test that tells you whether a tool has it.
Read post → -
Asset Scripts vs Worksheet Scripts: when to use which
Two Syncro tools with similar names and different purposes. Picking wrong wastes a week. A practical decision rule for when each one is the right tool.
Read post → -
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.
Read post → -
Why we call them "AI techs," not "AI employees"
The naming gap between "AI Employee" and "AI Tech" isn't cosmetic. It sets expectations for what the system can do, and how to think about it when it's wrong.
Read post → -
How AI helpdesk projects actually fail
The first 60 days of an AI helpdesk project are vibes. The next 60 are the real test. Three failure patterns from inside the queue, and how to design around them.
Read post → -
What "Syncro-native" actually means
The bolted-on vs built-in distinction matters more than vendors admit. Here's what "Syncro-native" should mean, and the one question that tells you whether a tool actually is.
Read post → -
Prismatic's debut AI L1 demo: automated password reset that runs inside Syncro
How a multi-agent loop handles password reset tickets end-to-end inside Syncro: the architecture, the verified user flow, and what the demo assumes.
Read post → -
A three-bucket frame for deciding what to automate, delegate, or hand to an AI agent
After 18 years in IT, the most useful frame I've found for triaging repetitive helpdesk work: what to automate, what to delegate, and what to hand to an AI agent.
Read post →