Most of the public conversation around Cursor focuses on two things: shipping fast with “vibe coding,” and tightening the loop on everyday edits. Both are real. Vibe coding is oddly fun—almost addicting—and small, surgical changes are where an AI-assisted editor pays for itself every week.

The use case that has actually changed how I run products is quieter: internal administrative tooling I would never have budgeted calendar time to build by hand.

What I mean by “admin”

I am not talking about a polished customer-facing feature with design reviews and a roadmap slot. I mean the unglamorous screens and actions that keep production support humane: refunding a payment with one click as opposed to following a support procedure and manually updating the database; confirming that a location actually matches a pin dropped on a map; reconciling a stuck job; or answering “what happened to this user’s data?” without connecting to production and writing SQL by hand.

Those workflows are almost always possible with SQL, spreadsheets, and a sequence of support procedures, but they are slow, error-prone, and they do not compose into anything you can hand to someone else on the team.

Why Cursor made the difference

Historically I would have filed that work under “someday,” because the honest ROI never cleared the bar against feature work. With a capable model in the loop, the cost shifts: scaffolding CRUD, wiring policies, calling payment APIs, dropping a map component next to a read-only record, and adding an audit log line are all tedious rather than unknown. The hard parts—business rules, naming, and not doing anything stupid in production—stay mine. The mechanical parts stop being the reason the tool does not exist.

The result, in practice, has been that I stay almost completely out of the database for day-to-day support. Manual runbooks (“look up X, then Y, then paste into Z”) have turned into one or two clicks inside the app, with the same guardrails I would insist on for any other code path.

The takeaway

If you are evaluating Cursor only as a faster way to sketch UI or tweak a controller, you are underselling it. The highest leverage I have seen is turning “we will just do this carefully in prod once in a while” into software—small, boring, and worth more than almost any greenfield feature because it compounds every time support touches the system.

For broader context on how Cursor fits into my current stack, see 2026 Dev Tooling.