A software company that happens to use AI

Before SGH Logic existed as an LLC, the work did. Production document extraction for crop insurance paperwork. Computer-vision takeoff reading construction drawing sets, with published per-category accuracy. An operations and financial platform for a private club, with double-booking made structurally impossible at the database level. A field service platform with scheduling, payments, and photo-verified completion. A real-time AI product load-tested to a thousand concurrent sessions with every failure counter at zero.

That is the difference between this and a vendor who has read about AI. The case studies carry measured numbers because the systems were measured.

One engineer, on your project, from discovery to production. That is a constraint as much as a promise — it means we take on a small number of clients at a time, and it means the person in your discovery meeting is the person who writes the code.

Source code goes in your GitHub organization from day one. Infrastructure goes in your accounts. If we stop working together, you keep everything and any competent developer can pick it up. A vendor with a bus factor of one is fine for the vendor and unacceptable for the client, so we deliberately remove the hostage situation.

We publish accuracy numbers including the unflattering ones, because an AI system that makes a mistake the user was told would not happen is a system that gets abandoned.

Because it is home, and because being nearby genuinely changes the work. Watching someone do the task you are about to automate surfaces things no interview ever does — the workaround nobody mentions, the second spreadsheet, the customer who always calls at 4:50.

We work remotely too, and plenty of the delivery happens that way. But discovery is better in your building, and most Midlands businesses are inside an hour of the office.