Small Bridges, Big Friction Removed

Situation

Three separate owner-led businesses, three versions of the same disease: a human manually ferrying data between two systems every single day. A restaurant re-keying online orders for the kitchen. An IT shop re-typing a SharePoint spreadsheet into Square. A property business hand-copying supplier invoices into records.

Problem

None of these problems is big enough to buy enterprise software for, and that is exactly why they persist. Each one costs minutes per occurrence, occurrences per day, every day, forever. And each is a place where a typo becomes a wrong order, a wrong price, or a wrong payment. The fix has to be small, cheap, and run without a developer on retainer.

Approach

Build the smallest bridge that removes the human from the loop, deploy it on infrastructure the client controls, and document it for handoff. A thermal-printer bridge that turns online orders into kitchen tickets automatically. A sync service that makes SharePoint the single source of truth and pushes it into Square. An extraction pipeline that reads supplier invoices into structured records.

Architecture and key decisions

• Right-sized scope. Each bridge does one job. No platform, no dashboard sprawl, no subscription a small business will resent in a year.
• Source-of-truth discipline. The catalog sync did not speed up re-keying. It abolished it, by declaring one system canonical and making the other follow.
• Print is still an API. The restaurant bridge speaks directly to ESC/POS thermal hardware because the kitchen’s interface is paper, and meeting a business where it actually operates beats forcing new screens on it.
• Handoff as standard. Each delivery includes documentation the owner’s people can follow without me.

What shipped

Three integrations, each built for a real client and documented for handoff: the thermal-printer order bridge, the SharePoint-to-Square catalog sync, live in production, and the invoice extraction pipeline.

Outcome

The catalog sync abolished a daily re-keying loop and runs in production. The printer bridge and invoice extractor are built and delivered; per-business usage and time-saved metrics get cited only as run history accumulates, measured numbers only.

What this demonstrates

Operational AI rarely starts with a moonshot; it starts with the 20-minute daily task nobody questioned. I find those, fix them at the right size, and use them to teach owners what automation can do, the diagnostic conversation that often precedes the bigger transformation.