ARBA is a two-person firm that designs and runs AI systems for work that repeats. Instead of sending a proposal, we build a small working version on real data and show you what it found. The subject changes from one project to the next. The way we work does not.
How we start
In most firms the first thing you receive is a document describing what would happen if you paid. We do it the other way round. We build the first version before anyone asks for it, run it on real public data, and hand over what it produced — with the source and the date on every line, so you can check any one of them in about a minute and decide for yourself.
Two owners, no employees, nobody on a timesheet. Nothing here becomes more profitable by taking longer, and there is no account manager standing between you and the document.
The first system we built reads a government publication every week and reports which official decisions have changed while a public database still returns the old ones. We built it before we had anyone to sell it to. The next one will be about something else entirely — the subject is the project, not the firm.
How we work
The first working version exists before there is a customer for it. What you look at is the output, not a description of the output.
The document it came from, and the day it changed. Nothing is condensed to the point where you would have to take our word for the summary.
What it read, what it skipped, and what it could not reach. A system that only speaks when it succeeds is not telling you much.
How we engage