On Standing Up Applied AI for Real Businesses

A great deal of software was never built for the businesses that keep the economy running. We take the opposite starting point. We work alongside owners and operators to put AI where the work actually happens: the reports someone rebuilds by hand every Monday, the questions that never get answered because the data lives in six places, the workflows that quietly depend on one person remembering everything.

Our deliverables are concrete. First, a unified data lake: one governed place the organization’s data lives, read-only and answerable in plain language. Second, internal AI applications: dashboards, assistants, and workflows a team can build, use, and publish for itself. Where we partner closely, we rebuild the public face too.¹

We do not deliver a black box. Everything we build stands on Setoku, an open-source context layer we author and maintain. It provides a governed path between a model and a company’s data, self-hosted on infrastructure the client controls.

Three properties are load-bearing. The system is self-hosted, so data never leaves the client’s walls. It is open source, so there is no lock-in and no privileged seat the client cannot inspect. And it is human-in-the-loop: the model proposes, and a person approves before anything is committed. When an engagement ends, what we installed is still standing, still theirs, still auditable.

As a demonstration of the method, we designed and built Hedgy, an AI recruiter, from end to end. It runs today in production on the same stack we would stand up for a client. We offer it not as a case study with borrowed numbers, but as a working artifact: the shortest version of our claim that we ship what we describe.² The reader is encouraged to inspect it directly at hedgy.works.

We take on a small number of engagements at a time. Correspondence is welcome at hello@hedgylabs.com. Tell us what is slow, what is manual, or what you wish you could see; we will tell you what we would stand up first.