Innish Labs
A quiet argumentfor boutique spatial AI
Most organizations don't need a platform. They need the right small team, working in the open, delivering a specific system that works. Innish Labs is a two-to-five-person practice at the seam of GIS and AI, hired by directors who've been burned by bigger firms and now want to work with people who will actually ship.
The practice
Grounded in the territory
Spatial systems fail when they're designed from the org chart instead of the ground. We start every engagement in the data (parcels, networks, easements, the field notes nobody digitized) because that's where the real constraints live. A week spent with the actual records saves a quarter spent arguing about requirements.
What we do
Four ways in
Every engagement is scoped to the thing that has to work. These are the shapes the work usually takes.
Spatial retrieval
Retrieval systems that understand geography, so a zoning or asset question comes back grounded in the right place, with the source attached.
Agent operations
Agents that do real work inside real constraints: permitting queues, inspection scheduling, records requests. Boundaries first, then autonomy.
Field-to-decision systems
Getting what a crew records in the field into the hands of the person who decides, the same day, without a nightly batch job in between.
Team enablement
We'd rather make your analysts dangerous than make you dependent on us. Engagements are designed to end.
A closing note
Built to ship,sized to care
The work we care about (the map that settles an argument, the record that survives an audit, the answer that arrives while the decision is still open) rarely comes out of a platform. It comes out of a small number of people who understand the ground, sit with the constraint, and stay until the thing runs without them. We keep the practice small on purpose.