Independent software studioDavidson, North Carolina35°29′57″N 80°50′55″W

A small surveyor's plat: a five-sided closed traverse with bearings and distances along each course.N 12°11′01″ E180.06′N 72°21′00″ E184.69′S 32°54′19″ E161.99′S 47°51′45″ W169.92′N 84°09′38″ W176.92′P.O.B.2345
Plate IClosed traverse · 1.16 ac.

Intelligent tools for the physical world.

An independent software studio building AI-native tools for geospatial work, infrastructure, and real-world operations.

Sta. 1+00The Studio

AI should work for you and take the friction out of technology.

mIIcai is a small, independent studio building software where AI handles the tedious parts and people stay in charge.

Field Agent, for public utilities, is the first product. The studio has room for other AI and GIS projects as it grows.

Studio
mIIcai
Type
Independent software studio
Funding
Bootstrapped
Based
Davidson, North Carolina
Founder
Dustin Stephens
Current work
Field Agent, in development

Legend

  1. Practical intelligence

    Built to get real work done, not to show off the technology.

  2. The physical world

    Assets have locations, and the location usually matters.

  3. Open technology

    Open standards and open-source geospatial tools, so systems stay connected and maintainable.

  4. Human control

    Agents work inside permissions and approvals, and every action is on the record.

Sta. 2+00Field Agent

First product

Field Agent

In development

Operations and asset management for public utilities, built on GIS, with an AI agent that uses the application the way your staff do.

Ask it what’s near a main break, have it pull up a job, or tell it what you did today and let it draft the log. It works through the same tools as the screens, under your permissions, and nothing saves until you confirm. It’s built to run on open models a utility can host itself.

Field AgentWater · DistributionConcept
Conceptual utility mapA neighbourhood street grid with a water main, valves, hydrants, a sewer line with manholes, and service connections. A reported main break near Elm Street and Third Avenue is marked, with the nearby main and two valves highlighted within a 200-foot search radius.ELM STBIRCH STCEDAR ST3RD AVE4TH AVE200 ftV-1V-2C
  1. Request · Resident

    “Water coming up through the street near Elm and 3rd.”

  2. Supervisor

    Work order created. What’s near the break?

  3. Agent

    An 8″ main on Elm St, with valves V-1 and V-2 within 200 ft. Showing them on the map.

  4. Crew · Voice

    “Started at seven. Two hours, one 8-inch repair clamp.”

  5. Agent

    Drafted: start work, 2.0 h labor, 1 × 8″ repair clamp.

    Confirmed and saved by technician
  6. Supervisor · Review

    Reviewed and closed.

A supervisor reviews and closes the work. Approvals stay with people, and every step is in the history.

Conceptual illustrationThe intended workflow, not a screenshot of shipping software.

What makes it different

  1. An agent that uses the application

    The agent calls the application’s own tools, about thirty of them, with the signed-in user’s permissions. It searches records, answers spatial questions, opens the right screen, moves the map, and drafts updates. No screen-scraping and no model-written SQL.

  2. The agent drafts; people decide

    AI changes data only through proposals a person confirms. Completing, approving, and assigning stay with people, every tool call is logged, and each new agent capability has to pass a test set with zero safety failures.

  3. Geography first

    Every asset, request, work order, and inspection has a location. Built on PostGIS and an open MapLibre map, with field edits reviewed and kept in each asset’s history.

  4. Works offline

    Jobs, inspections, notes, and photos stay on the device through dead zones and sync on reconnect. Conflicts are shown to the user, never silently overwritten.

Where it stands · October 2026

Working in development

  • An agent that searches, maps, navigates, and drafts across the application
  • Service requests, work orders, and inspections
  • Asset maps with reviewed edits and history
  • Offline field work

Next

  • Testing on real field devices
  • A first deployment

Starting with water and wastewater utilities, then other public works.

Interested in Field Agent?

I’m always glad to compare notes with people who work in utilities and GIS.

Get in touch

Sta. 3+00Founder

Why I’m building this.

I grew up around land and surveying. As a teenager I spent time surveying with my father, and the interest in maps and the ground they describe never left. GIS was a natural next step.

I’ve spent about fifteen years in GIS and IT, nearly a decade of it in municipal water and wastewater utilities: running enterprise GIS and asset management systems, leading GIS teams, and supporting the crews who rely on those systems every day.

I kept wondering how much better those systems could be if they were built for how people actually work, with agents that can help. mIIcai is where I’m finding out, starting with Field Agent.

Sta. 4+00Contact

Let’s build something useful.

Startup programs, technical collaboration, mentorship, or just comparing notes on AI, GIS, and infrastructure. I’d like to hear from you.