Ask your GIS questions in plain English.
GeoChat sits beside your map and answers questions about your ArcGIS layers, during a wildfire emergency or on an ordinary Tuesday. Your team gets answers in seconds, and your GIS staff get out of the "can you pull me a map" queue.
The live demo runs on public California power-grid data. Pilots run on your own layers.
2 lines and 2 plants are within 2 miles.
- Lines: one 230 kV, one 115 kV
- Plants: one solar, one hydro
The AI plans the analysis. Your ArcGIS does the math.
Questions run against your own feature services. The model only sees counts, statistics and a few sample rows; geometry goes straight from your services to the map.
Ask
"How many miles of our lines run through Tier 3 fire-threat areas in this county?" Name a place, click a feature, or draw a point, line or area on the map.
The assistant runs the tools
It picks from query, statistics, proximity, buffer, overlay and multi-layer screening tools, and checks every filter against your real field names before it runs.
Act on the result
Results are drawn on the map and listed in a table. Copy a short summary or a one-page PDF, or export the full data as CSV, GeoJSON, Shapefile, KMZ or a PDF field report.
One tool for the EOC and the everyday queue
The same assistant handles the wildfire at 2 a.m. and the routine "how many of our assets are in…" request on Monday morning.
Incident mode
Pin the fire perimeter once, and everyone in the EOC asks against the same incident. Re-run screenings as perimeters update. Every question and answer is logged for the after-action review.
Self-serve answers
Dispatchers, planners and field crews get counts, crossings, buffers and environmental screenings themselves. The GIS team sets up the layers once and stays in control of what's connected.
Try these in the live demo
Each link opens the demo and asks the question. For environmental screening, open the demo, draw a route on the map and ask "Which environmental layers does this route affect within 500 ft?"
Your data stays yours
GeoChat is designed for organizations that can't put grid data in someone else's cloud app.
Your credentials, your portal
In production, every query runs with the signed-in user's own ArcGIS token, so the assistant can only see what that user can see.
Deploy it inside your walls
Runs on your servers or in your cloud account, behind your SSO. Bring your own model through Amazon Bedrock or Google Vertex AI.
Read-only by default, and logged
The built-in tools only read your data. Anything that would change your portal, like creating a layer or running one of your own web tools, waits for a person to confirm. Every question, step and answer is logged.
The public demo uses public data and no sign-in.
Questions we get asked
Does our data train the AI model?
GeoChat calls the model you choose through your own cloud account: Amazon Bedrock or Google Vertex AI. Under those services' standard terms, your prompts aren't used to train their models; confirm this against your own agreement. The model sees your question, your layer and field names, and short summaries (counts, statistics and a few sample rows), never your full datasets or geometry.
Which ArcGIS products does it work with?
Feature services in ArcGIS Enterprise and ArcGIS Online, through the standard ArcGIS REST API. Your GIS team lists the layers once in a configuration file, and GeoChat reads each layer's fields and values itself.
What happens when the AI gets something wrong?
Every answer shows the steps it ran and the exact filters it used, so anyone can check it. Filters are checked against your real field names before they run, the assistant asks when a name is ambiguous, and its built-in tools only read your data. Results are screening-level answers for people to act on, not a replacement for engineering review.
How accurate are the locations on the map?
Results come straight from your feature services, so they're as accurate as your data. The public demo uses public datasets, some of which are simplified for national-scale maps; your own data will usually be more precise.
What does it cost?
Two parts: the GeoChat license, and model usage billed by your cloud provider per question. Pricing depends on the number of users and deployments. The pilot measures your real usage so the numbers are based on your questions, not estimates.
What do you need from us for a pilot?
Read access to the feature services you want to use, a place to run it (your server or cloud account), and a short list of the questions your team asks most. We handle the configuration.
Run a supervised pilot
We configure GeoChat against your ArcGIS, with your layers and your territory, and your team uses it on real questions for 30 days, with us on call.
- Fixed fee, credited in full toward the annual license
- Your data, your portal, your SSO; nothing shared
- Evaluation report on 30 of your real questions, with timed answers
- No exclusivity, no revenue share, cancel anytime