Local governments have spent the past two years deploying generative AI chatbots for document summaries and employee support. Now some are connecting those assistants to geographic information systems, letting residents ask questions tied to a specific location, such as "What construction projects are happening within a mile of my house?" or "Is my property in a flood zone?"
Maricopa County, Arizona, and Washington, D.C., have both launched tools in the last year that add this spatial capability. Leaders in both jurisdictions say the systems make government data easier to access, but they also introduce new challenges around data quality and ensuring AI-generated answers match official records.
Mari and Marco in Maricopa County
Kacie Baker, geographic information officer for Maricopa County, and Aaron Judy, the county's chief of innovation and emerging technologies, began experimenting with conversational interfaces between 2017 and 2019, including Amazon Echo voice assistants and IBM Watson models. The work accelerated after Judy returned to the county's Office of Enterprise Technology & Innovation with a directive from leadership to "get this conversational AI off the ground."
The result is two twin agents built on IBM Watson conversational models: Mari, the public-facing assistant, and Marco, the internal employee-facing version. Both operate as orchestration agents with multiple sub-agents and tools, including a Retrieval-Augmented Generation search agent for text and a set of Esri GIS tools for spatial queries.
After testing 11,000 prompts, the system launched in June 2026 and now averages 706 users per month. Mari has fielded more than 12,000 questions since launch. About 14% concerned property taxes, nearly 8% involved pet adoption, and 3% related to county jobs.
The pet adoption queries illustrate how the spatial layer works in practice. A resident can ask Mari for available huskies, and the system returns a list showing where each dog is located, its age, whether it's a mix, and a link to the adoption record. Baker and Judy call the broader category "nearest neighbor" searches: the system converts an address into a map point, calculates distances to relevant data points, and returns the closest matches.
Because Mari handles personal information such as addresses, the county applies policy warnings, access controls, and limited retention schedules. Conversations are accessible only to designated staff through an internal curation dashboard and are retained for roughly 13 months before being purged. Guardrails keep the assistant grounded in official county data and prevent it from offering legal advice or disclosing sensitive information.
The contextual chaining is what Judy finds most useful. "The cool thing is, I can chain these together contextually, right? So now that she has my address, she knows my parcel number," Judy said. "I'm chaining together multiple spatial tools without having to change context. It's carrying it forward through the session. It's really powerful because then I don't have to tell you my life story."
DC Compass opens city data
Washington, D.C.'s Office of the Chief Technology Officer took a different route, building its AI assistant on top of the city's open data portal rather than a general constituent services platform. The tool, DC Compass, launched as a private beta in March 2024 in partnership with Esri, then moved to a public beta before entering full production in March 2026.
Chief Technology Officer Stephen N. Miller said the project traces back to Mayor Muriel Bowser's 2016 order establishing a chief data officer and open data policies. Chief Data Officer Matt Sokol, who took the role in April 2023, heard repeatedly from residents that the open data portal was difficult to navigate. DC Compass was designed to address that feedback directly.
The tool connects to more than 700 public datasets and allows residents to query that data in plain language and multiple languages. It sits on top of the city's Esri-based GIS environment, acting as what Miller called the AI "front door" to spatial and tabular datasets. Residents can ask for nearest-neighbor searches - where the closest parks or schools are to a given address - or request maps of crime data, traffic incidents, and 311 requests over a specific period.
OCTO is now planning to embed Compass into GIS hub sites and city webpages, turning existing geospatial infrastructure into additional conversational experiences. The District also envisions using the technology with higher-sensitivity internal data, though those applications would face stricter controls.
Why this matters for government professionals
These deployments show that generative AI and LLM tools are moving beyond text-only interfaces into spatial reasoning, which changes how residents interact with government data. For agencies considering similar projects, the Maricopa and D.C. examples point to two distinct implementation paths: a general constituent assistant with GIS tools attached, or an AI layer built directly on an open data portal. Both approaches require attention to data quality, retention policies, and guardrails that keep AI answers grounded in official records rather than model-generated assumptions. The work also signals a shift in how AI for government is being evaluated - not just on answer accuracy, but on whether residents can find the right data without knowing which department or dataset to search.
Your membership also unlocks:
