![Streets of Washington DC](/sites/g/files/omnuum10826/files/styles/hwp_21_9__1920x825/public/2025-08/vlad-tchompalov-C6zoz9Rw2bk-unsplash.jpg?itok=Bg8zc3uu) # Washington, DC Redefines Open Data with GenAI The District of Columbia's new AI-powered assistant, DC Compass, is transforming open data access by making city information easier to explore and interact with. August 11, 2025 [ Betsy Gardner ](/people/betsy-gardner) Open data is entering a new era; from the mid 1990s to late 2010s, governments have been releasing publicly available data through different open data portals, budget transparency sites, and online operational dashboards. Yet the explosion of generative artificial intelligence (GenAI) in 2022, enabled by large language models (LLMs), has ushered in a paradigm shift in open data. The District of Columbia is one of the early leaders in this new era, thanks to the [DC Compass AI assistant](https://opendata.dc.gov/pages/compass). By marrying GenAI with the city’s open data portal, [Open Data DC](http://opendata.dc.gov), residents can now engage with city data like a data scientist and analyst, regardless of technical training and experience. ## Improving Open Data “How do we find new ways for residents to be able to interact with and explore our datasets?” This was the foundational question for DC’s Chief Data Officer Matt Sokol, who spearheaded the DC Compass pilot. Previously, Data-Smart City Solutions has written about the [challenges with simply releasing open data](https://datasmart.hks.harvard.edu/news/article/equal-access-open-data-and-equity) without contextualizing the information or offering instruction. Training and outreach programs, while effective, are difficult to enact at scale and are cost and labor intensive. In 2003, the District’s Enterprise Data team released the first version of an open data catalog that focused on centralized GIS data. Now, DC Compass is a way to expand the reach and usefulness of the Open Data DC platform, which has more than 1,400 datasets, maps and apps with thousands of records. Two important factors make DC Compass an effective tool for helping users unlock open data. The first is natural language, which allows users to navigate through the data by asking questions in common vernacular. Built for ease and accessibility, users can query the tool using voice or text modes, and in multiple languages. The second is contextualization; DC Compass will correlate datasets with specific city initiatives. By doing this, the tool not only provides a technical answer but also helps link the data to the relevant department or project, and directs users to the related initiative’s “hub site.” For example, if a user asks about pothole repair, they would be provided with information on infrastructure projects and a link to those pages. ![DC Compass answering "What species of trees are on the block of 220 I Street SE?" with a list of tree types and a map showing their locations](/sites/g/files/omnuum10826/files/2025-08/CD%20Compass%20beta.png) ## Implementing AI Tools To achieve what Sokol described as a “balanced approach to manage both the excitement and risks of a new technology like AI,” the OCTO team did significant work leading up to the DC Compass pilot. A fundamental component of this work is the leadership from Mayor Muriel Bowser, who issued a [mayoral order articulating DC’s AI values](https://techplan.dc.gov/page/mayors-order-articulating-dc%E2%80%99s-artificial-intelligence-values-and-establishing-artificial) at the start of 2024. By establishing whole-of-government guidelines, which include “clear benefit to the people,” “safety and equity,” and “privacy and cybersecurity,” the OCTO team had a clear path to follow. For example, DC Compass has an obvious value to users by promoting equitable access and understanding of the open data, encouraging residents to better engage with city initiatives. Users cannot modify any data, all of which is classified as ‘Level 0’, meaning it contains no personally identifying information. Also, according to Sokol, DC Compass performs “very lightweight AI functions with Retrieval Augmented Generation,” which limits the environmental impact. The [DC Compass AI Values Alignment Report](https://techplan.dc.gov/sites/default/files/dc/sites/itstrategicplan/publication/attachments/Final_Draft_DCCompass_Alignment.pdf) details how the tool aligns with District values. The next critical piece of this work was DC’s partnership with Esri, the geographic information system (GIS) company. “Most datasets are geographically-related,” said Sokol, “even if it’s not explicitly ‘GIS’.” For example, data about 311 calls or retail developments might not have exact coordinates, but there are still locational aspects to the answers. Instead of asking an LLM to perform complex geo-spatial analysis, the Compass combines Esri’s GIS power with the natural language. In this way, the LLM element can act as a translator for the user, intaking the question, determining the type, and then handing it off to the GIS apparatus. Once the analysis is done, that information is sent back to the LLM which translates back to the user. ![DC Compass pulling up all datasets related to schools, including info on independant and public schools, heads of school, and test scores with a map of locations](/sites/g/files/omnuum10826/files/2025-08/DC%20Compass%20beta%202.png) Finally, clean and high-quality data is necessary for optimal functioning; an AI tool can only be as good as the data it pulls from. “There's a lot of prep that we had already done,” said Sokol, “we've spent a lot of time creating good quality data sets and curating the metadata.” Thankfully, due to long-standing rigorous data standards and responsible data management, DC’s open data was well positioned to incorporate AI tools. ## Sharing Lessons Learned DC Compass was launched as a pilot just over one year ago, so Sokol shared reflections on the past several months and plans for the future. 1. **Data Quality is Key**: to successfully utilize GenAI, the data — and meta-data specifically — must be high-quality. Thankfully DC had put in this work already, but if this isn’t the case in other municipalities, cleaning and organizing the data is the first step. 2. **Leadership Alignment**: Due to Mayor Bowser’s leadership on AI and GenAI, the OCTO team was able to move forward with clear standards, knowing they were in alignment with mayoral guidance from the beginning. “These policies and orders which have been put in place by the mayor are really helpful” in framing and shaping the Compass work, said Sokol. 3. **Think Holistically**: Lockstep with each pilot, tool, or test is a values-guided policy. Sokol described the process as “an all-encompassing, holistic look at how these tools are being launched, how they're being used, and how we as the government are going to continue to deliver value.” Sokol expects that DC Compass will continue to improve, as both GenAI and the tools and capabilities from Esri continue to evolve. “We hope to take advantage of those new capabilities and features for DC Compass to continue to innovate and improve it,” he said “to take our ability to use open data to the next level.” See also:- [ Artificial Intelligence ](/topics/artificial-intelligence) - [ Data Visualization ](/topics/data-visualization) - [ GIS ](/topics/gis) - [ Open Data ](/topics/open-data) Share on:- [ Facebook ](#) - [ Twitter ](#) - [ Linkedin ](#)