Watch the Deep Dive of this post
Inside GSAi: How a New Ai Tool is Revolutionizing Government Data and Decisions
The U.S. government is increasingly looking towards artificial intelligence to boost efficiency and effectiveness. One exciting development comes from the General Services Administration (GSA), which recently launched its own internal generative AI tool, known as GSAi. While currently reserved for internal use, GSA leaders believe this tool holds significant potential for application across the entire federal workforce.
Launched in March, GSAi was designed with a clear goal: to increase internal productivity. It features a user-friendly chatbot, an application programming interface (API) for integration, and a sophisticated administration console. This console is particularly noteworthy because it allows GSA to evaluate different AI models for various purposes.
What sets GSAi apart is its deliberate focus on quality. Instead of concentrating solely on safety concerns, the tool prioritizes assessing the performance quality of each AI model based on the specific functional areas of the agency. For instance, GSAi can rigorously test how well a particular AI model answers questions related to procurement.
This emphasis on quality means GSA can continuously evaluate a model’s ability to perform its intended function over time, detect undesirable changes (known as model drift), and ultimately determine if the model is truly fit for purpose. This evaluation process helps identify whether interventions like fine-tuning or retrieval-augmented generation (RAG) are needed to improve performance. The aim is clear: to avoid simply giving the workforce a tool that might provide random or unreliable advice. GSA wants to be transparent about the tool’s capabilities and limitations. Having a robust quality measurement of performance is considered critical.
But GSAi has yielded a surprising and welcome secondary benefit. Its implementation has become a powerful catalyst for crucial, and perhaps overdue, conversations about data quality and metadata. GSA is deliberately using AI as a kind of “Trojan horse” to bring these topics to the forefront. Putting an AI engine to work on existing data stores quickly highlights the strengths – and more often, the weaknesses – in the underlying data.
This direct demonstration of data quality issues has generated significant energy and attention within GSA leadership (“the A suite”), prompting discussions about the essential need to address metadata. It’s been described as a “really nice little happy accident”.
In essence, GSAi isn’t just another AI chatbot; it’s a strategic tool focused on ensuring the reliability and suitability of AI models for specific government tasks. Furthermore, by highlighting the foundational importance of clean, well-structured data, it’s driving essential infrastructure improvements necessary for effective AI adoption across the government. While still in its early stages, the approach taken with GSAi focusing on measured quality and leveraging AI to expose data needs offers a promising model for government AI implementation nationwide.
AI in Government, GSAi, GSA, Generative AI, Data Quality, Metadata, Government Technology, Zach Whitman, AI Quality, Fit for Purpose