Government organizations face unique constraints around security, connectivity, and GPU access that make large language models impractical. Purpose-built small language models (SLMs) offer a viable alternative by running locally, requiring less compute, and enabling verifiable source grounding. SLMs can be tailored to specific agency needs, support advanced search over unstructured data, comply with regulations like GDPR, and reduce hallucinations by working from verified sources. Gartner predicts specialized small models will be used three times more than LLMs by 2027. The recommended starting point for public sector AI is improving search rather than deploying chatbots.
Sort: