Why AI matters at this scale

The Maryland Department of Assessments and Taxation (SDAT) is a mid-sized state agency responsible for two critical public finance functions: uniformly assessing all real property in the state and chartering all Maryland businesses. With a staff of 501-1000 employees, it manages immense volumes of structured and unstructured data—from property characteristics and sales records to business entity filings. At this operational scale, manual processes become a significant bottleneck, risking delays, inconsistencies, and increased operational costs. AI presents a transformative lever to enhance accuracy, efficiency, and public service in a sector traditionally slow to adopt new technology. For an agency of this size, targeted AI applications can automate routine tasks, freeing expert staff for complex judgment calls and improving service delivery to citizens and businesses across Maryland.

Concrete AI Opportunities with ROI Framing

1. Automated Initial Property Valuation: Manually assessing over two million properties is resource-intensive. Machine learning models trained on historical sales data, property attributes (square footage, location, condition), and macroeconomic indicators can generate preliminary valuation estimates. This augments assessors' work, reducing time spent on routine valuations by an estimated 20-30%, allowing them to focus on complex properties and appeals, directly boosting productivity and potentially improving assessment uniformity. 2. Intelligent Document Processing for Business Filings: The agency processes thousands of business formation and annual report documents. AI-powered optical character recognition (OCR) and natural language processing (NLP) can automatically extract, validate, and populate data fields from uploaded PDFs and scans. This reduces manual data entry errors, cuts processing time from days to hours, and improves the experience for business filers, leading to higher compliance rates and lower administrative overhead. 3. Proactive Anomaly Detection: AI algorithms can continuously analyze assessment rolls and business filings to identify outliers—such as a property assessment deviating significantly from neighborhood trends or a business report with suspicious officer information. Flagging these for investigator review shifts the model from reactive to proactive, helping combat fraud, ensuring tax fairness, and protecting the integrity of the business registry with a relatively low implementation cost.

Deployment Risks for a 501-1000 Employee Public Agency

Deploying AI in this context carries unique risks. Budget and Procurement Constraints: As a government agency, SDAT operates under fixed appropriations and lengthy procurement cycles, making it difficult to acquire and iterate on cutting-edge AI tools quickly. Legacy System Integration: Core systems for property assessment (CAMA systems) and business registration are often decades old, creating significant technical debt and integration challenges for modern AI APIs and data pipelines. Change Management and Skills Gap: Staff may be accustomed to established procedures, and the agency likely lacks in-house data science expertise, requiring extensive training or costly consultants. Algorithmic Accountability and Bias: Any AI used for valuations or decisions must be transparent and auditable to maintain public trust. Biases in historical data could be perpetuated, leading to legal and reputational risk, necessitating robust governance frameworks.