Andhra Pradesh Demonstrates AI-Driven GST Administration

Economic Times CFO reports that the department demonstrated AI-integrated return scrutiny, audit, and inspection processes spanning case selection, initiation, discrepancy identification, notice issuance, taxpayer responses, assessment, and adjudication. It also demonstrated an AI agent for generating reports and dashboards, plus a knowledge hub containing sectoral studies used in audits.

The publication reports that AI engines developed by Microsoft and Google support the processes. It also reports that Andhra Pradesh requested an API to integrate GST data with its own systems.

The Hans India reported that GST Council Secretariat Additional Secretary Pankaj Kumar Singh described the initiative as a potential model for the Centre and other states after a two-day departmental visit. The same report said officials watched demonstrations of return scrutiny, audit, inspection, investigation, and legal case-management systems, and were briefed on using GST data in real-time economic analysis, including Gross State Domestic Product estimation.

The reported implementation illustrates a broader public-sector analytics pattern: value depends not only on a model's ability to surface anomalies, but also on reliable access to transaction data, case-management workflows, and evidence that can be reviewed during assessment and adjudication. In comparable tax-administration deployments, practitioners need to distinguish predictive or anomaly-detection outputs from the rules, data lineage, and human review processes that support enforceable decisions.

Andhra Pradesh's demonstrations covered operational systems rather than a disclosed model architecture or published performance evaluation. The available reporting does not specify the models used, error rates, false-positive rates, or the methodology behind the reported revenue figure. Those details would be material for evaluating how readily the approach can be reproduced across jurisdictions with different data quality, tax rules, and integration constraints.