How AI is reshaping tax scrutiny—and why humans still matter

The report highlights how generative AI—the technology that powers tools like ChatGPT—can process enormous volumes of tax-related information in seconds. Unlike humans, who may take hours or days to analyse complex filings, AI can rapidly read, sort and interpret financial data, allowing it to flag suspicious patterns or anomalies almost instantly. This makes it a powerful early-warning system that could help administrators catch errors or evasive behaviour before they slip through the cracks.

Generative AI is trained on vast amounts of information to “understand" and generate human-like language. In simple terms, it functions like a super-intelligent research assistant for tax departments.

One of its most immediate applications is document summarization. In its “assistant" role, AI can scan hundreds—or even thousands—of pages of information and produce concise, easy-to-read summaries, saving officials significant time and effort.

Digital detective

In a more analytical “consultant" role, AI acts as a digital detective. Here, it examines complex datasets to identify subtle inconsistencies or potential fraud indicators.

For example, a tax officer can feed a full year’s worth of financial declarations into an AI model, which can immediately highlight anomalies—mismatches between the Annual Information Statement (AIS) and the Income Tax Return (ITR), incomes that do not align with reported expenses, or unusual spikes in transactions that deviate from typical taxpayer behaviour. These insights allow human officers to focus their scrutiny on genuinely high-risk cases.

At its most advanced stage, AI can operate as an autonomous agent—similar to a self-driving system for administrative tasks. In this role, it can automatically process routine functions such as verifying documents, matching data across systems or sending alerts, significantly reducing the manual workload on tax officials.

India’s groundwork

India has already laid the foundation for AI-assisted tax administration. Over the past few years, the Income Tax Department has aggressively digitized records, uploaded historical data, and expanded reporting frameworks for securities, property and foreign exchange transactions.

Combined with the ongoing redrafting of the Income-tax Act to improve clarity, these steps signal a clear move toward a more robust, technology-driven compliance ecosystem.

Limits of automation

However, AI is not foolproof. A recent Bombay High Court ruling highlighted this limitation. The Court set aside an assessment order in which the department relied on case laws it could not produce before the Court.

The Court cautioned: “In this era of AI, one tends to place much reliance on the results thrown open by the system. However, when one is exercising quasi judicial functions, it goes without saying that such results [which are thrown open by AI] are not to be blindly relied upon, but the same should be duly cross verified before using them. Otherwise mistakes like the present one creep in."

This reflects a broader issue known as AI “hallucinations," where systems generate incorrect or fabricated information—an issue documented globally and one that underscores the need for human oversight.

Human judgment

AI-based risk profiling is set to transform public finance in the years ahead. Governments may increasingly use AI to pre-fill tax returns, detect underreported income or inflated deductions, and provide data-backed insights to policymakers.

But while AI can compile, analyse and interpret information at scale, one essential truth remains: only humans can ensure that the right questions are being asked—and that automated outputs are applied judiciously.

Rakesh Nangia, managing partner and Neeraj Agarwala, partner, Nangia & Co LLP