Ask any practising CA how long it takes to research a moderately complex tax question (say, the TDS implications of a cross-border software payment to a non-resident) and the answer is usually "a few hours, sometimes half a day."
In 2026, with AI tools available for almost everything, why does tax research still take this long?
The Fragmentation Problem
Indian tax law is not one law. It is a web of interconnected legislation:
- The Income-tax Act, 2025 (replacing the 1961 Act)
- The Income-tax Rules, 2026
- CBDT Circulars (hundreds of them, many still in force from the 1960s)
- CBDT Notifications
- The CGST Act, 2017 + SGST Acts for each state
- IGST Act, 2017
- GST Rules
- GST Circulars and Notifications
- Double Taxation Avoidance Agreements (DTAAs with 90+ countries)
- Finance Acts (annual amendments)
- Allied acts: Black Money Act, Benami Transactions Act, Customs Act, etc.
A single client question can require searching across 5-6 of these sources. The TDS question above requires checking the IT Act (Section 393-395 for TDS, Section 196 for non-resident payments), the relevant DTAA (say, India-US DTAA Article 12 for royalties), CBDT circulars on software payments (Circular 2/2024), and possibly the equalisation levy provisions.
The Cross-Referencing Problem
With the IT Act 2025 transition, CAs now face a double cross-referencing burden:
1. Old Act sections cited in existing case law, circulars, and client documents must be mapped to new Act sections
2. New Act sections must be understood in the context of old interpretations and precedents
There is no official comprehensive concordance table. CBDT published a partial mapping, but for less common sections, CAs must figure it out themselves.
The Notification Overload
CBDT issues notifications regularly, amending rules, extending due dates, and clarifying provisions. In 2025 alone, over 100 notifications were issued. Each one potentially changes the answer to a question a CA researched last month.
Keeping track of which notification supersedes which, and whether a particular circular has been withdrawn or modified, is a full-time job in itself.
Why Google Does Not Work
Google is good at finding documents, not at answering tax questions. When you search "TDS rate on software payment to non-resident India," you get:
- Tax advisory firm blog posts (often outdated)
- Quora answers (often wrong)
- Government PDFs (right content, hard to parse)
- News articles about Budget changes (not directly useful)
You still have to read through multiple sources, verify which law was in force, check for amendments, and cross-reference with DTAAs.
Why ChatGPT Does Not Work
ChatGPT is impressive for general questions. For Indian tax law, it has three critical problems:
1. Outdated training data: It does not know about the IT Act 2025. Its section references are from the old Act.
2. Hallucinated citations: It will confidently cite "Section 194C(1)" when the question requires 194J, or invent circular numbers that do not exist.
3. No cross-referencing: It cannot check whether a CBDT circular has been superseded, or whether a DTAA overrides a domestic provision.
For a CA, using a wrong section number in a notice response or a client advisory is not just embarrassing. It can have financial consequences.
The Real Cost
A CA spending 3 hours researching a question at a billing rate of Rs. 2,000-5,000 per hour is spending Rs. 6,000-15,000 on research alone. For a firm handling 200-300 clients, this adds up to lakhs per month in research time.
The solution is not to skip research; that is dangerous. The solution is to make research faster and more reliable. That is the problem TaxMarg is built to solve.