Tax and audit teams use AI for research and evidence, and so do the tax authorities and audit firms looking at you.
Tax uses
- Research: summarize legislation, rulings and guidance; compare positions across jurisdictions, with every citation checked.
- Provision support: gather data, draft workpapers and explanations.
- Indirect tax: classify products and transactions; check invoices for the right tax treatment.
- Document review: read contracts for tax-relevant terms (permanent-establishment risk, withholding).
Audit and controls uses
- Full-population testing: analyze all transactions rather than a sample to find outliers.
- Evidence gathering: assemble support for auditor requests.
- Control documentation: draft and update narratives and risk-control matrices.
The other side uses AI too
Tax authorities use machine learning to choose whom to audit. The IRS said in September 2023 that it was using AI to help select large partnerships for examination. Audit firms are investing in AI tools too, though in 2024 the PCAOB staff found their use was still mostly administrative and research-focused, with firms expecting AI to augment, not replace, auditors.
Practical rules
- Never file on an unverified AI answer. Tax research output needs citations checked against primary sources.
- Assume the authority's analytics will see patterns in your data that a sample would not; make your own data consistent first.
- Agree with your auditor how AI-generated evidence and analytics will be relied on.
Measures
Research time per question, share of citations verified, provision cycle time, exceptions found by full-population testing, and audit requests answered within the agreed time.