Tax practice in Malaysia is in the middle of two simultaneous shifts. The first is regulatory — the rollout of e-invoicing under LHDN, the broadening of SST under Royal Malaysian Customs, and the steady tightening of transfer pricing requirements. The second is technological — the maturation of AI tools that genuinely accelerate tax research, return preparation, and advisory work.
This article is the practitioner's view of where AI is delivering value in Malaysian tax work in 2026, and where it must stop because the regulatory or professional bar is too high. Like our audit article, the framing is conservative — AI as accelerant, never as substitute for professional judgement.
1. Tax research
The clearest first win. Searching across Income Tax Act 1967 sections, LHDN Public Rulings, Practice Notes, Technical Guidelines, and case law to build a position on a specific client question. AI tools that index these sources and respond to natural-language queries can compress what previously took half a day of research to half an hour.
The discipline that matters: every AI-generated research output must be verified against the underlying source. The tax adviser remains the author of the position. AI cannot cite Section 33(1) and Section 39(1)(c) interactions correctly without supervision; the technology drafts, the practitioner reviews and concludes.
2. Return preparation
AI accelerates the production work — extracting structured data from financial statements, preparing schedules, populating return forms — while human review remains for sign-off. The patterns that work in Malaysian tax practice in 2026:
- Capital allowance schedules from fixed asset registers, with AI handling categorisation per Schedule 3 and the practitioner verifying.
- Reconciliation of accounting profit to taxable income, with AI proposing the standard adjustments and the practitioner reviewing.
- Section 132 (DTA) and Section 133 (unilateral) credit calculations, where AI handles the arithmetic and currency work.
- RPGT and stamp duty computations on transactions, where AI handles base calculations and the practitioner reviews exemptions and reliefs.
3. Transfer pricing documentation
Among the heaviest documentation burdens in Malaysian tax practice. Master file, local file, country-by-country reporting where applicable. AI changes the production economics significantly. Functional analysis sections can be drafted from interview transcripts. Industry analysis sections can be drafted from public databases. Comparable company analyses can be accelerated through AI-assisted comparable searches and rejection commentary drafting.
The professional discipline: the practitioner remains responsible for the arm's-length analysis and the conclusion. AI accelerates documentation; AI does not replace the economist's judgement on what the appropriate range is.
4. e-Invoicing readiness
Specific to Malaysia in 2026. As LHDN's e-invoicing rollout has expanded across taxpayer groups, AI has become useful for: validating e-invoice content against business rules before submission; reconciling submitted e-invoices to the GL and AP/AR records; surfacing exceptions and rejections that need investigation. For taxpayers with high transaction volume, AI-assisted e-invoicing validation is increasingly the default.
5. SST treatment analysis
The complexity of SST treatment across goods, services, exemptions, and zero-rating creates substantial advisory volume. AI tools can take a description of a transaction and draft a position on the SST treatment with reference to the relevant orders and guides. The practitioner's role is to verify the analysis, identify edge cases the AI may have missed, and conclude.
6. Tax controversy support
For LHDN audits, queries, and disputes, AI accelerates the documentation work — pulling relevant transactions, drafting position papers, summarising prior correspondence. The substantive defence and the discussions with LHDN officers remain with the human practitioner. The combination is faster, more thorough preparation than the same practitioner could produce manually.
7. The lines that must not be crossed
- AI cannot conclude on tax positions. Professional sign-off remains the practitioner's.
- Client confidential data on consumer LLMs is a breach. Enterprise tiers only.
- Privileged communications must remain privileged. Counsel-led advisory work has additional confidentiality requirements that AI deployment must respect.
- AI hallucinations on tax law are real and dangerous. Always verify against primary sources. The Income Tax Act takes precedence over any AI output.
8. The 90-day starting plan
- Days 1–30: Vocabulary alignment with the tax team. Approved-platform list. Pilot use case — usually tax research or return preparation reconciliation.
- Days 31–60: Run the pilot through one complete tax cycle (a quarterly return or a specific advisory engagement). Track time savings against comparable prior work.
- Days 61–90: Partner-level review of pilot outcomes. Decide on extension to additional use cases. Establish quarterly review of LHDN and MIA guidance updates.
For Malaysian tax practitioners formalising this, our AI Agentic Automation programme covers the workflow stack and our AI Vibe Coding programme equips tax staff to build internal tools. HRDC SBL-KHAS claimable for eligible employers.
