AI for Accountants in Oman: Daily Uses, Client Confidentiality and What to Check
Accountants can use AI every day for the parts of the job that are mostly reading, drafting and reorganising:
- extracting data from invoices and statements;
- explaining reconciliation differences;
- drafting variance commentary and client emails;
- summarising contracts and regulations;
- writing spreadsheet formulas.
AI should not be the source of a number in a filed return, or the authority on a tax treatment.
The working rule is simple: AI drafts, the accountant decides. Use approved business tools, keep client data out of public ones, recompute every figure that matters, and check tax positions against the law and the Tax Authority’s own guidance. Then keep the AI-assisted workpapers with the records the law already requires you to retain, which for VAT is ten years.
The short answer
- Best uses: document extraction into drafts, reconciliation exceptions, first drafts of commentary and client letters, contract summaries for accounting review, spreadsheet formulas, research starting points.
- Never: let AI post entries without review, decide a tax position, or make a professional judgement such as materiality or impairment.
- Ethics still apply. ICAEW’s guidance maps generative AI onto the profession’s five fundamental principles. Among them: “Generative AI tools do not make up for a lack of professional knowledge and skill.”
- Confidentiality: no client data in public AI tools; business tools only, with data terms you have read.
- Oman specifics: VAT records kept for ten years (fifteen for real estate), e-invoicing obligations from 2027, and the Personal Data Protection Law for client and employee data.
Where AI helps in a finance team or practice
| Task | How AI helps | What the accountant checks |
|---|---|---|
| Invoice and receipt data | Extracts supplier, date, VAT number, lines and totals into draft entries | Validation rules, duplicates, a sample against the originals |
| Bank reconciliation | Suggests matches and explains unmatched items in plain language | Every match before posting; nothing auto-posts |
| Month-end commentary | Drafts variance explanations from figures you supply | Figures re-checked; explanations confirmed with budget holders |
| Client correspondence | Drafts requests for documents, reminders and explanations in Arabic or English | Accuracy, tone, and no advice beyond the engagement |
| Contract and lease review | Lists payment terms, escalations, durations and options for accounting assessment | The accountant applies the standards and records the judgement |
| Spreadsheets | Writes and explains formulas, lookups and checks | Tested on data with known answers before use |
| Research starting points | Points to the provisions and topics a question touches | The law or guidance itself, read in full |
| Procedures | Turns notes into a draft month-end checklist or policy | The process owner approves the final version |
AI drafts, the accountant decides
ICAEW’s generative AI guide applies the five fundamental principles of professional ethics to these tools. Its wording is a useful checklist for any firm:
- Integrity: “being transparent and honest in the use of Generative AI is essential.”
- Objectivity: be alert to bias in AI outputs.
- Professional competence and due care: “Generative AI tools do not make up for a lack of professional knowledge and skill.”
- Confidentiality: “consideration must be given to the type of data that is input into Generative AI tools”, which means “not loading confidential information into public Generative AI tools”.
- Professional behaviour: avoid any use that would discredit the profession.
The same guidance calls for “exercising professional scepticism when reviewing the output”. In practice, that means treating an AI draft the way you would treat work from a new junior: useful, fast, and checked line by line before it carries your name.
Oman specifics an AI workflow must respect
VAT records. Article 70 of the VAT Law, issued by Royal Decree 121/2020, requires tax invoices and accounting records to be kept for ten years after the end of the tax year in which the return was filed, and fifteen years for real estate. If AI-assisted workpapers support a return, keep them with the records, in a form that will still be readable in a decade. Quarterly returns are due within 30 days of the quarter end, so checking time has to fit inside that window.
E-invoicing. Under Tax Authority Decision 189/2026, issuing tax invoices electronically becomes a legal obligation on 1 April 2027 for businesses with annual supplies above OMR 5 million, and on 1 October 2027 for the rest. As more invoices arrive as structured data, AI extraction becomes a fallback for foreign suppliers and older documents. AI can also help with readiness work, such as finding gaps in customer and supplier master data. E-invoicing in Oman covers the timetable.
Personal data. Payroll files, individual customers and employee claims contain personal data. Sending them to a model hosted outside Oman is a cross-border transfer under the Personal Data Protection Law and its Executive Regulation. When you process clients’ data on their behalf, your engagement terms should say how, including which AI tools are used. A policy for staff using AI tools and what to put in your software vendor’s contract cover the controls.
Professional confidentiality. Your engagement letters and your professional body’s rules on confidentiality apply whatever tool you use. If a tool’s terms allow it to keep or train on your inputs, it is not the place for client work.
What not to use AI for
- The numbers in a return. Recompute every figure from the ledger, outside the AI tool.
- A tax position. AI can point you to the relevant provisions; you read them, and the Tax Authority’s guidance, yourself.
- Professional judgements. Materiality, provisions, impairment and going concern can be drafted as considerations, but the judgement and its documentation are yours.
- Posting without review. Draft entries, never final ones.
- Anything in a public tool with client data in it.
A prompt that works for variance commentary
You are drafting month-end commentary for our management accounts. Use only the figures in the table below. For each line where the variance against budget is above 10% and OMR 1,000, write one sentence explaining the variance, using only the explanations listed under “Notes from budget holders”. Where no explanation is listed, write “Explanation needed” — do not suggest one. Keep figures exactly as given. Plain English, under 250 words.
Table: (paste the figures here) Notes from budget holders: (paste here)
The “Explanation needed” rule matters most. It turns the model’s tendency to supply a plausible reason into a visible gap that someone must fill. Prompt engineering for business explains the technique, and checking AI output covers the review.
A practical rhythm for a finance team
- Daily: document extraction into drafts, with exceptions reviewed the same day.
- Weekly: reconciliation suggestions reviewed and posted by a person; debtor reminder drafts.
- Month-end: first-draft commentary from final figures, checked against the ledger and with budget holders.
- Quarterly: a VAT checklist, and a reconciliation of the VAT accounts to the return, recomputed by hand.
- Ongoing: shared, tested prompts in a team prompt library, and a log of who reviewed what.
If extraction and reconciliation run at volume, they are better built into your accounting system than done in a chat window. From prompts to workflows and choosing accounting software cover that step.
Frequently asked questions
Can AI prepare our VAT return? It can help prepare the working papers, such as extracting invoice data and drafting checklists. Every figure in the return should be recomputed from the ledger, and every tax treatment checked against the law and Tax Authority guidance, by a person.
Is it safe to put client ledgers into an AI tool? Only into a business tool your firm has approved for client data, on terms that exclude training on your inputs and set out retention and location. Public AI tools are not the place for confidential client information.
Does using AI change our professional responsibility? No. As ICAEW’s guidance puts it, generative AI tools “do not make up for a lack of professional knowledge and skill”. The accountant who signs remains responsible for the work.
How do we keep AI-assisted work auditable? Keep the prompt, the output and a note of who reviewed it with the working papers, and retain them as long as the records they support. For VAT, that is ten years after the end of the relevant tax year.
Which tasks should a finance team start with? Document data extraction into drafts, reconciliation suggestions, and first drafts of commentary and client emails. All three save time and are easy to check.
Will AI replace junior accountants? The International AI Safety Report 2026 found no effect yet on overall employment, but some signs of declining demand for early-career workers in some AI-exposed occupations. Firms should plan how juniors will learn judgement if routine tasks move to AI, rather than assume the work simply disappears.
This article states the position under the laws in force on 28 October 2026, the date of publication. It is not tax, accounting or legal advice.
Sources: ICAEW — Generative AI and ethics · VAT Law, Royal Decree 121/2020 — Oman Tax Authority · Oman Tax Authority — E-invoicing · Royal Decree 6/2022 — Personal Data Protection Law (Arabic text) · International AI Safety Report 2026 — executive summary
Want AI in your finance processes, with controls an auditor would accept? ZAWAT builds AI and automation and business systems with draft-and-approve steps, validation rules and audit trails designed in. Book a call.