What is an AI bookkeeper?

The short definition

Bookkeeping is the recording of a business's financial transactions. An AI bookkeeper does that recording with the help of a large language model: you describe what happened, or hand it a document, and it proposes the accounting entry — which account is debited, which is credited, for how much, with what tax.

The word *proposes* matters. The better tools treat the model's output as a draft that a person approves. The model is good at reading messy inputs (a crumpled receipt, “paid the electricity bill yesterday”) and mapping them to structure; it is not an authority on your tax position.

What an AI bookkeeper does

• **Interprets plain language.** “Received 3,000 SAR from Al Noor for invoice 12” becomes a receipt against accounts receivable.
• **Reads documents.** Receipts, supplier invoices and statements are extracted into vendor, date, amounts and VAT.
• **Classifies.** It chooses accounts and categories, ideally reusing what you approved for the same vendor before.
• **Handles mechanics.** Currency conversion, VAT splits, due dates, duplicate detection.
• **Answers questions.** “How much did I spend on software this quarter?” — computed from the ledger, not guessed.

What it does not do

• It does not decide your tax treatment or file returns. Calculations such as a VAT summary or a Zakat estimate are inputs for you or your adviser.
• It does not see transactions you never record. Completeness is still your job (or your bank statement's).
• It is not infallible. Models misread digits and pick wrong accounts; review is part of the design, not a fallback.

AI bookkeeper vs “AI features” in accounting software

Many established accounting products add AI for suggestions inside a form-based workflow: auto-categorising a bank feed, suggesting a match. An AI bookkeeper flips the interface — conversation, upload and email are the primary ways in, and the form becomes the exception. Both can sit on proper double-entry books; the difference is who does the data entry.

How to judge one

1. **Is there a real ledger underneath?** Ask for a trial balance. If it can't produce one, it's a categoriser, not a bookkeeper.
2. **Do you approve entries?** Look for a draft/approve step and an audit log.
3. **Does it explain itself?** An entry you can't understand is an entry you can't check.
4. **Where does data go?** Which model provider, under what terms, and whether your data trains models.
5. **Does it know your region?** VAT rates, currencies and language you actually use.

How Qaid fits the definition

Qaid is an AI bookkeeper in exactly this sense: chat, receipt upload, an inbox email address and statement import feed drafts that you approve into a double-entry ledger. Every draft shows its reasoning, changes are logged, and the security page states what happens to your data. It calculates VAT and estimates Zakat; it does not file.