As of 13 August 2026, AI can only partly categorise your business purchases.
This still needs a person who signs their name to it.
Can you do it?
5 minutesto a draft.
30 minutesto something you’d act on.
Cost, all in£0
Skill neededchat-fluent
Who has to check ityou
What the alternative costsA bookkeeper or accountant remains the alternative; no comparable price is supplied in the available data.
If this goes wrong: purchases are posted to the wrong categories or VAT treatment, and correcting the records may affect your accounts or tax reporting.
What to actually do
Hand it to a person
The route this page recommends
A person who owns the outcome does this end to end, worth it when the failure is dear.
Use a tool built for this
Second choiceDo it yourself
The distant thirdA chat interface, chat-fluent skill, and roughly 30 minutes until you can act on the result.
How to actually do it
- Open your accounting software and export the relevant purchase transactions as a CSV, including date, supplier, description, amount and VAT fields where available.
- Open your chart of accounts and copy the category names and any existing purchase-coding rules into a separate document.
- Gather the matching invoices and receipts, then redact unnecessary personal or payment details before copying their text or attaching the files.
- Add your business type, VAT registration status and any rules from your accountant to the business context section of the prompt.
- Paste the chart of accounts, business context and purchase data into the prompt and ask the AI to produce the categorised table and exceptions table.
- Compare every proposed category and VAT entry with the original invoice, receipt and chart of accounts, then correct rows where the supplier description does not establish the business purpose.
- Send the exceptions and any rows affecting VAT, capital expenditure, personal use or unusual purchases to your bookkeeper or accountant before importing the approved categories into your accounting software.
Prompt
Categorise the business purchases in the data below using the chart of accounts and rules I provide. Return a table with these columns: transaction date, supplier, description, amount, VAT amount if stated, category, confidence, and reason. Use only information in the supplied data. Do not invent missing details, VAT treatment, business purpose or categories. If a purchase is ambiguous, unusual, personal-looking, capital expenditure, mixed-use, or potentially subject to a special accounting or VAT treatment, put it in an exceptions table and explain exactly what I need to confirm. Keep the category names exactly as they appear in my chart of accounts. Do not delete, merge or alter transactions. At the end, list duplicate-looking or incomplete records separately. Chart of accounts and categorisation rules: [PASTE YOUR CHART OF ACCOUNTS AND RULES] Purchase data: [PASTE A REDACTED CSV, SPREADSHEET EXTRACT, INVOICE TEXT OR RECEIPT DATA] Business context: [STATE THE BUSINESS TYPE, WHETHER YOU ARE VAT REGISTERED, AND ANY RULES YOUR ACCOUNTANT HAS GIVEN YOU] Before producing the final table, state which fields or rules are missing and ask no more than five targeted questions if they prevent reliable categorisation. Otherwise proceed and mark those rows as exceptions.
Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.
What it gets wrong
- AI cannot know the business purpose of a vague transaction from a supplier name alone.
- AI cannot reliably decide the correct treatment of mixed-use purchases, capital expenditure or unusual VAT cases without facts and accounting judgement.
- AI cannot take responsibility for inaccurate accounts, VAT returns or tax reporting.
- AI cannot replace the audit trail created by retaining the original invoice, receipt and reason for each categorisation.
What caps this at PARTLY: judgement under ambiguity, verification cost and stakes of error.
How we scored this
Five axes, each scored nought to two by hand: ten means AI carries the task cleanly, and the thresholds that turn a total into YES, PARTLY or NO are published in the methodology. Each axis name links to its definition.
| Axis | Score (0–2) |
|---|---|
| Output | 2 |
| Inputs | 1 |
| Verification | 1 |
| Liability | 1 |
| Effort delta | 2 |
| Total | 7 / 10 |
The methodology and its thresholds are published in full.
FAQ
- Can ChatGPT categorise my business expenses?
- Yes, it can sort transaction descriptions into your chart of accounts and flag uncertain rows. It cannot reliably infer the business purpose or VAT treatment of an ambiguous purchase, so you must check the output before using it in your accounts.
- Can AI categorise expenses from bank statements?
- It can categorise the descriptions and amounts in a bank-statement export, especially when you provide your chart of accounts and prior coding rules. Bank descriptions often lack the invoice details needed to confirm business purpose, VAT and mixed-use treatment.
- Is it safe to use AI for bookkeeping?
- It is suitable for a first pass and for finding missing or duplicate-looking records, but not as an unchecked posting system. You remain responsible for the records and should refer uncertain VAT, capital expenditure and personal-use items to a bookkeeper or accountant.
- What is the best AI tool for extracting invoice data?
- Parseur is designed to pull data out of invoices, emails and PDFs, which can give you cleaner input for categorisation. It extracts information rather than replacing your accounting judgement, so the resulting fields and categories still need checking.
Nearby answers
Assessed by gpt-5.6-luna (gpt-5.6-luna) on 2026-08-13, second-checked by an independent model. Wrong somewhere? Email [email protected] and it gets re-checked.
The newsletter
AI news, new answers and product picks, straight to your inbox.