As of 13 August 2026, AI can only partly create invoices from completed orders.
This still needs a person who signs their name to it.
Can you do it?
30 minutesto a draft.
1 hourto something you’d act on.
Cost, all in£0
Skill neededpower-user
Who has to check ita colleague
What the alternative costsThe supplied tool data gives no price for an invoicing alternative.
If this goes wrong, customers may receive invoices with incorrect items, totals or tax details and your records may need correction.
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, power-user skill, and roughly 1 hour until you can act on the result.
How to actually do it
- Open the order system and export the completed orders, including order references, dates, customer billing details, line items, quantities, prices, discounts, tax information, currency and payment status.
- Open your current accounting or invoicing system and gather the business name, address, registration details, invoice numbering rule, payment terms and approved tax settings.
- Remove cancelled, refunded and duplicate orders from the export, then check that customer and product data can be shared with the AI service under your organisation's privacy arrangements.
- Paste the prompt and the cleaned order data into an AI tool, replacing the bracketed business-rule section with your actual rules.
- Compare the drafted invoice fields and totals against the original orders and your accounting system, and resolve every flag without allowing the model to guess.
- Ask a colleague who understands your invoicing process to check the invoice sequence, customer details, tax treatment, totals and payment terms.
- Import or enter the approved invoices into the accounting system, keep the order references attached, and send them only after the system accepts the records.
Prompt
Create an invoice workflow from the completed order data below. Use only the supplied facts and do not invent customer details, products, prices, discounts, tax rates, payment terms or invoice numbers. First define the field mapping from each order field to the invoice. Then produce one draft invoice per completed order in a clean, copyable format and a validation table showing the order reference, customer, line items, quantities, unit prices, discounts, subtotal, tax treatment, total and payment terms. Flag any missing or conflicting value instead of guessing. Keep cancelled, refunded and duplicate orders out unless I explicitly include them. Preserve the currency shown in the data. Do not send, publish or mark any invoice as paid. Use these business rules: [insert invoice numbering rule, payment terms, registered business details and tax rules]. Here is the completed order data: [paste CSV, spreadsheet rows or API response].
Open it prefilled in ChatGPT or Claude, or copy it into Gemini, which takes no prefill link.
What it gets wrong
- AI cannot access your order and accounting systems unless you provide an authorised export or build and approve an integration.
- AI cannot know whether a particular UK sale has the correct VAT treatment from an order description alone.
- AI cannot take responsibility for invoice numbering, customer disputes, tax records or corrections.
- AI cannot safely decide how to handle refunds, partial fulfilments, mixed tax treatments or duplicate orders without explicit business rules and testing.
- AI cannot approve sending invoices on behalf of your business.
What caps this at PARTLY: private data access, verification cost and legal accountability.
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 create invoices from completed orders?
- Yes, it can turn supplied order data into draft invoices and validation tables. It cannot safely access your systems, determine missing business rules or take responsibility for sending incorrect invoices.
- Can AI automatically send invoices after an order is completed?
- It can help build an automation that does this, but the connection, permissions, exception handling and approval step still need to be configured and tested. Keep sending disabled until you have checked totals, tax treatment, duplicate handling and payment status.
- Can AI calculate VAT on invoices?
- It can apply VAT rules that you provide and show the calculation, but it should not decide the correct treatment for an unusual transaction from incomplete order data. Use an accountant for a serious or disputed VAT case, and treat any AI output as not professional advice.
- What information does AI need to create an invoice?
- It needs the completed order reference, customer billing details, products or services, quantities, prices, discounts, currency, tax treatment, payment terms and your business details. It also needs clear rules for invoice numbering and for excluding cancelled, refunded or duplicate orders.
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.