Reading clients' scanned invoices: what to set up
What a firm must set up before Achekam can read clients' scanned invoices and receipts, what happens without it, and why every journal stays a candidate.
By Achekam · · 4 min read · For Accountants & audit firms in South Africa, Nigeria, Kenya
A client's month of supplier invoices arrives in every possible form: PDFs e-mailed by suppliers, scans from the office copier, photos of till slips taken on a phone in a car park. A firm switching on document intake expects all of them to be read. Then the first batch of photos produces nothing but review items, and someone concludes the feature does not work.
It does work, but not without a piece of set-up that is easy to miss. This article explains the difference between documents Achekam can read as they are and documents it needs help with, what to configure, and what happens when nothing is configured.
Text files versus scans and photos
There are two kinds of document, and they look the same on screen.
Documents with text in them. Most PDFs produced by a supplier's accounting system contain the text itself, not just a picture of it. You can usually tell by trying to select a word in the PDF. These can be read into their fields directly.
Images of documents. A scan from a copier, a photo from a phone, or a PDF made by "printing" a scan contains only a picture. To read it, the picture has to be turned into text first. That step is optical character recognition, or OCR.
Achekam can read supplier invoices and receipts into their fields and propose a journal for each. It reads scans and photos only when an outside OCR service or an image-reading AI key is set up.
An outside OCR service or an image-reading AI key
There are two ways to give Achekam the ability to read images.
An outside OCR service. A dedicated text-recognition service is connected and used to turn images into text before the fields are read.
An image-reading AI key. Achekam's AI features go through one AI engine that can use Anthropic, OpenAI, Google Gemini, DeepSeek or xAI, using the firm's own API key first and the platform's key only where allowed. Keys are set by administrators under Settings, in "AI models, keys, routing & spend". A model that can read images can then read scanned documents. With no key at all, every AI feature says so and drafts nothing.
Which route to choose is a firm decision, often settled by what the firm already uses and what its policy on sending client documents to an outside provider allows. Make that decision deliberately, with whoever is responsible for the firm's data protection obligations, before switching anything on.
Where staff use Document Intelligence to have AI read a receipt, invoice or statement, Achekam re-checks totals, VAT and dates and flags any figure not in its own transcription. Restricted data classes, such as payroll and bank data, are refused unless allowed per purpose. That check is worth explaining to staff: the AI's reading is compared with the document, and differences are shown rather than smoothed over.
The review item raised when no engine is set up
If neither an OCR service nor an image-reading key is set up and a scan or photo arrives, Achekam does not guess. It raises a review item named OCR_ENGINE_NOT_CONFIGURED.
Recognise that item for what it is. It is not a problem with the client's document, and re-uploading the same file will produce the same result. It means the firm has not yet given Achekam a way to read images. The remedy is set-up, or, for a single urgent document, asking the supplier or client for a text-based PDF.
A set-up checklist
- Decide between an outside OCR service and an image-reading AI key.
- Agree the decision with whoever owns the firm's data protection policy.
- Have an administrator configure it.
- Upload a test batch that includes a phone photo and a copier scan.
- Confirm the fields were read and no OCR_ENGINE_NOT_CONFIGURED item was raised.
- Tell the bookkeeping team what changed.
VAT that is not printed stays empty
One rule applies however a document is read. VAT that is not printed on the document stays empty. Achekam does not work VAT out from the total.
That is the right behaviour. Whether VAT can be claimed depends on what the supplier's document shows, not on what the reader assumes the rate should be. A till slip with no VAT shown, or an invoice from a supplier who is not registered, should not produce input VAT. In South Africa, SARS's VAT 404 Guide for Vendors covers tax invoices, documentary proof and input tax limits among its topics. Nigerian and Kenyan rules on what an invoice must show differ, and in Kenya input VAT depends on electronic tax invoices; check the documents against your own country's requirements.
When the VAT field is empty, the reviewer checks the original. If VAT is shown and was missed, correct it. If it is not shown, leave it empty and, where the amount matters, ask the client for a proper tax invoice.
Every journal is only a candidate
Reading a document well does not post anything. Supplier invoices and receipts uploaded to Financial Intake become suggested postings. AI can suggest, but only a person approves and posts. Journals proposed from documents, bank lines or adjustments are only candidates, and AI is refused both the approval and the posting.
Two places hold the person's work. Low-confidence fields and possible duplicates wait in the Intake Review Queue, described in working the intake review queue. The candidates themselves are approved under FINANCIAL INTAKE > Posting Candidates, covered in posting candidates a person approves.
That design means set-up improves speed, not authority. Once images can be read, staff spend less time typing and more time checking, which is where an accountant's time is better spent.
Ask your administrator this week whether an OCR service or an image-reading key is configured, and if neither is, put the decision on the next partners' agenda.
General guidance only, not advice on your own circumstances. Tax and payroll rules change: check the current position with the authority or your adviser.
