Kaunt codes every invoice line — GL account, cost centre, VAT code, any dimension your customers use inside your platform, under your brand. You ship the feature. We run the AI.
Coding accuracy
>98%
Across GL accounts, cost centres and any customer-defined dimension. It improves the longer it runs.
Time to value
<24hours
From the first historical data upload to live AI coding predictions.
Setup
0rules
Historical data is the only input. Nothing to template, nothing to map.
—Embedded AI
Your interface. Your brand.
Our intelligence
Ship AI account coding as part of your own product. Kaunt's coding interface drops into your screens under your own branding, so your customers never leave your platform. No second login, no other vendor's logo.
- One iframe tagNothing to bundle and no SDK to keep up to date. Improvements ship without you releasing any frontend code.
- Keys never leave your serverAPI keys stay server-side. The browser only ever gets a short-lived, revocable token, scoped when it is issued and never widened.
- Whole workflowsNot just widgets — validation, preview and the audit trail come with the embedded UI.
—Live in 24 hours
Send historical data. We do the rest
Upload your customers' historical invoices and postings, and training starts on its own. There is nothing to configure, because the history is the configuration.
—Continuous learning
Every correction makes the next proposal better
When one of your users changes a code, that correction goes straight back into that customer's own model. The next invoice from the same vendor arrives already carrying the corrected code, at a higher confidence. No retraining request and no release on your side — the next invoice simply knows.
—Any format
An .xml, a .pdf, a photo.
One proposal.
Structured XML, PDFs, scans, phone photos, your own JSON — however the invoice arrives, you get the same coded lines back. Leave the format to us.
—Confidence you can act on
We tell you how confident we are. You decide what happens
Every proposal comes back with a confidence on each dimension it predicted — HIGH, MEDIUM or LOW. The model knows the difference between a line it has seen a thousand times and one it is guessing at, and it says so. What each band should mean in your workflow is your decision, not ours.
—Advanced postings
The difficult postings,
proposed too
Most invoice lines post to a single account. The exceptions are where the time goes. Kaunt proposes the split as well as the code — freight across the goods it relates to, a shared service across cost centres, a subscription across twelve months.
—Posting text
Posting text written the way your customers write it
Kaunt composes the posting text from the invoice line itself — the vendor, the description, the date, the amount — in each customer's own house style: their length, their word order, their abbreviations.
—A model per customer
Trained on their data. Only theirs.
Kaunt trains a separate model for each of your customers, on that customer's postings alone. Nothing is pooled — one customer's data never reaches another customer's model.
—Built API-first
One REST call. Nothing else required.
A standard REST API, documented the way your developers expect, with no SDK to adopt and no proprietary protocol to learn. Authentication, pagination and error handling behave exactly as your team expects.