The client provides merchant cash advances to small businesses. Every application arrives with a set of supporting documents, sent in by message, by email or through the client's portal.
The Problem
Each application involves a pile of documents in different formats: identity documents, application forms and bank statements from many different banks. Staff had to sort these documents, work out what each one was, and capture the transactions from every bank statement before a credit assessment could even begin.
The more applications came in, the more people the client needed to process them. Growth was tied directly to headcount.
Our Solution
Elucidate built four solutions that work together as one platform.
- Document intelligence. Incoming documents are automatically classified into one of many categories, such as an ID, an application form or a bank statement, using a combination of machine learning and LLMs. Each document is then renamed in a consistent, readable format. A bank statement, for example, is named after the bank and the date range it covers.
- Bank statement processing. A dedicated pipeline learns the layout of each bank's statements and extracts transaction data with 99% accuracy.
- Transaction categorisation. Every extracted transaction is categorised, turning raw statements into a clear picture of each business's finances.
- Credit decisioning engine. The categorised data feeds a smart credit decisioning engine that the client uses to make credit decisions.
The Results
- Processing time cut by 78%
- More than 10,000 applications processed every month, without adding overhead
- Bank transactions extracted with 99% accuracy
- Documents classified and named automatically, from every channel
- Credit decisions made on clean, structured, categorised data