Automation
AI-Powered Credit Report Automation
Finaccsol · Financial services
Producing the analysis of each credit report (a PDF of more than 13 pages) took about 30 minutes. We designed an AI-powered automated workflow that extracts the key data, generates an executive infographic and delivers it by WhatsApp or email in about 2 minutes.
The challenge
Finaccsol produces credit risk analyses from long reports, PDFs of more than 13 pages. Preparing each analysis took about 30 minutes of manual work, with a volume of around 1,000 reports a month.
What we did
- We designed an end-to-end workflow that turns each credit report into an executive infographic.
- We chose a self-hosted architecture (n8n, PostgreSQL and Redis) because of the client's budget and the privacy of its financial data.
- Requests arrive by email (payment receipts) or through a web form and are assigned automatically.
- A language model (Anthropic API) extracts and structures the key metrics, with extraction strategies that reduce token usage.
- A rendering engine built with QuickChart and Browserless generates custom infographics in real time.
- Delivery to the client is automatic, by WhatsApp and email.
Result
The time to produce each analysis dropped from about 30 minutes to 2 minutes.