← All work Case study · AI agents & automation

Turning a chat message into an approved invoice

Case study · AI agents & automation · Twinslytics

A company keyed every vendor invoice into their ERP by hand. We built a bot that reads a message or a PDF, extracts the line items, and routes it for approval before it ever touches the ERP. Same idea works on QuickBooks, NetSuite, SAP, or whatever finance runs on.

The problem

This is one of the most common finance-ops gaps: every vendor invoice gets keyed in by hand — read the PDF, find the vendor, pick the right account, create the entry, get it approved. Multiplied across every bill that shows up in a month, it adds up to real accounting hours spent on data entry instead of finance.

What we built

A chat bot — we built this one for Slack, but Teams works the same way — that turns a conversation or an uploaded PDF into a purchase invoice in the ERP:

The result

Vendor invoices go from a PDF or a chat message to an approved ERP entry in one step instead of five, with an audit trail that didn't exist before. The exact ERP and chat tool change the integration, not the approach.

Chat message → approved ERP entry — one step instead of five

Still keying invoices in by hand?

We build the AI agents that read, decide, and act inside your existing tools — with every write gated behind a human confirmation.